INDEPENDENT RESEARCH ARM Holdings 7th December 2015 Cash me if you can Semiconductors Fair Value 1310p (price 1,122p) BUY Coverage initiated
Bloomberg ARM.LN ARM Holdings designs and licences processors used by a large Reuters ARM.L majority of industry players. The group was one of the main winners 12-month High / Low (p) 1,205 / 848.5 of the smartphones wave such that the share price has rocketed from Market capitalisation (GBPm) 15,766 Enterprise Value (BG estimates GBPm) 14,799 75p in 2009 to more than 1100p these days. As we go into 2016, Avg. 6m daily volume ('000 shares) 4,640 questions primarily concern sources of future growth. We have Free Float 79.4% identified in particular, network infrastructure, servers, IoT and the 3y EPS CAGR 17.3% Gearing (12/14) -44% increasing adoption of ARM technologies enabling a rise in the Dividend yields (12/15e) 0.66% royalty rate per chip. In all, we have factored in average EPS growth of 16% for the next three years and value the share at 1310p (upside of YE December 12/14 12/15e 12/16e 12/17e 18%). Revenue (GBPm) 795.20 965.82 1,104 1,226 EBITA GBPm) 400.4 498.1 559.6 634.9 Op.Margin (%) 50.3 51.6 50.7 51.8 Further potential in smartphones! Although we have noted a Diluted EPS (p) 24.12 30.21 33.95 38.89 slowdown in smartphone sales momentum, we believe the group could EV/Sales 18.99x 15.32x 13.08x 11.45x benefit from 1/ moves upscale in processors (from 32-bits to 64-bits), 2/ EV/EBITDA 34.6x 27.3x 23.6x 20.3x EV/EBITA 37.7x 29.7x 25.8x 22.1x the multiplication of cores and 3/ the adoption of ARM's Mali graphics P/E 46.5x 37.1x 33.0x 28.9x technology. In all, we expect average growth potential of 11% between ROCE 36.6 46.7 53.2 62.5 2014 and 2018e in royalty revenues from smartphone processors.
Fresh sources of growth at hand! Whereas the market is focusing on 1215.7 momentum in global smartphone sales and is waiting impatiently for the 1165.7
1115.7 emergence of servers based on ARM's processors, we believe that 1065.7 network infrastructure offers more opportunities for the group in the 1015.7 short term. Significant changes are underway in usages and technologies 965.7
915.7 used in this segment and these are clearly beneficial to ARM. In contrast, 865.7 we do not expect the servers segment to be visible before 2018, when we 815.7 expect it to genuinely take off. 765.7 04/06/14 04/09/14 04/12/14 04/03/15 04/06/15 04/09/15 04/12/15 ARM HOLDINGS SXX EUROPE 600 Unrivalled value creation. In historical terms, the group has proven that it has created a sufficiently flexible business model to generate FCF of close to 100% of net profit on a stable basis. With an ROIC/WACC differential of almost 18 points on average over the past five years, ARM's business model clearly generates value for its shareholders. As such, we believe that with a profile such as this and 12m forward P/E of 33.3x, i.e. close to five-year low levels, the share harbours upside potential. Our valuation (DCF and peer comparison) points to upside of more than 18%.
Analyst: Sector Analyst Team: Dorian Terral Richard-Maxime Beaudoux 33(0) 1.56.68.75.92 Thomas Coudry
[email protected] Gregory Ramirez
r r
ARM Holdings
Simplified Profit & Loss Account (GBPm) 31/12/12 31/12/13 31/12/14 31/12/15e 31/12/16e 31/12/17e 31/12/18e Revenues 577 715 795 966 1,104 1,226 1,364 Change (%) 17.3% 23.9% 11.3% 21.5% 14.3% 11.1% 11.3% Adjusted EBITDA 282 445 437 543 611 692 781 Adjusted EBIT 263 351 400 498 560 635 718 EBIT 208 153 309 405 450 520 589 Change (%) 39.6% -26.3% 101% 30.9% 11.3% 15.4% 13.4% Financial results 13.6 13.2 11.0 12.1 13.2 13.5 13.6 Pre-Tax profits 221 163 317 412 458 527 598 Tax (71.4) (73.4) (68.6) (82.1) (91.7) (97.3) (110) Net profit 161 105 255 340 379 443 503 Adjusted net profit 205 291 343 428 481 551 622 Change (%) 19.7% 41.7% 17.9% 24.9% 12.4% 14.5% 12.8%
Cash Flow Statement (GBPm) Depreciation & amortisation 18.8 94.3 36.6 44.9 51.3 57.0 63.4 Change in working capital (80.7) 33.2 (45.5) 7.6 6.1 5.4 6.1 Operating cash flows 157 315 342 476 526 601 681 Capex, net (25.9) (45.3) (30.4) (38.6) (38.6) (42.9) (47.8) Free Cash flow 131 270 311 437 488 558 634 Acquisition, net (7.5) (24.8) (12.8) (3.5) 0.0 0.0 0.0 Financial investments, net 8.8 (3.4) (2.8) (0.69) (5.5) (6.1) (6.8) Dividends (51.8) (68.9) (86.1) (104) (134) (166) (172) Issuance of shares 8.3 5.9 (60.1) (42.2) 0.0 0.0 0.0 Issuance (repayment) of debt (3.3) (4.4) (7.6) (2.7) 0.0 0.0 0.0 Other (50.9) 28.8 (57.1) 10.3 13.2 13.5 13.6 Net debt (381) (584) (668) (967) (1,329) (1,729) (2,197) Balance Sheet (GBPm) Tangible fixed assets 36.1 33.6 43.4 24.1 2.0 (22.5) (49.8) Intangibles assets & goodwill 531 609 644 661 672 684 698 Investments 164 149 221 221 225 229 234 Company description Deferred tax assets 84.0 72.2 64.8 64.9 64.9 64.9 64.9 ARM is a UK group specialised in the Other non-current assets 2.0 1.6 1.7 1.7 1.7 1.7 1.7 Cash & equivalents 386 588 675 969 1,331 1,731 2,199 design of processor architectures and Current assets 264 185 187 227 259 287 319 graphic chips. Virtually all Total assets 1,467 1,638 1,837 2,169 2,555 2,976 3,467 semiconductor players are licencees of Shareholders' equity 1,206 1,311 1,528 1,823 2,172 2,558 3,012 the group and use the designs Provisions 24.2 45.1 45.6 45.6 45.6 45.6 45.6 Deferred tax liabilities 16.6 18.9 32.3 32.3 32.3 32.3 32.3 developed by ARM to help them Current liabilities 214 259 225 265 304 337 376 design their own chips. ARM has L & ST Debt 5.8 4.2 6.5 2.0 2.0 2.0 2.0 benefited massively from the boom in Total Liabilities 1,467 1,638 1,837 2,169 2,555 2,976 3,467 smartphones, 85% of which use Capital employed 826 728 860 857 843 830 815 processors based on an ARM Ratios architecture. To continue expanding, Gross margin 94.83 94.79 95.52 96.10 96.20 96.30 96.60 Adjusted operating margin 45.58 49.10 50.35 51.57 50.70 51.80 52.60 the group now needs to develop new Tax rate 32.32 45.14 21.67 19.92 20.01 18.45 18.36 growth sources including IoT, servers Adjusted Net margin 35.56 40.68 43.10 44.32 43.60 44.97 45.56 and networking infrastructure. ROE (after tax) 13.32 7.98 16.71 18.67 17.43 17.33 16.70 ROCE (after tax) 21.58 26.98 36.58 46.69 53.24 62.55 72.04 Gearing (31.55) (44.51) (43.73) (53.02) (61.18) (67.56) (72.94) Shareholders Pay out ratio 31.78 65.09 33.36 30.45 35.05 37.11 34.02 Number of shares, diluted 1,396 1,412 1,421 1,417 1,417 1,417 1,417 BlackRock Baillie Gifford 5.0% 5.1% Thornburg Data per Share (p) Investment EPS 11.68 7.50 18.16 24.21 26.92 31.52 35.76 5.0% Restated EPS 14.73 20.59 24.12 30.21 33.95 38.89 43.86 % change 18.3% 39.8% 17.1% 25.2% 12.4% 14.5% 12.8% FMR BVPS 0.86 0.93 1.08 1.29 1.53 1.81 2.13 4.9% Operating cash flows 0.11 0.22 0.24 0.34 0.37 0.42 0.48 FCF 0.09 0.19 0.22 0.31 0.34 0.39 0.45 Treasury shares Net dividend 3.71 4.88 6.06 7.37 9.43 11.70 12.17 0.6% Other free float 79.4% Source: Company Data; Bryan, Garnier & Co ests.
2
ARM Holdings
Table of contents
1. Investment Case ...... 4
2. A specific business model ...... 5
2.1. Snapshot: digital architect ...... 5
2.1. Licence to kill… competition ...... 7
2.2. Royal(ties) family: processors for everything! ...... 9
3. What sources of fresh growth for ARM? ...... 11
3.1. Enterprise infrastructure: where are the servers? ...... 11
3.2. Network infrastructure: connection underway! ...... 15
3.3. Mobile: adjusting the mix to underpin growth ...... 17
3.4. Internet of Things: ARM's "one more thing" ...... 23
4. Ultimately, a genuine cash cow? ...... 25
4.1. Scenario: 2014/18e CAGR of 16% in EPS ...... 25
4.2. Free cash cow ...... 28
4.3. A professional in value creation ...... 28
4.4. GBP717m returned to shareholders over 10 years ...... 29
5. Upside potential to capture ...... 31
5.1. We have adopted a Buy recommendation ...... 31
5.2. DCF valuation of 1,319p ...... 31
5.3. Peer comparison of 1,302p ...... 33
5.4. Valuation on historical multiples of 1,272p...... 34
5.5. What risks/catalysts for the share? ...... 35
Bryan Garnier stock rating system...... 39
3
ARM Holdings
1. Investment Case
The reason for writing now As we go into 2016, the market is questioning the group's ability to find fresh sources of growth, thereby pushing the share price valuation (P/E) to levels 20% lower than the five-year historical average. Independent of smartphone sales, we believe that growth potential exists in value terms in this segment (improvement in royalty rate per chip) and that IoT, networking infrastructure and servers should rapidly help strengthen momentum.
Valuation Our Fair Value of 1,310p stems from the equi-weighted average between a DCF valuation (FV of 1,319p with WACC of 9.0%) and peer comparison (FV of 1,302p), which takes account of ARM Holdings' historical premium to peers. Based on our estimates, the share is trading on 2016e EV/EBIT of 25.5x and 2016e P/E of 32.7x.
Catalysts In our view prospective catalysts are 1/ the adoption of new ARM technologies in smartphones (ARMv8, big.LITTLE, Mali, octa-core…) boosting the royalty rate, 2/ the ramp-up in ARM chips in network infrastructure and IoT, and 3/ further out, market share gains in the servers segment.
Difference from consensus Our 2015/17e sales estimates are in line with market expectations. We are slightly more cautious than the consensus in terms of EPS (-1.5%).
Risks to our investment case Risks to our scenario are: 1/ a slower than expected ramp-up in network infrastructure, servers and IoT, 2/ a slower-than-expected adoption of new ARM technologies in smartphones, 3/ a strong macro-economic slowdown affecting volumes, especially smartphones and 4/ a hike in the GBP relative to the USD.
4
ARM Holdings
2. A specific business model 2.1. Snapshot: digital architect ARM is a UK group specialised in the design and improvement of micro-architectures, namely chip design. ARM focuses on research and does not produce the chips itself but sells the right to use its designs that are particularly well-liked by the industry and which serve as the basis for the design of a large share of digital chips sold today.
Fig. 1: ARM designs and licences processor architectures
Equipment goes into fabs IDM • Traditional manufacturer of Integrated circuits owned by IDM and Fabless • Design the chips (Integrated Device • Operate fabs Manufacturers)
• Design the chips Fabless • Outsource production of chips • Do not have fabs Fabless outsource production to Foundries • Only produce for clients : the Fabless Foundries • Usually do not realize the design nor the selling
Equipment • Semiconductor specific equipment manufacturers • Customers are IDM and Foundries Manufacturers
IP blocks are also used IP & EDA by IDM and Fabless to • IP : Design IP blocks accelerate IC design. (Electronic Design Automation & • EDA : Make and sell software design tools EDA tools are used to Intellectual Property) design ICs at IDM and Fabless.
Source: Bryan, Garnier & Co. Today, the lion's share of electronic equipment carries embedded software (a sequence of tasks to carry out that is sometime very basic). This software is executed by integrated circuits known as 1/ microcontrollers 2/ processors and 3/ SoC (System on Chip) when they are more advanced (housing several chips on a same component: several application processors, several graphic processors, memory, a 3G/Wifi/Bluetooth connection etc.). From a material viewpoint, the processors are designed to execute a set of instructions that are defined on conception (adding, subtracting etc.), while the material implementation (cabling) is known as micro-architectures or instruction set architecture (ISA).
ARM's business is to Even the simplest chips now need to be capable of executing a minimum amount of instructions, design the most complex such that their design has become increasingly complicated. The extent of this is such that when a IP blocks, especially new chip is designed, virtually no manufacturer develops the entire architecture itself, but uses pre- calculation processors designed and available blocks (memory, processors etc.) known as intellectual property blocks (IP (Cortex) and graphics blocks). ARM's business is to design the most complex IP blocks, especially calculation processors (Mali). processors (Cortex) and graphics processors (Mali).
5
ARM Holdings
In order to access and use Although these predesigned blocks are not available free of charge, chip manufacturers (Qualcomm, ARM's IP blocks, chip Broadcom, STMicroelectronics, Infineon…) readily use them since they save a huge amount of time manufacturers acquire and money (R&D costs). In addition, they ensure the compatibility between software that was licences. Thereafter, when developed for the architecture selected (vs. an architecture created internally and for which all the chip design is programmes need to be redeveloped). In order to access and use ARM's IP blocks, chip completed, they pay royalties on each of the manufacturers acquire licences. Thereafter, when the chip design is completed, they pay chips produced royalties on each of the chips produced (a rate on the unit selling price of the chip).
Fig. 2: ARM does not produce components, but markets an architecture and chip designs
Technology Collaboration
Chip
Chip design maker OEM basis Smart chip
IP block
Technology needs
Licence Fee Royalty per chip
Cash flow Knowledges Products
Source: Bryan, Garnier & Co.
ARM also accompanies chip manufacturers in the design process and up to the production launch with its physical IP division. The group is then involved in adapting its architecture to the manufacturer's various production technologies (process technologies).
In this report, we describe The calculation processor is the most critical part of a chip and also the most complex to ARM chips or ARM develop. This can be simplified by stating that there are currently two architectures available: 1/ processors. These are not Intel's x86, which is extremely complex but powerful, on which the majority of computer processors chips manufactured by are designed, and 2/ the ARM architecture, which is simple and lighter, on which virtually all ARM, but chips designed smartphone processors are based (smartphones’ SoC). on the basis of an ARM architecture and produced by the group's clients. Easier to manufacture and more energy efficient, the vast majority of chips designed on the basis of an ARM architecture are therefore used in smartphones, although they are also found in cars, industrial robots, bank cards, wearables and even a few entry-level laptop computers (notebooks).
ARM Holdings is split into four divisions but has six sources of revenue generation: 1/ the processor division (PD) with a royalties and licence section, 2/ the physical IP division (PIPD) also with a royalties and licence section, 3/ software & tools (design tools) and 4/ a services division (consulting, training etc.).
6
ARM Holdings
Fig. 3: Four businesses, six sources of sales generation
Processors Physical IP Software & Tools Services
Physical IP Software Royalties Services Processors Licencing 11.3% and Tools 4.8% 4.6% 80.1% 34% 4.0% Royalties Licencing 46% 6.4% % of group sales
Licences Royalties Licences Royalties 600 600 100 100 100 100
500 500 80 80 80 80
400 400 60 60 60 60 300 300 40 40 40 40
Revenue 200
(in USDm) 200 20 20 20 20 momentum FY12 FY13 FY14 FY12 FY13 FY14 FY12 FY13 FY14 FY12 FY13 FY14 FY12 FY13 FY14 FY12 FY13 FY14
- Cortex A (processor) - Development tools - Training services - Cortex R (processor) - Physical designs bridging IP (logic, memory, specialised for ARM - Maintenance Key - Cortex M (processor) interface, etc.) to hardware (wafer level) architecture - Consulting services products - Mali (graphics)
- Apple - Samsung - PD or PIPD - PD or PIPD - Qualcomm - Foundries Licensees Licensees - MediaTek - Small semiconductor companies - Systems companies - Systems companies - Nvidia - Systems designers - Systems designers (examples) … Key consumers - Nearly all major semiconductor companies
- Semiconductor - Increasing complexity of chips - Semiconductor - Momentum of creation of semiconductor start- Market and especially - ARM ecosystem expansion Market and especially Key ups ARM chips - Adoption of ARM based chip in new segment ARM chips momentum catalysts momentum
- Demand for consumer goods - Demand for consumer goods - ARM ecosystem - ARM ecosystem - Automotive market momentum - Automotive market momentum momentum momentum Strong sensitivity to
Source: Bryan, Garnier & Co. 2.1. Licence to kill… competition For the group, licences For a chip manufacturer, obtaining a licence from ARM is often the first stage. For the group, represent 40% of sales. licences represent 40% of sales. The reason the group is so successful and why virtually all semiconductors players have ARM licences lies in the fact that when a manufacturer decides to launch a new product (i.e. a new chip), the choice of the basic architecture is essential (see previous section).
Among the two architectures that we describe above, x86 is primarily reserved for desktop and laptop computer processors, and an ARM is therefore almost systematically chosen for other chips. ARM is not the only option vs. x86 (there is also MIPS architecture for example), however, ARM's architecture boasts a highly developed ecosystem and hence a wide range of 1/ processor designs, 2/ development and design tools, 3/ qualified engineers and support services and 4/ third party software. As a result, ARM based processors for tablets and smartphones are widely adopted today.
7
ARM Holdings
Fig. 4: All major semis players are licensees
Processor Some public licensees Cortex-A57 AMD, Broadcom, HiSilicon, STMicroelectronics, Samsung, MediaTek, Huawei Cortex-A53 AMD, Broadcom, Samsung, Altera, STmicroelectronics, MediaTek, Qualcomm, Xilinx Cortex-A17 VIA, MediaTek, Realtek, Rockchip Cortex-A15 Texas Instruments, ST-Ericsson, nVIDIA, Samsung Electronics Broadcom Corporation, Freescale, NEC Electronics, nVIDIA, STMicroelectronics, Texas Cortex-A9 Instruments, Toshiba, Mindspeed Technologies, ZiiLABS, Open-Silicon, eSilicon, Altera, Xilinx Broadcom Corporation, Freescale, Panasonic, Samsung Electronics, STMicroelectronics, Texas Instruments, Cortex-A8 PMC-Sierra,ZiiLABS Cortex-A7 Broadcom, Freescale, Fujitsu, HiSilicon, LGE, Samsung, STEricsson, Texas Instruments Cortex-A5 Cambridge Silicon Radio, Open-Silicon, eSilicon Broadcom Corporation, Texas Instruments, Toshiba, Infineon, Open-Silicon, eSilicon, Samsung, Marvell, LSI, Cortex-R Fujitsu, Xilinx Cortex-M7 STMicroelectronics, Freescale, Atmel Cortex-M4 NXP, STMicroelectronics, Texas Instruments, Freescale, Open-Silicon, eSilicon Accent Srl, Actel Corporation, Broadcom Corporation, Cypress Semiconductor, Ember, Energy Cortex-M3 Micro, Fujitsu, NXP, Fuzhou Rockchip Electronics CO. Ltd., STMicroelectronics, Texas Instruments, Toshiba, Zilog, Open-Silicon, eSilicon Austriamicrosystems, Chungbuk Technopark, NXP, Triad Semiconductor, Melfas, Open- Cortex-M0 Silicon, eSilicon, Cypress, Infineon,Nuvoton, STMicroelectronics Cortex-M0+ Freescale, NXP, Atmel, Spansion, Silicon Labs
Source: Company data
We have identified five types of licences authorising the use of ARM technologies:
Architecture licences: These licences give the right to design processors which use the ARM instruction set. Licensees are in charge of their own implementation. This type of contract notably ties ARM to Apple, Qualcomm and Samsung.
Subscription licences: These provide access to a broad range of existing and future ARM technologies to design a specific product over the subscription period. We estimate the price of these licences can exceed USD40m.
Perpetual licences: They provide perpetual rights to design and produce chips on a selected range of ARM technologies. We estimate the selling price for the Cortex-A perpetual licence selling price is at between USD5m and USD10m.
Term licences: These licences provide access to a selected range of ARM's products and enable the design of chips for a limited period (generally three years, production is not limited). We estimate the Cortex-A term licence selling price at USD2.5-5m.
Per use licences: These provide the possibility of designing chips on the basis of a specific ARM product. As for subscription and architecture licences, the duration of design development is limited over time but chip production is not subject to restrictions as long as it uses the designs developed on the basis of the ARM technologies licences (as for others licences). We estimate the Cortex-A term licence price at between USD0.5m and USD1m.
For several years now and despite demanding comparison with previous years, the group has enjoyed healthy momentum in licensee recruitment. This stems from the diversity of ARM technologies requiring new licences (apart from under licence agreements). In all, the group's sales growth in processor licences has been maintained an average level of 18% over almost 10 years.
8
ARM Holdings
Fig. 5: Faster momentum in new licence signings and stable growth level in volume terms over 10 years at around 10-15%
180 30% 160 25% 140 120 20% 100 15% 80 60 10% 40 5% Number of new licences signed 20 0% YoY growth of cumulative processor licences FY00 FY01 FY02 FY03 FY04 FY05 FY06 FY07 FY08 FY09 FY10 FY11 FY12 FY13 FY14 New processor licences signed (net) YoY growth of cumulative processor licences
Sources: Company Data (licences signed include Mali licences); Bryan, Garnier & Co ests. In order to continue this momentum, ARM invests massively in innovation. In order to continue this momentum, ARM invests massively in innovation (R&D 21% of Usually, R&D costs for sales, around GBP200m in 2014) in order to improve its technology, like the ARMv8-A technology, one specific technology which provides new functionalities (including the handling of 64-bit instructions) relative to the are fully recovered from previous ARMv7-A (32-bit) technology. Usually, R&D costs for one specific technology are fully the 10th-12th licence of this recovered from the 10th-12th licence of this technology. technology. 2.2. Royal(ties) family: processors for everything! The group divides its processors into four units, Cortex-A, Cortex-R, Cortex-M and Classic. Each of these is destined and designed for specific applications apart from the Classic segment which houses previous architectures still in distribution, but slowing rapidly in favour of the other three segments (90% in 2010, around 30% in Q3 2015).
Royalty units Cortex-A processors are capable of executing complex systems such as mobile OS and therefore target applications that require a maximum amount of embedded intelligence (primarily smartphones and tablets). The different Cortex-A designs help choose the right Classic balance between performance and energy saving. A number of these, such as Cortex-A53, A7 Cortex-M Family 33% 43% and A5 can function alone or in combination with their more powerful peers such as Cortex- A72 and Cortex-A57 based on the principles of the big.LITTLE technology (energy efficient Cortex-A processors manage the majority of tasks and powerful processors are activated during important Family 17% tasks such as games on a smartphone). On average, we estimate that the royalty rate on Cortex-R Family Cortex-A chips stands at around 2.5% bearing in mind that chips are generally sold at a 7% Source: ARM (Q3-15) high average price (USD5-20).
Cortex-R are destined for products requiring a high level of reliability and significant calculation power such as microcontrollers used in cars, trains, industrial robots, aircraft etc.
Finally Cortex-M are processors built with the aim of reducing energy consumption as far as possible and therefore have limited computing power but enough for numerous applications (Bluetooth chips, audio, bank cards…). Although they account for a large share of volumes, the low value of the chips makes them fairly unprofitable. In this case, the group's strategy is one of volumes rather than value added.
9
ARM Holdings
Fig. 6: A range of 17 different processors, Cortex-A remain at the heart of ARM's portfolio
Family Cortex-A
Category High Efficiency Balanced High Performance Cortex- Cortex- Cortex- Cortex- Processor Cortex-A9 Cortex-A72 Cortex-A57 Cortex-A17 Cortex-A15 A53 A35 A7 A5 ARMv8-A ARMv8-A ARMv7-A ARMv7-A ARMv7-A ARMv8-A ARMv8-A ARMv7-A ARMv7-A Instruction set (64/32-bit) (64/32-bit) (32-bit) (32-bit) (32-bit) (64/32-bit) (64/32-bit) (32-bit) (32-bit) DMIPS*/Mhz 2.3 - 1.9 1.6 2.5 7.4 4.7 4.0 3.5
Example of 4K Smart- DVD Automotive High-end Mid-range Low-end 4K Set-top Embedded Routers devices/usages TV player Infotainment Smartphone Smartphone Smartphone box
Family Cortex-R Processor Cortex-R7 Cortex-R5 Cortex-R4 ARMv7-R ARMv7-R ARMv7-R Instruction set (32-bit) (32-bit) (32-bit) DMIPS*/Mhz 3.8 2.5 2.5 Example of usages Automotive ABS Hard-drive Printer
Family Cortex-M
Processor Cortex-M7 Cortex-M4 Cortex-M3 Cortex-M0+ Cortex-M0 ARMv7-M ARMv7-M ARMv7-M ARMv7-M ARMv7-M Instruction set (32-bit) (16-bit) (16-bit) (16-bit) (16-bit) DMIPS*/Mhz 3.2 2.0 1.9 1.4 1.3 Example of usages High-End audio headset High-perf. motor control Basic Smartband Bluetooth chip Touchscreen controller
*Dhrystone Millions of Instructions per Second Sources: Company data, Bryan, Garnier & Co. ests.
This very broad range of products leads to an inconsistent comparison in royalty value per chip over time. Indeed, ARM is booking more and more royalty revenues from low priced microcontrollers vs. high priced application processors for smartphones. In addition, we estimates that the royalty rate for a Cortex-A is close to 2.5% whereas it is closer to 1% for Cortex-R and Cortex-M. On a comparable basis, royalty per chip increases over time in our view.
10
ARM Holdings
3. What sources of fresh growth for ARM? Since 2008, the group has capitalised on the quality of the platform it has developed and has widely benefited from the advent of smartphones. As we approach 2016, the mature smartphones market and the question of credible sources of fresh growth are now an issue for ARM. For several years, the group has been announcing its entry into the servers market, but potential customers have so far been reticent to make public announcements about their intention to adopt ARM based servers.
We believe that ARM is closer than ever to its goal. Recently, we have finally seen announcements for ARM platform servers capable of upsetting Intel's hegemony. We also believe that the group is capable of capitalising on its server chips to strengthen its positions in the network infrastructure segment, currently in the throes of change. In addition, although the smartphones market is slowing, we believe ARM has the tools necessary for continuing its growth thanks to an improvement in its mix. Finally, the new craze for connected objects should offer the group an efficient source of fresh growth, although it only accounts for low volumes at the moment.
3.1. Enterprise infrastructure: where are the servers? ARM's customers’ platforms target specific workloads or what some prefer to call the “scale-out” market, namely datacentres whose architecture enables a ramp-up by adding similar machines, generally based on modest servers, contrary to centralised architectures where the ramp-up stems from improvements to the machines deployed.
3.1.1. The market is ready to welcome ARM’s solutions The servers segment has changed enormously over the past two decades. In the early 1990s, Intel had virtually zero market share in servers but now boasts a virtual monopoly in the segment.
Although Intel has already managed to take the market by storm and boasts dominant positions, we consider that a number of factors are not as beneficial for ARM today as they were for Intel 20 years ago. At the time, the market was fragmented with numerous manufacturers and processor references, all based on a simplified RISC architecture (different to Intel's x86 architecture). Intel has managed to dominate the competition thanks to 1/ its highly developed industrial facilities (processors for personal computers) enabling economies of scale and a technical edge in terms of production (350nm vs. 500nm), 2/ low competition in terms of sales and marketing clout, and 3/ rivals using a similar RISC architecture but with specific features meaning that applications have to be specifically developed for each platform.
ARM has been ARM has been communicating for a long time on servers adopting its technology and the first communicating for a long rumours even circulated in 2010 concerning ARM servers at Facebook. To our knowledge, the first time on servers adopting official server was launched by Calxeda in 2013 based on a quad-core Cortex-A9 (4MB L2). Not very its technology and Intel powerful, this had the advantage of working with very low energy consumption. Intel immediately immediately took the took the threat seriously and its response was the Atom S1260, a theoretically more powerful threat seriously processor, but which actually proved to be 20-40% slower while consuming double the amount of power (TDP of 15W vs. 8.3W). Although this processor, or SoC more precisely, was not convincing, ARM has nevertheless not managed to impose its platform.
11
ARM Holdings
If this is the case, we believe that the real competition for low-power servers stemmed more from the numerous virtual servers launched on physical servers based on Intel Xeon where energy consumption was shared. Moreover, it was at this time that virtualisation began to become more widespread in datacentres. To obtain similar performances to Intel's servers, ARM equipment therefore needed to be multiplied and the material cost therefore played against these configurations (even if they were less energy-consuming). Added to a limited RAM support (4Go), these configurations were therefore relegated to niche markets such as file servers.
ARM has developed its As we move into 2016, ARM and Intel have both muscled up their offers: Intel has expanded its chips to such an extent portfolio in entry level products to meet needs for modest servers (scale-out), while ARM has that they are just as developed its chips to such an extent that they are just as powerful as Intel processors for PC. powerful as Intel Now that the virtualisation wave is under control, the battle for the best material architecture is now processors for PC. beginning again.
3.1.2. ARM / Intel: David against Goliath? The servers market is currently totally dominated by Intel processors since AMD Opteron processors started to disappear from the landscape in 2009. To lay the foundations from the start, we believe that ARM processors clearly do not target the same market as Intel's upscale solutions (Xeon series 5000 and 7000 available at ASPs of between USD1,200 and USD7,200), but rather Intel's entry level range, in other words the Xeon series 3000 (c. USD420-550) and D (c. USD200-580).
At present, the Intel entry-level range is built as follows:
Xeon E3 processors, the characteristics of which are similar to the Core i5/i7 (quad-core/8MB L3/2 DDR3 channels) with the support of ECC and VT-d in addition (faster memory and enhanced virtualisation capabilities). The main defaults of these processors are still the 32Go RAM limit and a still high power dissipation (Thermal Dissipation Power) of around 45W.
Xeon D SoCs, which are chips destined specifically for datacentres with a scale-out architecture and hence built with modest servers. The Intel SoCs have a minimum TDP of 25W.
Atom C2000 SoCs, for which TDP stands at between 6W and 20W and which also embed all Intel has a broad portfolio the controllers necessary USB, Ethernet, SATA and others. Compared with the Xeon D, their but ARM has numerous power is far more limited and their software compatibility is restricted (no instruction sets for partners. virtualisation in particular). This does not prevent Intel from positioning its processors at a high price of around USD170 such that the processor kit/mothercard rapidly exceeds USD300.
At ARM, the catalogue is currently made up of chips produced by Cavium, AppliedMicro, Broadcom, TI, Marvell and AMD. While the offer remains very limited today, we believe this is only the start. As such, although a multitude of different Intel x86 server configurations exist at HP, Lenovo, Dell and consorts, we have only identified a few server references using ARM chips, including HP ProLiant (built around an Applied Micro SoC), Cirrascale RM1905D (Applied Micro), Softiron 64-0800 (AMD), Wiwynn LN1148-10SL (Marvell), Micac Datun (Applied Micro), as well as three Gigabyte references (Applied Micro, Cavium and Annapurna).
12
ARM Holdings
More recently, Qualcomm has also announced its solution, an SoC with 24 cores, which in Although the ARM offer turn is aiming to undermine Intel's hegemony and its Xeons in this market segment. The remains limited, it finally Qualcomm design is based on a main ARM SoC, but also FPGAs by Xilinx (to ramp-up calculation seems to have the necessary appeal for an power). Note that Field Programmable Gate Arrays (FPGA) are reprogrammable chips that can be increasing number of optimised for different types of calculation. OEM's to take interest and thereby offer a wider Although the ARM offer remains limited, it finally seems to have the necessary appeal for an panel of configurations. increasing number of OEM's to take interest and thereby offer a wider panel of configurations.
3.1.3. Hardware: Scale-out or scale-up, all seems to favour ARM As explained previously, we believe that the change in technologies and practices, especially virtualisation, have modified the traditional approach to rolling out a datacentre (scale-up), causing the emergence of a new approach (scale-out). In both cases, ARM boasts at attractive positioning.
Scale-up: this is the traditional deployment approach for a datacentre. In very basic terms, the idea is to roll out a server whose resources exceed current requirements in order to maintain leeway in terms of rising future requirements. Thereafter, when the limits of this equipment A new market trend is have been reached, the equipment (generally evolving) is improved either by replacing certain emerging aimed no longer components, or by adding additional material. However, in this configuration, a new market at changing the processors trend is emerging aimed no longer at changing the processors of a server for other more of a server for other more powerful ones but to add to its CPUs the chips dedicated to pure calculation, namely powerful ones but to add FPGAs (capable of embedding more than 100 calculation cores). We believe this is one of the to its CPUs the chips dedicated to pure most efficient strategies (costs/power/flexibility) for multiplying the calculation power of a calculation, namely server. It is indeed this rationale that triggered the latest acquisition by Intel, Altera, for FPGAs. It is indeed this USD16.7bn. However, these chips are mostly produced in an ARM architecture (Altera rationale that triggered the included). At present, the main players in this segment are Altera, Xilinx (ARM architecture) and latest acquisition by Intel, Kalray (proprietary architecture). At first glance, the risk is that Intel abandons the Altera Altera, for USD16.7bn. technology based on an ARM architecture, but we do not expect this to be the case. Firstly, because Intel has made it known that it would like to continue to develop this technology, and secondly, because the US group proved it could react on a very rational manner. Indeed, it has already undertaken a similar acquisition, namely Infineon's Wireless business, and has maintained an ARM architecture, judged more efficient, for its modems.
Scale-out: this approach is increasingly used and consists of associating several modest servers in order to create a calculation pool. Virtualisation now offers the possibility of making a multitude of physical machines work together without this being visible for users. These servers have modest unit performances, but once they are aggregated, they develop very high power. This architecture offers the advantage of 1/ flexibility, whether in terms of scalability or event management, and 2/ often better controlled consumption, sometimes at the expense of simplicity in start-up (virtualisation) and administration. However, by definition, a modest server is an appliance built around a chip with modest performances and ideally that consumes little energy and is therefore perfectly adapted to an ARM architecture.
13
ARM Holdings
3.1.4. Software: the ecosystem is in place During the 1990s, In order to establish itself in the servers market, Intel also benefited from the appeal of network administrators Windows, which gradually adapted itself to the world of servers. During the 1990s, the Windows NT welcomed and already duo and the success of Pentium Pro went down particularly well. Although we believe that the knew the Windows hardware offering for a server could rapidly change to the benefit of ARM chips, the question environment well, therefore concerns the software ecosystem. Indeed, while ARM servers are competitive and offer a whereas Intel processors offered good number of advantages on the material front (ideal power/consumption ratio), for some time already, performances we believe that the main brake to adoption lies in compatibility with software and we are indeed seeing things change gradually.
With increasingly high storage needs in order to meet BigData and Cloud requirements, servers specialised in storage are increasingly common. And as luck would have it, their usage is perfectly suited to ARM chips: servers available 24/7/365, which do not undertake heavy calculations but need to consume and heat as little as possible, take up a small amount of space (to free space for additional hard drive) and be capable of managing a large quantity of data. At the same time, various light ARM based chips are equipment destined for development or the consumer market (especially NAS) have been developed perfectly suited for Cloud on an ARM base for cost saving reasons. servers. This has prompted manufacturers This has prompted manufacturers of this equipment to gradually adapt software bricks to the to gradually adapt ARM devices such that today, the large majority of bricks necessary to build a server are software bricks for ARM compatible with ARM chips: from the famous Apache web server, to the OS Citrix and including devices. Ubuntu and MySQL. In this respect, we believe that the start-up phase, which is probably the most complex during the launch of an ecosystem, has already been validated.
3.1.5. However winning market share takes time Helped by a change in paradigm in hardware terms and an already installed software ecosystem, we believe that the base of installed ARM servers should gradually grow and the target of 25% market share by 2020 seems feasible to us. We are forecasting a servers components market worth USD20bn in 2020e, pointing to average annual growth of 6% on average. Paradoxically, this is a market where changes are fairly slow. As such, we believe the ramp-up in this business is likely to be gradual. For the next few years, we have estimated a scenario in which royalty revenues from server chips account for USD41m out to 2018e, or 4% of the group's overall sales and <9% market share (vs. ests. Fig. 7: Starting from zero, ARM should grow faster than the market Server market ARM Royalty revenues in Server 25,000 70 Servers’ royalty revenues CAGR 14/20e: 6% represent 4% of total 60 20,000 group sales in 2018e 50 15,000 40 A progressive ramp up heading to 9% market 30 10,000 share in 2018e. 20 market valueUSDm) (in 5,000 Royalty revenue (in USDm) 10 0 0 FY14 FY15e FY16e FY17e FY18e FY19e FY20e FY14 FY15e FY16e FY17e FY18e Server maket Server royalty revenues Source: ARM; Gartner; Bryan, Garnier & Co ests. 14 ARM Holdings To resume, we believe that this segment is complicated to penetrate. Whereas ARM now has all the right assets for success, we do not expect steady market share gains as of today, but rather a gradual rise in momentum followed by a strong ramp-up between 2018 and 2020. We have factored in the sensitivity of impacts on EPS of a lower or higher than expected market share in servers by 2020 as well as a lower or higher royalty rate. The details of these various scenarios are set out in section 5.5. 3.2. Network infrastructure: connection underway! Apart from servers, we believe another market exists that is fairly similar and in which ARM potentially has good opportunities, namely network infrastructure. This market is currently valued at USD15bn (electronic components for network equipment only). Based on the various projects of its clients, the group is forecasting market share of more than 45% by 2020 vs. around 10% at present. 3.2.1. Usages are changing and oblige infrastructures to strengthen With a rising number of connected devices, restrictions and requirements on networks are increasingly high, especially for mobile networks. According to Cisco, almost half a billion additional devices connected to the various global networks during 2014, bringing the total number of connected devices to 7.4 billion. For 2019, Cisco is forecasting 3.2 billion in various connected objects in addition to 8.2 billion mobile phones, or around 11.4 billion devices. Fig. 8: 2014/19e CAGR of 9% in connected devices 14.00 12.00 10.00 8.00 6.00 4.00 Connected devices (in Bn ofunits) 2.00 0.00 2014 2015e 2016e 2017e 2018e 2019e Non-smartphones Smartphones M2M Laptops Tablets Other Portable devices Source: Cisco In addition, changes in usages with the success of Cloud Computing and video on demand (Youtube, Netflix, Catch-up TV…) are prompting an even faster increase in data traffic. By 2019, Cisco forecasts that Cloud applications could account for 90% of data traffic vs. 81% in 2014. 15 ARM Holdings Fig. 9: Stress on networks is set to intensify, especially due to video World mobile data consumption World mobile data consumption per usage 30 30 24.3 25 25 20 20 Mobile File Sharing 16.1 15 15 Mobile Audio 10.7 Mobile Web/Data/VoIP 10 10 Exabytes per Month 6.8 Exabytes per Month Mobile Vidéo 4.2 5 2.5 5 0 0 2014 2015e 2016e 2017e 2018e 2019e 2014 2015e 2016e 2017e 2018e 2019e Source: Cisco This multiplication 1/ in connected devices and 2/ in data consumption requires the roll-out of more lasting infrastructure that is stable but above all flexible, open, programmable and evolving in order to accelerate future network expansion. In addition, power consumption is also an integral part of the issues encountered by operators. 3.2.2. The revolution in network infrastructure is an opportunity for ARM In response to these new In response to these new restrictions, historical networks are in the throes of change. Like the restrictions, historical increasingly virtualised datacentres, we believe that network infrastructure should also adopt a networks are in the throes virtualised structure in a cloud mode (distributed, scalable and redundant systems). of change. A classic network architecture is made up of several clusters of specialised devices (routers, load- balancers, gateways…) spread over and duplicated at multiple points of the networks. In order to simplify the structure and network administration, we note that operators and equipment makers are adopting three trends: We note that operators NFV (Network Function Virtualisation) aiming to replace specific network equipment by and equipment makers are applications launched on traditional servers. adopting three trends and we believe ARM should SDN (Software Defined Network), which consists of controlling the network routing (path fully benefit from them. taken by the traffic) via software in order to improve efficiency and control. Cloud-RAN (or C-RAN) is a third example of a strong trend within the networks. This is a breakthrough architecture, which takes a number of network elements currently coupled to RF modules (antenna) to a centralised site in order to facilitate their control, administration and reparation and to offer a more dynamic approach to available resources. All of the network modules function on a Cloud architecture, which is very close to a datacentre architecture on a hardware perspective. Operators are therefore in the process of adopting a new architecture and we believe that ARM should fully benefit since its technologies provide an answer to all the issues for operators: 1/ better steering of resources, 2/ improving QoS (Quality of Services), and 3/ improving network efficiency. 16 ARM Holdings From a material perspective, these trends are beneficial to non-specialised server clusters whereas at present, network equipment is primarily fitted with very specific high-performance chips known as Application Specific Integrated Circuits (ASIC). These are particularly costly since they are designed and manufactured on a customised basis in limited volumes. The majority of network software is currently executed on these specialised chips whereas the new uses have clear interest in executing them on traditional architecture chips in terms of: 1/ cost cutting, 2/ network flexibility, and 3/ control thanks to a centralised structure. For an operator, relative to ASIC, ARM processors offer the possibility of virtualising and making the various network bricks interoperable and connected (a gateway is therefore capable of informing a router of the state of the network it manages) since these bricks are all executed on a homogenous equipment architecture. In addition, we note that Asian operators view the arrival of ARM's solutions positively. Moreover it is quite feasible to imagine that the Qualcomm solution (described in section 3.1.2) could be used to build a C-RAN network. 3.2.3. A market worth USD18bn in 2020e (2014/20 CAGR of 3%) So far ARM has positioned itself in certain network segments, especially mobile Fronthaul (relay antennas), fixed Fronthaul (connection equipment right to the subscribed) as well as company network equipment. The group currently boasts market share of around 10% and is targeting 45% by 2020. We believe that things Although we expect market share gains in the servers market to remain difficult for ARM in should be easier in coming years, we believe that things should be easier in network infrastructure. Indeed, the network infrastructure group already boasts market share of 10% in this segment, the software ecosystem is in place than in servers. Indeed, and market segmentation is far greater. the group already boasts market share of 10% in this segment, the software In addition, we believe that operators are also tending to develop specific infrastructure for video, ecosystem is in place and which already represents a large majority of usages. As such, content distribution networks (CDN) market segmentation is far belonging to operators are seemingly positioned around the network core and are responsible for greater. distributing the video contents frequently consulted. We believe this is a significant opportunity for ARM. Indeed, this equipment is similar to data servers that route contents and therefore need to be energy efficient, which is the strong point of ARM's servers. As such, we have factored in market share of 37% out to 2018e, pointing to royalty revenues on network equipment chips of USD181m (vs. USD47m in 2015e). 3.3. Mobile: adjusting the mix to underpin growth 3.3.1. Market losing momentum Smartphone is the Since 2008, ARM has clearly benefited from robust growth in this market, thanks to smartphones in product category that has particular. This product category has enjoyed the fastest take-up by the mass market in enjoyed the fastest take- history. In comparison, it took more than 50 years for fixed-line telephony to exceed a penetration up by the mass market in rate of 50% in the US, 20 years for cars, refrigerators and mobile phones, but just seven years for history. smart phones to exceed this threshold, namely almost a third of the time (Asymco data). However, the However, the smartphones market is currently slowing. According to compiled data and IDC smartphones market is estimates, growth in the market stood at 75% in 2010 but is now close to zero. We believe this is a currently slowing. mechanical and expected trend. Indeed, current growth is penalised by disadvantageous comparison 17 ARM Holdings with previous years with a high penetration rate in developed countries (more than 60% of data compiled by GSMA). As such, while the market is currently witnessing a plateau in volume terms, a second wind is likely as of 2017 with the wide-scale equipment of a number of high-potential regions such as India, according to IDC expectations. Although volume growth is slowing, pressure on prices stemming from entry-level smartphones should automatically lessen. This has indeed been the case since early 2015, since the mix of low-end and high-end smartphones has stabilised. In value terms, growth therefore stems primarily from moves upscale during renewals, a trend that is likely to continue and which is beneficial to ARM. Indeed, top-end smartphones house more intelligent chips and thereby generate higher revenues for the group. Note also that although in historical terms Qualcomm was a good proxy for the health of the smartphones market and potentially ARM, we believe that this position is no longer true. Indeed, we have noted that OEMs are increasingly verticalising the design of application processors (AP) such as Apple and Samsung: Huawei with Kirin APs (developed in partnership with HiSilicon), Xiaomi is also working on its own AP based on work by Leadcore, while LG is developing its Nuclun AP. Similarly, following TSMC was historically a good indicator in the segment, but likewise, we have noted that Samsung is winning market share and is therefore making the reading of momentum in the sector more complicated. For ARM, the impact is high, especially on royalty revenues, but despite slowing volumes, the group is managing to maintain its growth. Fig. 10: Despite low momentum in smartphones weighing on ARM's momentum, the group is managing to maintain its growth 2.400 60% 2.200 50% 2.000 40% 1.800 30% 1.600 20% 1.400 10% YoY shipments growth Shipments (in m units) 1.200 0% 1.000 -10% FY10 FY11 FY12 FY13 FY14 FY15e FY16e FY17e FY18e Global mobile shipments (lhs) Mobile growth (rhs) ARM Mobile Processors royalites growth (rhs) Sources: Gartner; Bryan, Garnier & Co ests. Over five years, mobile In addition, the impact of this slowdown on momentum is partly cancelled out by the fact that shipments have dropped mobiles represent an increasingly small share of ARM's sales, particular in favour of connected objects from 62% to 45% and the corporate segment (network infrastructure and servers). The mobile segment now only accounts for 45% of ARM architecture chips shipments (volumes), whereas it accounted for almost two-thirds of chips delivered in 2010. 18 ARM Holdings Fig. 11: Over five years, mobile shipments have dropped from 62% to 45% 100% 90% 80% 70% 60% 50% 40% 30% 63% 62% 57% 53% 49% 45% % of total ARM shipments 20% 10% 0% FY09 FY10 FY11 FY12 FY13 FY14 Mobile Embedded Entreprise Home Sources: Company Data; Bryan, Garnier & Co ests. 3.3.2. Two tools for preserving growth In value terms, we estimate that the mobile segment accounts for 50% (around 40% for smartphones) of the group's royalty revenues. To offset weak momentum in volumes of devices manufactured, ARM has two tools: 1/ multiplying the number of ARM chips per device and 2/ a rising value of IP per chip helping to increase the royalty rate per chip. Multiplication in the number of chips per device: Whereas 10 years ago, a basic telephone only contained a single fairly basic micro-controller, today's smartphones include ever-more intelligent chips. Our research leads us to believe that in 2006, there were around 1.6 ARM chips per mobile on average, whereas this figure has now doubled. Fig. 12: Multiplication of chips designed on ARM architecture 7.000 3.3 3.2 6.000 Our research leads us to 3.1 believe that in 2006, there 5.000 3.0 were around 1.6 ARM chips per mobile on 4.000 2.9 2.8 average, whereas this 3.000 figure has now doubled. 2.7 Chips per phone (in units) Shipmentsbilion (in units) 2.000 2.6 1.000 2.5 FY10 FY11 FY12 FY13 FY14 FY15e FY16e FY17e ARM chip shipments (lhs) Global mobile shipments (lhs) ARM chips per phone (rhs) Sources: Company Data; IDC; Bryan, Garnier & Co ests. This has notably enabled ARM to outperform the market in volume terms for more than 10 years. We believe this trend should continue since we have noted a rise upscale in smartphones in 2015. In addition, we believe that moves to equip emerging markets where smartphone penetration rates remain low should also provide a favourable base. However, smartphones include more intelligent 19 ARM Holdings chips than classic mobiles, such that although on a comparable basis, the switch to a smartphone from a classic phone does not prompt an increase in mobile volumes, it does increase the number of chips on which ARM receives royalties. Rising value per chip: The chip market for mobile phones breaks into two sub-segments, application processors (AP), which are the heart of a smartphone (around USD15Bn) and other chips (around USD10bn). Combined, we estimate that these two sub-segments should witness average growth of around 6% between now and 2020, primarily driven by an increase in the value of APs (2014/20e CAGR of +9%). Meanwhile, ARM has a number of levers for improving its royalty rate per AP. We estimate the royalty rate on a chip using ARMv7 is close to 1.5% of the overall selling price of a chip. We believe that the group can increase this royalties rate to almost 5% thanks to 1/ the switch to the ARMv8 instruction set (+100bp - see section 2.1), 2/ the constant increase in the number of cores in SoCs and in particular, with the big.LITTLE technology, which helps couple powerful calculation cores with energy-saving cores on the same chip in order to balance the workload between these cores to benefit from substantial calculation power and maximise the device's autonomy (+30/50bp), 3/ the adoption of ARM's Mali graphic processor (+100bp), 4/ the addition of a connectivity core for Wifi, BT, 3G/4G … (+50bp) and 5/ the use of ARM's IP physical services. Fig. 13: Five levers for increasing the royalty rate per chip from 1% to 5% 5.0% 0.5% 0.5% 4.0% 1.0% 3.0% 0.5% 5.0% 2.0% 1.0% 1.0% 1.5% 0.0% Standard ARMv7 ARMv8 Multi-core with Additional Additional Physical IP Full service Royalty rate per chip (as% ofatotal chip price) based chip big.LITTLE Graphic radio core core Mali Total royalty rate per chip Incremental Processor royalty rate Incremental Physical IP royalty rate Source: Bryan, Garnier & Co ests. At present, we estimate that the average royalty rate for a mobile chip stands at around 2.5% bearing in mind that the adoption rate for the ARMv8 technology in smartphones should be close to 75% at the end of 2015. Virtually all APs for Concerning the multiplication of cores, we estimate that virtually all APs for smartphones now smartphones now work work on a multi-core basis. Note that among the highest-selling smartphones in H1 2015 (primarily on a multi-core basis but upscale), less than half were octa-core. On this point, we estimate that around 40% of the Cortex-A less than half are octa- delivered at present are octa-core. Note also that high-volume APs do not always include ARM's core big.LITTLE technology. This is a source of growth for ARM if the group manages to impose this technology. 20 ARM Holdings Fig. 14: The big.LITTLE technology has not yet been widely adopted in chips for the highest selling smartphones # of Rank Model Application processor big.LITTLE Architecture details core 1 Apple iPhone 6 Apple A9 2 (2x Custom) 2 Apple iPhone 6 Plus Apple A9 2 (2x Custom) 3 Samsung Galaxy S6 Samsung Exynos 7420 8 (4x Cortex-A57 + 4x Cortex-A53) 4 Samsung Galaxy S6 Edge Samsung Exynos 7420 8 Yes (4x Cortex-A57 + 4x Cortex-A53) 5 Xiaomi Redmi 2 Qualcomm Snapdragon 410 4 (4x Corex-A53) 6 Samsung Note 4 Samsung Exynos 5433 8 (4x Cortex-A57 + 4x Cortex-A53) 7 Apple iPhone 5S Apple A8 2 (2x Custom) 8 Samsung Galaxy S5 Samsung Exynos 7580 8 (8x Cortex-A53) 9 Xiaomi Redmi Note MediaTek MT6795 Helio X10 8 (8x Cortex-A53) 10 Microsoft Lumia 435 Qualcomm Snapdragon 200 2 (2x Cortex-A7) Sources: Top selling smartphones ranking by Counterpoint; Bryan, Garnier & Co. Our extended research Our extended research with the main AP manufacturers prompts us to conclude that less with the main AP than half of the new SoCs launched in 2015 use a big.LITTLE architecture. However, we see manufacturers prompts us healthy momentum in its adoption which prompts us to be optimistic concerning the technology, also to conclude that less than bearing in mind that the big. LITTLE technology fits perfectly with the multi-core trend. half of the new SoCs launched in 2015 use a big.LITTLE architecture. At the two main independent AP manufacturers, things are more nuanced. Only two MediaTek APs out of five launched in 2015 include this technology. At Qualcomm, the situation is similar also with two APs out of five using the big.LITTLE technology. Fig. 15: Less than half of the APs launched in 2015 use the big.LITTLE technology Release # of Application Processor big.LITTLE Micro-architecture Graphic* date core Apple A9 Q3-15 2 No 2x Custom (Twister) PowerVR Apple A9X Q4-15 2 No 2x Custom (Twister) PowerVR HiSilicon Kirin 930 Q1-15 4+4 No 4x Cortex-A53 (@2.0Ghz) + 4x Cortex-A53 (@1.5Ghz) Mali HiSilicon Kirin 950 Q4-15 4+4 Yes 4x Cortex-A72 + 4x Cortex-A53 Mali MediaTek MT8173 Q1-15 2+2 Yes 2x Cortex-A72 + 2x Cortex-A53 Power VR MediaTek MT6735 Q2-15 4 No 4x Cortex-A53 Mali MediaTek MT6753 Q3-15 8 No 8x Cortex-A53 Mali MediaTek MT6755 Q4-15 8 No 8x Cortex-A53 Mali MediaTek MT6797 Q4-15 2+4+4 Yes 2x Cortex-A72 + 4x Cortex-A53 (@2.0Ghz) + 4x Cortex-A53 (@1.4Ghz) Mali Qualcomm Snapdragon 415 Q1-15 8 No 8x Cortex-A53 Adreno Qualcomm Snapdragon 425 Q1-15 8 No 8x Cortex-A53 Adreno Qualcomm Snapdragon 618 Q1-15 4+4 Yes 4x Cortex-A72 + 4x Cortex-A53 Adreno Qualcomm Snapdragon 620 Q1-15 4+4 Yes 4x Cortex-A72 + 4x Cortex-A53 Adreno Qualcomm Snapdragon 820 Q1-16 4 No 4x Custom (Kyro) Adreno Samsung Exynos 7420 Q2-15 4+4 Yes 4x Cortex-A57 + 4x Cortex-A53 Mali Samsung Exynos 7580 Q2-15 8 No 8x Cortex-A53 Mali Samsung Exynos 8890 Q1-16 4+4 Yes 4x Custom (Mongoose M1) + 4x Cortex-A53 Mali ARM Technology / Total 7/17 9/17 * PowerVR is Imagination’s graphic technology, Adreno is Qualcomm, Mali is ARM. Sources: Companies data; Bryan, Garnier & Co. As such, this table, which sets out the processors launched in 2015 (non-exhaustive) and programmed for Q1 2016, also prompts us to believe that there are ultimately very few opportunities for ARM in graphic processors. Indeed, slightly more than 50% of these use the 21 ARM Holdings Note importantly that Mali graphic technology, although we do not see Qualcomm opting for Mali on the near term given these APs are to be that it has developed its own graphic technology Adreno internally (while we see it possible on the integrated into LT). Finally, the only large client remaining is Apple, which is nevertheless known for its loyalty and smartphones launched in its reticence to change, and currently uses PowerVR by Imagination Technologies for its graphic 2016 and thereby enable technology. Another negative point is that we believe Qualcomm could win further market share at ARM to increase its Samsung in 2016 and therefore reduce the size of Mali via a domino effect. royalty revenues during the next year. Note importantly that these APs are to be integrated into smartphones launched in 2016 and thereby enable ARM to increase its royalty revenues during the next year. 3.3.3. Mali saturated in smartphones... but is finding other sources of growth The main rivals in these ARM entered the graphic platforms segment in 2006 following the acquisition of Falanx, a Norwegian segments are: Imagination company that developed the first Mali chips (launched in 2005). Today, the Mali product family is the Technologies with leader in embedded and mobile graphic processing and we estimate that the group should benefit PowerVR, Qualcomm from growth in Mali chip volumes, which are set to exceed 700m units in 2015e (vs. 550m in with Adreno and Nvidia 2014 and <500m for the second player Imagination in 2014). with Tegra. As explained previously, for 2016, we are forecasting a beneficial effect on royalty revenues associated with Mali thanks to the announcement of new chips embedding ARM's graphics technology in 2015. ARM's graphics business works in a similar way to the processor business, namely with a licence segment and a royalties segment. However, it does not stand out as a business unit since the revenues generated by Mali are generally coupled with a processor business (indeed, a few rare exceptions exist such as Intel Atom X3 which includes an Intel processor and a Mali graphic part). Mali is therefore visible by adding basis points to the royalty rate generated by the chip. Note that we estimate that by adding a Mali graphic option, the royalty rate increases by 100bp on average. Fig. 16: Sharp historical growth pointing to healthy momentum over 2015/16/17e 250 1000 900 200 800 24 700 150 25 600 200 25 100 207 500 19 950 Licences (in unit) 17 400 850 50 300 550 50 500 200 400 0 FY11 FY12 FY13 FY14 FY15e FY16e FY17e 100 Shipments chip of with Mali graphic (munits) 150 0 45 Cumulative Licences New licences Estimated Cumulative Licences FY11 FY12 FY13 FY14 FY15e FY16e FY17e Sources: Company data at the end of Q3-15; Bryan, Garnier & Co. ests. Although ARM has so far won market share in the mobile segment as shown by the agreement signed last June with Samsung for the Exynos (which until then used PowerVR by Imagination Technologies), the group also boasts healthy momentum with Mali in other devices such as cameras since the addition of graphic calculation capacity enables new uses such as real-time editing of photos. Overall, we model a slowdown in additional volumes of Mali chips (>= +150M/year since 2013) 22 ARM Holdings from 2017e (+100M/year). Then, we believe that for Mali, the majority of growth as of 2017 is set to stem from IoT (see Fig 15 and comments in the table). The group has also developed video and screen controller chip technologies. Given that these technologies are very young, we have not factored them in as sources of fresh growth. 3.4. Internet of Things: ARM's "one more thing" The Internet of Objects (IoT) is now a clear trend. While the concept was still difficult to define 18 months ago, the industry now agrees in defining it as a range of connected objects with the exception of computers (PC, servers etc.), smartphones and tablets. 3.4.1. High growth potential The number of connected objects is set to rise from 10 billion at present to 30 billion in 2020, representing average annual growth of 15-20% over the period (source: Cisco). According to IBS, the components market for IoT could exceed USD40bn in 2020, or more than double the current market. Fig. 17: IoT: components market to double between 2014 and 2020e Source: IBS In our sector report published last June, we identified four key components for IoT: sensors, logic chips, connectivity chips and power management chips. 23 ARM Holdings Fig. 18: Building blocks of connected objects: four main component types Sensors Logic Connectivity Power Management Collecting real world data Process collected data and Transmit data to the cloud or Efficiently powering and such as motion or sound doing basic analysis master device (mainly short controlling energy supply of (analog data) (mostly 32bit MCUs) range RF) other components Example of European players Example of European players Example of European players Example of European players AMS ARM Dialog Semiconductor Dialog Semiconductor Infineon Infineon NXP Infineon STMicroelectronics STMicroelectronics STMicroelectronics STMicroelectronics All components working together with a strong integration Source: Bryan, Garnier & Co. ARM’s technology is currently the only one present in three out of the four components to a virtually systematic extent. Indeed, logic, connectivity and power management chips for IoT can be designed in an ARM architecture. Concerning logic chips, these are generally microcontrollers, in which we estimate ARM currently has a market share of slightly more than 25%. Penetration is slightly better in radio chips where market share is estimated at more than 50%. In these chips, rival architectures are often very basic older architectures, which have been sufficient until now but which are now showing their limits to ARM's great benefit. The main difference with For the group, the main difference with chips for smartphones or tablets is the selling price of chips for smartphones or components. In the majority of cases, these are components whose unit value is less than USD1. In tablets is the selling price comparison, we estimate that an AP for smartphone is sold for around USD15. As such, while IoT of components. offers a significant opportunity, ARM receives royalties on the basis of a percentage of chip selling prices and this is likely to rapidly cap opportunities. 3.4.2. For ARM, security is not an option, it is an opportunity to create value! We believe the group has We believe the group has identified this issue and is now looking to improve revenues per identified this issue and is chip by offering new functionalities such as security. now looking to improve revenues per chip by ARM's credo is that in order for Big Data to be efficient, we should above all trust Little Data. offering new However, in coming years, Little Data suppliers are primarily set to be connected objects. In addition, functionalities such as security. at each new major technology stage, we have noted that security is left to one side initially in favour of rapid technological change. ARM has therefore created the ARMmbed platform, which helps identify and certify the authenticity of the connected device and secure its connection with a server. In addition, in order to improve its skills in security, the group recently acquired a company specialised in equipment security and software for SosC destined for IoT. We forecast additional revenue of about USD3m out to 2018e, recorded in ARM’s Software & Tools division. 24 ARM Holdings 4. Ultimately, a genuine cash cow? 4.1. Scenario: 2014/18e CAGR of 16% in EPS Our top-line scenario takes account of the growth rate in the various market segments in which ARM's chips are sold, their respective growth rates, ARM's market share within these segments, as well as the royalty rate per chip (PIPD royalties et PD royalties). Fig. 19: 2014/17e CAGR of 14.4% in sales (in USDm except if stated) FY14 FY15e FY16e FY17e FY18e CAGR 14/18e Semiconductor SAM per segment (in USDbn) Application Processors 15.0 16.5 17.6 18.5 19.7 7.0% Networking Infrastructure 15.0 15.8 16.2 16.7 17.3 3.6% Servers 14.0 14.9 15.8 16.7 17.8 6.1% Embedded Intelligence 15.0 17.4 18.9 20.6 22.5 10.6% Automotive 10.0 11.4 12.2 13.0 14.0 8.7% Other Mobile Chips 10.0 10.1 10.1 10.1 10.1 0.2% Non-Mobile Application Processors 10.0 10.0 10.0 10.0 10.0 0.0% Chips into Other Markets 7.0 7.2 7.7 8.1 8.6 5.3% Total 96.0 103.2 108.5 113.9 119.8 5.7% ARM Market Share Application Processors 85% 85% 85% 85% 85% +0ppt Networking Infrastructure 10% 16% 21% 28% 37% +27ppt Servers 1% 2% 4% 6% 9% +8ppt Embedded Intelligence 25% 25% 25% 25% 25% +0ppt Automotive 5% 5% 5% 5% 5% +0ppt Other Mobile Chips 65% 65% 65% 65% 65% +0ppt Non-Mobile Application Processors 50% 50% 50% 50% 50% +0ppt Chips into Other Markets 40% 40% 40% 40% 40% +0ppt Royalty rate Application Processors 2.5% 2.6% 2.7% 2.8% 2.9% 3.5% Networking Infrastructure 2.0% 2.3% 2.5% 2.7% 2.8% 9.1% Servers 2.0% 2.2% 2.3% 2.4% 2.5% 5.8% Embedded Intelligence 1.0% 1.0% 1.0% 1.0% 1.0% 0.0% Automotive 1.5% 1.8% 2.0% 2.2% 2.4% 12.5% Other Mobile Chips 1.0% 1.0% 1.0% 1.0% 1.0% 0.0% Non-Mobile Application Processors 1.5% 1.8% 1.9% 2.0% 2.0% 7.5% Chips into Other Markets 1.0% 1.2% 1.3% 1.4% 1.5% 10.7% Royalties Processors 536 676 763 865 994 16.7% yoy change 8.2% 26.2% 12.9% 13.3% 14.9% Processor New licences signed 163 158 172 180 188 3.6% Revenue per licence signed (in USDm) 3.0 3.3 3.5 3.6 3.6 4.4% Processors Licences 497 529 600 640 681 8.2% yoy change 29.9% 6.5% 13.3% 6.8% 6.3% Penetration rate 40% 41% 41% 44% 44% 2.4% Physical IP royalty rate 0.5% 0.5% 0.5% 0.5% 0.5% 0.2% Royalties Physical IP 60 70 73 85 93 11.6% yoy change 6.6% 15.8% 5.4% 15.9% 9.6% Physical IP Licences 84 85 103 124 145 14.7% yoy change 28.5% 1.0% 21.6% 20.7% 16.9% Software and Tools 57 53 56 63 70 5.3% yoy change 0.4% -6.7% 5.1% 11.3% 12.7% Services 59 65 67 69 71 4.8% yoy change 9.3% 10.0% 3.0% 3.1% 3.4% ARM Revenue in USD 1293 1478 1662 1846 2055 12.3% yoy change 15.6% 14.3% 12.5% 11.1% 11.3% ARM Revenue in GBP 795 966 1104 1226 1364 14.4% yoy change 11.3% 21.5% 14.3% 11.1% 11.3% Sources: Company Data; Bryan, Garnier & Co ests. 25 ARM Holdings Fig. 20: Average EPS growth of 16% over the next four years in GBPm FY14 1Q15 2Q15 3Q15 4Q15e FY15e FY16e FY17e FY18e Licensing - Processors 309 71 82 81 108 346 398 425 452 seq change 26.5% -19.3% 15.8% -1.1% 32.5% 11.9% 15.1% 6.8% 6.3% yoy change 26.5% 2.0% 6.8% 9.0% 22.4% 11.9% 15.1% 6.8% 6.3% Licensing - Physical IP 52 16 14 12 13 55 68 83 97 seq change 26.5% 8.3% -12.8% -8.8% 4.0% 6.3% 20.0% 20.0% 15.0% yoy change 26.5% 38.1% 7.9% -10.1% -8.4% 6.3% 23.5% 20.7% 16.9% Royalties - Processors 326 110 103 121 107 442 507 574 660 seq change 2.7% 15.0% -6.9% 17.6% -11.1% 35.5% 14.7% 13.3% 14.9% yoy change 2.7% 43.2% 42.8% 47.7% 12.0% 35.5% 14.7% 13.3% 14.9% Royalties - Physical IP 37 11 11 11 12 46 49 57 62 seq change -10.5% 22.6% -3.5% 1.8% 2.1% 24.8% 15.0% 12.0% 5.0% yoy change -10.5% 15.2% 29.4% 27.3% 27.4% 24.8% 7.1% 15.9% 9.6% Software and tools 35 10 9 7 9 35 37 42 47 seq change -3.8% 5.5% -8.3% -15.9% 12.3% -0.1% 5.0% 2.0% 2.0% yoy change -3.8% -2.0% 10.0% -8.6% -2.4% -0.1% 6.8% 11.3% 12.7% Services 37 10 10 10 11 42 44 46 47 seq change 6.4% 4.3% 6.1% -1.9% 9.3% 15.9% 5.0% 6.0% 4.0% yoy change 6.4% 6.5% 13.0% 18.6% 16.3% 15.9% 4.6% 3.1% 3.4% Total Group Revenues 795 228 229 243 267 966 1,104 1,226 1,364 seq change 11.3% 0.7% 0.4% 6.4% 9.7% 21.5% 14.3% 11.1% 11.3% yoy change 11.3% 21.9% 22.1% 24.3% 18.1% 21.5% 14.3% 11.1% 11.3% Cost of Revenues -36 -10 -8 -9 -10 -38 -42 -45 -46 % of sales -4.5% -4.4% -3.7% -3.7% -3.8% -3.9% -3.8% -3.7% -3.4% Gross profit 757.4 217.5 220.1 234.0 256.6 928.2 1,061.7 1,180.4 1,317.9 Gross margin 95.5% 95.6% 96.3% 96.3% 96.2% 96.1% 96.2% 96.3% 96.6% Research and development expenses -165.1 -47.8 -51.1 -55.3 -62.4 -216.6 -253.8 -278.2 -307.0 Selling and marketing expenses -83.6 -21.6 -21.1 -23.6 -26.9 -93.2 -111.5 -117.7 -126.9 General and administrative expenses -110.5 -30.6 -27.1 -29.5 -33.1 -120.3 -136.9 -149.5 -166.4 Adjusted EBIT 400.4 117.5 120.8 125.6 134.2 498.1 559.6 634.9 717.6 % of sales 50.3% 51.6% 52.9% 51.7% 50.3% 51.6% 50.7% 51.8% 52.6% EBIT (IFRS) 309.1 101.5 92.3 102.0 108.8 404.6 450.3 519.7 589.4 % of sales 38.9% 44.6% 40.4% 42.0% 40.8% 41.9% 40.8% 42.4% 43.2% Adjusted EBITDA 437.0 117.5 120.8 125.6 134.2 543.0 610.9 691.9 781.1 % of sales 54.9% 51.6% 52.9% 51.7% 50.3% 56.2% 55.4% 56.5% 57.3% Interest income 11.0 3.0 3.1 2.8 3.2 12.1 13.2 13.5 13.6 % of sales 1.4% 1.3% 1.4% 1.2% 1.2% 1.3% 1.2% 1.1% 1.0% Adjusted Result before income taxes 411.4 120.5 123.9 128.4 137.4 510.2 572.8 648.4 731.3 % of sales 51.7% 53.0% 54.2% 52.8% 51.5% 52.8% 51.9% 52.9% 53.6% Income tax expense -68.6 -19.3 -20.3 -20.5 -22.0 -82.1 -91.7 -97.3 -109.7 Tax rate -16.7% -16.0% -16.4% -16.0% -16.0% -16.1% -16.0% -15.0% -15.0% Adjusted Net profit 342.8 101.2 103.6 107.9 115.4 428.1 481.2 551.1 621.6 % of sales 43.1% 84.0% 83.6% 84.0% 84.0% 44.3% 43.6% 45.0% 45.6% Net profit 255.4 85.0 77.1 85.9 92.4 340.4 378.5 443.3 502.9 % of sales 32.1% 37.4% 33.7% 35.3% 34.7% 35.2% 34.3% 36.2% 36.9% Basic EPS (in p) 18.16 6.04 5.46 6.11 6.57 24.21 26.92 31.52 35.76 seq change 142.2% 16.6% -9.5% 11.8% 7.6% 33.3% 11.2% 17.1% 13.4% yoy change 142.2% 36.2% 38.9% 32.6% 27.0% 33.3% 11.2% 17.1% 13.4% Adj. Diluted EPS (in p) 24.12 7.13 7.28 7.61 8.14 30.21 33.95 38.89 43.86 seq change 17.1% -1.1% 2.2% 4.6% 6.9% 25.2% 12.4% 14.5% 12.8% yoy change 17.1% 27.1% 34.1% 28.6% 13.0% 25.2% 12.4% 14.5% 12.8% Source: Bryan, Garnier & Co. ests. 26 ARM Holdings We note that between 2014 and 2018e, EPS growth should be generated by an increase of processor royalty revenues (59%), the rest come from other division sales increase (39%) and margin improvements (2%). Fig. 21: 59% of EPS improvement come from higher processor royalty revenues 50.0 +0.3 +0.3 45.0 +0.3 +19.7 +4.6 +1.6 40.0 +0.9 +0.5 +2.3 35.0 +3.8 +1.0 30.0 +4.2 25.0 EPS (in p) 20.0 15.0 Processor royalty revenues Other divisions Margin 24.1 24.1 10.0 +11.0p +7.2p improv. +0.3p 5.0 0.0 EPS 2014 Incremental EPS *Others include: automotive, other mobile chips, non-mobile AP and chips in other markets Source: Company Data; Bryan, Garnier & Co ests. What to expect in the short term? The ARMv8 technology During Q3, 3.6 billion ARM chips were delivered. Despite disappointing guidance from clients and (64-bit) is currently especially Qualcomm, which is set to incur a decline in volumes, we believe the group should adopted in more than continue to perform well thanks to internal factors, especially the ramp-up of ARMv8. Indeed, the 60% of APs and the ARMv8 technology (64-bit) is currently adopted in more than 60% of APs and the group is group is now targeting an now targeting an adoption rate for ARMv8 of 75% for the end of the year (vs. 50% previously) adoption rate for ARMv8 of 75% for the end of the bearing in mind that the new architecture provides an additional royalty rate of 50bp to the average year royalty rate of 2.5% in Q3 2015. In addition, an ever higher number of cores, especially octa- cores (est. at 40% of APs in Q3), and the increased adoption of Mali should also help the group maintain healthy momentum during Q4 2015. Note also that the group is stepping up its R&D spending slightly and this is set to increase operating expenses by GBP5m in Q4, thereby weighing on the margin during the period. As such, we expect coming quarters to carry this effect as well and expect a narrowing in EBIT margin in 2016e compared with 2015e. Despite R&D spending The 140bp impact on EBIT margin should be entirely offset by the increase in sales. As such, set to increase, we do not even though Q4 is likely to incur the highest impact on a sequential basis, we do not expect a QoQ expect a QoQ decline in decline in EPS (+2.9% in Q4 2015 vs. Q3 2015). EPS 27 ARM Holdings 4.2. Free cash cow Over the past five years, Over the past five years, the group has managed to generate free cash flow of close to 100% of the group has managed to net profit. In 2014, free cash flow exceeded GBP300m, a level that we consider sustainable over generate free cash flow of coming years. We have factored in a stable level of investments over coming years relative to 2014, close to 100% of net namely around 4% of sales. profit. Fig. 22: High FCF generation 700.0 160.0% 600.0 140.0% 120.0% 500.0 100.0% 400.0 80.0% 300.0 60.0% 200.0 Net result to FCF 40.0% 100.0 20.0% Operating FreeCash Flow (in GBPm) 0.0 0.0% FY05 FY06 FY07 FY08 FY09 FY10 FY11 FY12 FY13 FY14 FY15e FY16e FY17e FY18e Operating Free Cash Flow FCF as a % of net result Sources: Company Data; Bryan, Garnier & Co ests. 4.3. A professional in value creation A company creates value for its shareholders once its return on invested capital (ROIC) is higher than its cost of capital (WACC). ARM's ROIC has ARM's ROIC has averaged at more than 27% over the past three years, namely 18 points averaged at more than ahead of its cost of capital, which stands at around 9% (see Fig. 28). In this respect, we estimate 27% over the past three that the group creates value for its shareholders, on a stable basis. years, namely 18 points ahead of its cost of capital Fig. 23: Average ROIC of 27% (five year), 18 points ahead of the group's WACC 35% 31.0% 29.5% 30% 26.8% 25.9% RoIC 25% 24.6% 20% 22% 21% Shareholders value creation 17% 18% 15% 13% 10% WACC 5% 9% 0% FY10 FY11 FY12 FY13 FY14 Sources: Company Data; Bryan, Garnier & Co ests. 28 ARM Holdings 4.4. GBP717m returned to shareholders over 10 years The group has paid dividends to shareholders since 2004, distributing more than GBP389m over 11 years. Dividend per share has risen every year (CAGR of +25% over the past five years). We believe that the group aims to maintain the pay-out at around 25% of net profit (25% in 2011 and 2012, 24% in 2013 and 25% in 2014). Fig. 24: Dividend payments still close to 25% of net profit 16 45% 14 40% 35% 12 30% 10 25% 8 20% 6 15% % ofnet result DIvidend per share (in p) 4 10% 2 5% 0 0% FY06 FY07 FY08 FY09 FY10 FY11 FY12 FY13 FY14 FY15e FY16e FY17e FY18e Dividend per share (in pence) % of adjusted net result Sources: Company Data; Bryan, Garnier & Co ests. In addition, the group undertakes share buyback operations anecdotally. We understand that the group uses these tools especially to control the number of shares in circulation and stem any eventual dilution (stock-options). In historical terms, two periods stand out: 1/ 2005-2008 with buybacks totalling GBP262m over the period and 2/ 2014, with an amount of GBP67m spent on share buybacks. So far, the total amount of dividend payments made by the group stands at more than GBP380m including the dividends paid during the first half of 2015. Including share buyback programmes, the group has paid more than GBP717m to shareholders over 10 years. Fig. 25: GBP717m paid back to shareholders in 10 years 1400 1,164 1200 1000 717 800 600 496 338 400 Cumulative Return (in GBP) 200 36 0 FY05 FY06 FY07 FY08 FY09 FY10 FY11 FY12 FY13 FY14 FY15e FY16e FY17e Cumulative dividend return Cumulative Share buyback Sources: Company Data; Bryan, Garnier & Co ests. 29 ARM Holdings At the end of Q3, the group had net cash of GBP682m. In view of ARM's high cash generation model, we believe the group is capable of maintaining dividend returns of around 25% of net profit in coming years. This prompts us to forecast net cash of GBP2,196m for 2018e. Note also that over the past 10 years, the group has never been in a negative FCF position despite the subprime and sovereign debt crises. As such, given the current cash position and the change in CFO in January 2015 (Chris Kennedy previously at Easyjet who seems historically in favour of better shareholder returns), we believe it is possible that the group could further improve its pay-out rate. In addition, we know that management considers a gross cash level of more than GBP800m as comfortable. 30 ARM Holdings 5. Upside potential to capture 5.1. We have adopted a Buy recommendation Our Fair Value of 1,310p stems from an equi-weighted average of a DCF valuation and peer comparison. This valuation shows upside potential of 18% relative to the current price. In addition, note that these two valuation methods yield very similar Fair Values. In view of the positive momentum described previously and the share's low valuation, we have adopted a Buy recommendation on the share. Fig. 26: Overview of our valuation method Method Weight FV (in p) Upside Peer group 50% 1,302 17% DCF 50% 1,319 19% Average valuation 100% 1,310 18% Source: Bryan, Garnier & Co. 5.2. DCF valuation of 1,319p Our DCF valuation is based on the following common assumptions: Our base scenario, set out in section 4, includes our estimates out to 2018e. As for other semiconductor players that we cover, over the normal average period (from 2019e to 2024e), we have applied a pattern of cyclical growth featuring the same characteristics as the previous cycle. As such, for 2019e, we have applied a growth rate of 33% (equivalent to the peak-cycle growth rate over the past five years), followed by a linear reduction in this rate over the period in order to move towards our terminal growth rate of 3% (we have applied a terminal growth rate of 3% if the historical five-year CAGR is higher than 15%, 2010/14 CAGR in ARM sales: +18%). This scenario points to growth of 18% over 2015-24e with a usual cyclical growth pattern in the sector. We have applied average EBIT margin of 52% over 2015-24e, i.e. the margin in our 2015/18e scenario followed by a margin peak in 2019e (three percentage points ahead of the cycle-low margin), with a linear decline towards our long-term EBIT margin of 45%. This long- term EBIT margin is equivalent to the average EBIT margin over the past five years. Exceptionally, we have not applied a margin discount since we estimate that ARM has a different model to other semiconductor players that we cover since the group does not have a plant and therefore does not suffer the same scissors effect on cycle-low margins. WCR of close to 4.4% of sales over the entire period, equivalent to the five-year historical level. Investments equivalent to 4% of sales in 2015e, followed by 3.5% of sales over the remaining period, namely a level close to the historical five-year average of 3.8%. A corporate tax rate of close to 15%, corresponding to the normal-average corporate tax rate at ARM Holdings. The group benefits from a low corporate tax rate thanks to the advantages granted by the UK government to companies specialised in technological innovation. 31 ARM Holdings WACC of 9.0%. We have applied beta of 1.1x, a level similar to that applied to the valuations of ASML, a risk-free rate of 2.0% and a market risk premium of 6.4%. At the end of 2014, the group had a net cash pile of GBP668m. Fig. 27: WACC of 9.0% WACC European risk-free interest rate 2.0% Equity risk premium 6.4% Beta 1.1 Return expected on equity 9.0% Interest rate on debt 2.5% Market capitalisation (GBPm) 14,737 Net debt on 31/12/2014 (GBPm) -668 Enterprise value (GBPm) 14,069 WACC 9.0% Source: Bryan, Garnier & Co. ests. Fig. 28: DCF, a FV of 1,319p pointing to upside of 19% in GBPm (FYE 31/12) 2015e 2016e 2017e 2018e 2019e 2020e 2021e 2022e 2023e 2024e Sales 966 1,104 1,226 1,364 1,819 2,332 2,874 3,395 3,840 4,149 % chg yoy 21% 14% 11% 11% 33% 28% 23% 18% 13% 8% Operating profit 498 560 635 718 1,048 1,294 1,534 1,742 1,889 1,954 As a % of sales 51.6% 50.7% 51.8% 52.6% 57.6% 55.5% 53.4% 51.3% 49.2% 47.1% Tax -82 -92 -97 -110 -157 -194 -230 -261 -283 -293 Tax rate 16.0% 15.0% 15.0% 15.0% -15.0% -15.0% -15.0% -15.0% -15.0% -15.0% Net Operating income after tax 416 468 538 608 890 1,100 1,304 1,480 1,606 1,661 Depreciation 45 51 57 63 64 82 101 119 134 145 As a % of sales 4.7% 4.7% 4.7% 4.7% 3.5% 3.5% 3.5% 3.5% 3.5% 3.5% Capex -39 -39 -43 -48 -64 -82 -101 -119 -134 -145 As a % of sales 4.0% 3.5% 3.5% 3.5% 3.5% 3.5% 3.5% 3.5% 3.5% 3.5% WCR -43 -49 -54 -60 -81 -103 -127 -150 -170 -184 As a % of sales -4.4% -4.4% -4.4% -4.4% -4.4% -4.4% -4.4% -4.4% -4.4% -4.4% Change in WCR 8 6 5 6 20 23 24 23 20 14 Free cash flows 430 487 557 630 911 1,123 1,328 1,504 1,626 1,675 Discounted free cash flows 430 446 469 486 644 729 790 820 813 769 Total discounted FCF - 2015-2024e 6,396 Discounted Terminal value - 2025e+ 11,428 Enterprise value 17,824 - Provision (incl. pension plan) 0 - Fair value minority interests 0 WACC - Net debt on 31/12/2014 -668 [in EUR] 8.0% 8.5% 9.0% 9.5% 10.0% Equity value 18,687 43.0% 1,541 1,397 1,277 1,176 1,090 Nbr of diluted shares (m) 1,417.2 44.0% 1,568 1,420 1,298 1,195 1,107 Valuation per share (GBP) 13.2 45.0% 1,594 1,444 1,319 1,213 1,123 Valuation per share (GBX) 1,319 46.0% 1,620 1,467 1,339 1,232 1,140 Op. margin Upside 19% 47.0% 1,647 1,490 1,360 1,250 1,156 Source: Bryan, Garnier & Co. Our scenario points to a per share value of 1,319p pointing to upside of 19%. 32 ARM Holdings 5.3. Peer comparison of 1,302p Among the listed companies that have a similar business, ARM is the only one to have market capitalisation of more than USD2bn. ARM is therefore 10 times larger than the no. 2 player. If we extend our sample of peers to electronic design automation publishers (EDA - chip design software), whose customers are the same as IP sellers, the maximum market capitalisation stands at USD7.5bn, namely still half the amount of ARM's. Fig. 29: Breakdown of our sample of peers Subsector Company Market Cap. (EURm) P/E 15e P/E 16e Imagination Technologies 738 31.8x 34.5x Ceva 485 47.0x 32.7x Tessera Techs. 1,630 13.7x 15.9x PDF Solutions 332 15.1x 11.8x IP Vendors IP Interdigitals 1,765 18.7x 21.1x Dolby Laboratories 1,719 20.3x 21.9x Cadence Design Sys. 6,252 20.7x 18.8x Mentor Graphics 1,991 10.3x 13.8x EDA Synopsys 7,379 18.1x 16.8x Average Peer Group 2,694 21.7x 20.8x ARM Holdings 22,542 37.6x 32.7x Premium 74% 58% Source: Thomson Reuters I.B.E.S.; Bryan, Garnier & Co. ests. Since 2009, ARM has been trading on a significant premium to its peers. Fig. 30: The share has been trading on a premium since 2009 60.0x ARM is paid with a premium since 2009 50.0x 40.0x Since 2009, ARM has 30.0x been trading on a significant premium to its 20.0x peers. P/E ratio (12m fwd) 10.0x 0.0x 01/2005 01/2006 01/2007 01/2008 01/2009 01/2010 01/2011 01/2012 01/2013 01/2014 01/2015 Premium ARM P/E ratio Peer group average P/E ratio Source: Thomson Reuters I.B.E.S.; Bryan, Garnier & Co. ests. If we compare ARM Holding's 12m sliding P/E multiples to those of its peers over 10 years, two distinct periods stand out: 1/ before 2009, when ARM had not yet been identified as a the major winner in the smartphone trend, 2/ after 2009, when the market began to value the group at a premium relative to peers in order to take account of its better momentum. We explain this detachment relative to peers by the group's very different growth profile, and this is still true today. Over 2009/15e, we note a wide gap in top-line growth varying from eight percentage points to 21 percentage points. 33 ARM Holdings Fig. 31: A different growth profile than peers explains the premium Top line growth 2009 2010 2011 2012 2013 2014 2015 (12m fwd) Peer group average -3.0% 14.5% 14.6% 5.2% 4.7% 6.5% 9.5% ARM Holdings 11.0% 22.2% 24.8% 19.3% 25.5% 14.5% 19.4% Spread 14.0 ppt 7.7 ppt 10.2 ppt 14.1 ppt 20.8 ppt 8.0 ppt 9.9 ppt Source: Thomson Reuters I.B.E.S.; Bryan, Garnier & Co. ests. We have noted that As such, as of 2009, the emergence of this premium created an automatic decorrelation variance has narrowed between ARM's valuation and that of its peers. However, we have noted that variance has massively since 2014, narrowed massively since 2014, implying that the premium paid for ARM has stabilised over implying that the the past two years. premium paid for ARM has stabilised over the Fig. 32: Correlation strengthening again since 2014 past two years. Since 2014, 500 Before 2009, ARM valuation was Since 2009, ARM valuation is paid with a the premium 450 strongly correlated with peers. premium vs. peers. paid is stabilising 400 350 300 250 Variance 200 150 100 50 0 01/2005 01/2006 01/2007 01/2008 01/2009 01/2010 01/2011 01/2012 01/2013 01/2014 01/2015 Source: Thomson Reuters I.B.E.S.; Bryan, Garnier & Co. ests. For our valuation of ARM Holdings, we have therefore assumed the average premium of 80% in P/E terms, paid since early 2014. For information, the average premium since 2009 works out to 126%. Fig. 33: A valuation of 1,302p per share, pointing to upside of 17% Peer Group P/E ARM's historical ARM Valuation (GBX Implied P/E ratio This method yields a per 12m fwd premium vs. peers per share) share value of 1,302p ARM peer group valuation 21.3x 80% 38.4x 1,302 ARM current valuation 33.3x 1,109 Premium / Discount 17% Source: Thomson Reuters I.B.E.S.; Bryan, Garnier & Co. ests. 5.4. Valuation on historical multiples of 1,272p The group's structure and its momentum has not changed fundamentally over the past five years (the last sizeable acquisition dates back to 2004 with Artisan, for USD1bn). The group has therefore incurred no fundamental change in scope over the past five years. In addition, historical cycles in the semiconductors sector tend to last between two and five years. As such, using the method of average multiples over the past five years seems possible at first glance. However, note that for all multiples, standard deviation over five years is particularly high. Note especially that the P/E multiple hit a peak level of 60x over the period and a low point 34 ARM Holdings of 25x. As such, using average multiples over this period does not seem relevant and we are not taking into account this valuation method. Fig. 34: 12m sliding historical multiples EV/Sales EV/EBIT P/E 25x 50x 70x 60x 20x 40x 50x 15x 30x 40x 10x 20x 30x 20x 5x 10x 10x 0x 0x 0x Jul-11 Jul-12 Jul-13 Jul-14 Jul-15 Jul-11 Jul-12 Jul-13 Jul-14 Jul-15 Jul-11 Jul-12 Jul-13 Jul-14 Jul-15 Jan-11 Jan-14 Jan-15 Jan-11 Jan-12 Jan-13 Jan-14 Jan-15 Jan-11 Jan-12 Jan-13 Jan-12 Jan-13 Jan-14 Jan-15 EV/SALES Avg 5y EV/EBIT (FWD) Avg 5y P/E (FWD) Avg 5y Source: Company Data; Bryan, Garnier & Co ests. The group is currently trading below all of its historical multiples. This method points to a valuation per share of 1,272p, very similar to the valuation method presented above and that we have assumed. Fig. 35: Valuation of 1,272p per share, pointing to upside of 15% This method yields a per 12m fwd Historical multiples Current multiples Premium FV (in GBX) Upside EV/Sales 14.6x 12.0x -18% 1,201 8% share value of 1,272p EV/EBIT 30.8x 24.0x -22% 1,279 15% P/E 39.3x 33.3x -15% 1,336 20% Average -19% 1,272 15% Source: Thomson Reuters I.B.E.S.; Bryan, Garnier & Co. ests. 5.5. What risks/catalysts for the share? We have identified four main risks to our base case: The risk of losing market share in smartphones especially in application processors. This could notably stem from increased competition from Imagination Technologies, which is currently implementing an AP architecture on the basis of the MIPS technology recently acquired, as well as from Intel, which is continuing to work on mobile solutions, especially with its 4G Cat10 SoFIA chips, which could enjoy a certain degree of success. Note nevertheless that, all other factors remaining equal, a deviation of 20 percentage points relative to our target market share of 85% out to 2020e in AP would have a fairly limited impact of 3% on average on 2015/18e EPS. Similarly, deviation of 1% in the royalty rate received by ARM in APs would have an impact of 4.2% on our EPS estimates. 35 ARM Holdings Fig. 36: Sensitivity of our EPS estimates to change in the AP market ARM's market share in Application Processors 50.0% 65.0% 85.0% 90.0% 100.0% 2.0% -8.5% -6.7% -4.2% -3.6% -2.4% 2.5% -6.9% -4.8% -2.1% -1.4% -0.1% 3.0% -5.2% -3.0% 0.0% 0.7% 2.2% 3.5% -3.6% -1.1% 2.1% 2.9% 4.6% Royalty rate 4.0% -1.9% 0.7% 4.2% 5.1% 6.9% Source: Bryan, Garnier & Co. ests. Execution risk in market share gains in the networking infrastructure segment. ARM already has 10% market share in this market segment. The group has set the target of reaching 45% over five years, a target which we consider feasible in view of 1/ the upheaval in the telecoms equipment industry, and 2/ ARM's offer which is perfectly suited to new trends (see section 3.2). As such, although this is the scenario we favour, the group could nevertheless disappoint. Relative to our target of 45% market share out to 2020e, a deviation of 10 points would have an impact of 0.9% on our 2015-18e EPS estimates (all other factors remaining equal). Similarly, deviation of 1% in the royalty rate received by ARM in components destined for network infrastructure would have a 1.3% impact on our EPS. Indeed, revenues stemming from royalties generated by the network infrastructure business represent less than 10% of average sales over the period. Fig. 37: Sensitivity of our EPS estimates to changes in the network infrastructure segment ARM's market share in Networking Infrastructure 25.0% 35.0% 45.0% 55.0% 65.0% 2.0% -2.8% -2.0% -1.3% -0.6% 0.1% 2.5% -2.3% -1.5% -0.7% 0.1% 0.9% 3.0% -1.7% -0.9% 0.0% 0.9% 1.7% 3.5% -1.2% -0.3% 0.7% 1.6% 2.6% Royalty rate 4.0% -0.7% 0.3% 1.3% 2.4% 3.4% Source: Bryan, Garnier & Co. ests. Execution risk in market share gains in the servers segment. We do not expect ARM to benefit from a rapid ramp-up in this segment in coming years given the strong competition from Intel, which is particularly well established in this segment. We nevertheless consider the group capable of delivering its 2020 target for market share of 25%, thanks especially to increasingly high interest from OEMs for ARM's chips as well as the ecosystem that is now in place. In this scenario, the risk is a slower than expected expansion in the servers segment. Relative to our target of 25% market share for 2020e, a 10-point deviation would have an impact of 0.5% on our 2015/18e EPS estimates (all other factors remaining equal). Similarly, a deviation of 1% in the royalty rate received by ARM in components destined for servers would have an impact of 0.2% on our EPS estimates. 36 ARM Holdings Fig. 38: Sensitivity of our EPS figures to changes in the servers market ARM's market share in Servers 5.0% 15.0% 25.0% 35.0% 45.0% 1.5% -1.1% -0.7% -0.3% 0.1% 0.5% 2.0% -1.1% -0.6% -0.2% 0.3% 0.8% rate 2.5% -1.1% -0.5% 0.0% 0.5% 1.1% 3.0% -1.0% -0.4% 0.2% 0.7% 1.3% Royalty 3.5% -1.0% -0.3% 0.3% 0.9% 1.6% Source: Bryan, Garnier & Co. ests. Valuation associated with the GBP/USD exchange rate. The group publishes its accounts in GBP whereas 99% of its revenues are generated in USD as are around half of its operating expenses (or in associated currencies), the remainder being in GBP. As such, like STMicroelectronics, the forex impact is relatively high for our valuation of ARM Holdings. Our calculations show that a 1% change in the GBP/USD exchange rate has an inverse impact of around 1% on our valuation. Fig. 39: An inverse impact between changes in the GBP/USD exchange rate and our FV Sensibility of our FV to the GBP/USD exchange rate GBP/USD Exchange rate 1.355 1.431 1.506 1.583 1.657 Var vs. current rate -10% -5% 0% 5% 10% Fair Value 1,451 1,377 1,310 1,249 1,195 Impact 11% 5% 0% -5% -9% Source: Bryan, Garnier & Co. ests. Speculative appeal associated with an eventual acquisition of ARM by Apple or Intel. For several years, a number of investors have hoped to play an eventual acquisition of ARM Holdings by Apple or Intel. We do not really believe in this scenario since 1/ although Apple has the necessary cash, it is unlikely to acquire a company of ARM's size and also, we believe the group could develop its own architecture skills if it wanted to be independent from ARM and 2/ an acquisition project by Intel would probably be overturned by the competitions authorities given Intel's monopoly in PC/servers and ARM's in smartphones/tablets. 37 ARM Holdings 38 ARM Holdings Bryan Garnier stock rating system For the purposes of this Report, the Bryan Garnier stock rating system is defined as follows: Stock rating Positive opinion for a stock where we expect a favourable performance in absolute terms over a period of 6 months from the publication of a BUY recommendation. This opinion is based not only on the FV (the potential upside based on valuation), but also takes into account a number of elements including a SWOT analysis, positive momentum, technical aspects and the sector backdrop. Every subsequent published update on the stock will feature an introduction outlining the key reasons behind the opinion. Opinion recommending not to trade in a stock short-term, neither as a BUYER or a SELLER, due to a specific set of factors. This view is intended to NEUTRAL be temporary. It may reflect different situations, but in particular those where a fair value shows no significant potential or where an upcoming binary event constitutes a high-risk that is difficult to quantify. Every subsequent published update on the stock will feature an introduction outlining the key reasons behind the opinion. Negative opinion for a stock where we expect an unfavourable performance in absolute terms over a period of 6 months from the publication of a SELL recommendation. This opinion is based not only on the FV (the potential downside based on valuation), but also takes into account a number of elements including a SWOT analysis, positive momentum, technical aspects and the sector backdrop. Every subsequent published update on the stock will feature an introduction outlining the key reasons behind the opinion. Distribution of stock ratings BUY ratings 56.6% NEUTRAL ratings 34.1% SELL ratings 9.3% Research Disclosure Legend 1 Bryan Garnier shareholding Bryan Garnier & Co Limited or another company in its group (together, the “Bryan Garnier Group”) has a No in Issuer shareholding that, individually or combined, exceeds 5% of the paid up and issued share capital of a company that is the subject of this Report (the “Issuer”). 2 Issuer shareholding in Bryan The Issuer has a shareholding that exceeds 5% of the paid up and issued share capital of one or more members No Garnier of the Bryan Garnier Group. 3 Financial interest A member of the Bryan Garnier Group holds one or more financial interests in relation to the Issuer which are No significant in relation to this report 4 Market maker or liquidity A member of the Bryan Garnier Group is a market maker or liquidity provider in the securities of the Issuer or No provider in any related derivatives. 5 Lead/co-lead manager In the past twelve months, a member of the Bryan Garnier Group has been lead manager or co-lead manager No of one or more publicly disclosed offers of securities of the Issuer or in any related derivatives. 6 Investment banking A member of the Bryan Garnier Group is or has in the past twelve months been party to an agreement with the No agreement Issuer relating to the provision of investment banking services, or has in that period received payment or been promised payment in respect of such services. 7 Research agreement A member of the Bryan Garnier Group is party to an agreement with the Issuer relating to the production of No this Report. 8 Analyst receipt or purchase The investment analyst or another person involved in the preparation of this Report has received or purchased No of shares in Issuer shares of the Issuer prior to a public offering of those shares. 9 Remuneration of analyst The remuneration of the investment analyst or other persons involved in the preparation of this Report is tied No to investment banking transactions performed by the Bryan Garnier Group. 10 Corporate finance client In the past twelve months a member of the Bryan Garnier Group has been remunerated for providing No corporate finance services to the issuer or may expect to receive or intend to seek remuneration for corporate finance services from the Issuer in the next six months. 11 Analyst has short position The investment analyst or another person involved in the preparation of this Report has a short position in the No securities or derivatives of the Issuer. 12 Analyst has long position The investment analyst or another person involved in the preparation of this Report has a long position in the No securities or derivatives of the Issuer. 13 Bryan Garnier executive is A partner, director, officer, employee or agent of the Bryan Garnier Group, or a member of such person’s No an officer household, is a partner, director, officer or an employee of, or adviser to, the Issuer or one of its parents or subsidiaries. The name of such person or persons is disclosed above. 14 Analyst disclosure The analyst hereby certifies that neither the views expressed in the research, nor the timing of the publication of Yes the research has been influenced by any knowledge of clients positions and that the views expressed in the report accurately reflect his/her personal views about the investment and issuer to which the report relates and that no part of his/her remuneration was, is or will be, directly or indirectly, related to the specific recommendations or views expressed in the report. 15 Other disclosures Other specific disclosures: Report sent to Issuer to verify factual accuracy (with the recommendation/rating, No price target/spread and summary of conclusions removed). A copy of the Bryan Garnier & Co Limited conflicts policy in relation to the production of research is available at www.bryangarnier.com 39 London Paris New York Geneva New Delhi Beaufort House 26 Avenue des Champs Elysées 750 Lexington Avenue rue de Grenus 7 The Imperial Hotel 15 St. Botolph Street 75008 Paris New York, NY 10022 CP 2113 Janpath London EC3A 7BB Tel: +33 (0) 1 56 68 75 00 Tel: +1 (0) 212 337 7000 Genève 1, CH 1211 New Delhi 110 001 Tel: +44 (0) 207 332 2500 Fax: +33 (0) 1 56 68 75 01 Fax: +1 (0) 212 337 7002 Tel +4122 731 3263 Tel +91 11 4132 6062 Fax: +44 (0) 207 332 2559 Regulated by the FINRA and SIPC member Fax+4122731 3243 +91 98 1111 5119 Authorised and regulated by Financial Conduct Authority (FCA) Regulated by the Fax +91 11 2621 9062 the Financial Conduct and the Autorité de Contrôle FINMA Authority (FCA) prudential et de resolution (ACPR) Important information This document is classified under the FCA Handbook as being investment research (independent research). Bryan Garnier & Co Limited has in place the measures and arrangements required for investment research as set out in the FCA’s Conduct of Business Sourcebook. This report is prepared by Bryan Garnier & Co Limited, registered in England Number 03034095 and its MIFID branch registered in France Number 452 605 512. Bryan Garnier & Co Limited is authorised and regulated by the Financial Conduct Authority (Firm Reference Number 178733) and is a member of the London Stock Exchange. Registered address: Beaufort House 15 St. Botolph Street, London EC3A 7BB, United Kingdom This Report is provided for information purposes only and does not constitute an offer, or a solicitation of an offer, to buy or sell relevant securities, including securities mentioned in this Report and options, warrants or rights to or interests in any such securities. This Report is for general circulation to clients of the Firm and as such is not, and should not be construed as, investment advice or a personal recommendation. No account is taken of the investment objectives, financial situation or particular needs of any person. The information and opinions contained in this Report have been compiled from and are based upon generally available information which the Firm believes to be reliable but the accuracy of which cannot be guaranteed. All components and estimates given are statements of the Firm, or an associated company’s, opinion only and no express representation or warranty is given or should be implied from such statements. All opinions expressed in this Report are subject to change without notice. To the fullest extent permitted by law neither the Firm nor any associated company accept any liability whatsoever for any direct or consequential loss arising from the use of this Report. Information may be available to the Firm and/or associated companies which are not reflected in this Report. The Firm or an associated company may have a consulting relationship with a company which is the subject of this Report. This Report may not be reproduced, distributed or published by you for any purpose except with the Firm’s prior written permission. The Firm reserves all rights in relation to this Report. Past performance information contained in this Report is not an indication of future performance. The information in this report has not been audited or verified by an independent party and should not be seen as an indication of returns which might be received by investors. Similarly, where projections, forecasts, targeted or illustrative returns or related statements or expressions of opinion are given (“Forward Looking Information”) they should not be regarded as a guarantee, prediction or definitive statement of fact or probability. Actual events and circumstances are difficult or impossible to predict and will differ from assumptions. A number of factors, in addition to the risk factors stated in this Report, could cause actual results to differ materially from those in any Forward Looking Information. Disclosures specific to clients in the United Kingdom This Report has not been approved by Bryan Garnier & Co Limited for the purposes of section 21 of the Financial Services and Markets Act 2000 because it is being distributed in the United Kingdom only to persons who have been classified by Bryan Garnier & Co Limited as professional clients or eligible counterparties. Any recipient who is not such a person should return the Report to Bryan Garnier & Co Limited immediately and should not rely on it for any purposes whatsoever. Notice to US investors This research report (the “Report”) was prepared by Bryan Garnier & Co Limited for information purposes only. The Report is intended for distribution in the United States to “Major US Institutional Investors” as defined in SEC Rule 15a-6 and may not be furnished to any other person in the United States. Each Major US Institutional Investor which receives a copy of this Report by its acceptance hereof represents and agrees that it shall not distribute or provide this Report to any other person. Any US person that desires to effect transactions in any security discussed in this Report should call or write to our US affiliated broker, Bryan Garnier Securities, LLC. 750 Lexington Avenue, New York NY 10022. Telephone: 1-212-337-7000. This Report is based on information obtained from sources that Bryan Garnier & Co Limited believes to be reliable and, to the best of its knowledge, contains no misleading, untrue or false statements but which it has not independently verified. Neither Bryan Garnier & Co Limited and/or Bryan Garnier Securities LLC make no guarantee, representation or warranty as to its accuracy or completeness. Expressions of opinion herein are subject to change without notice. This Report is not an offer to buy or sell any security. Bryan Garnier Securities, LLC and/or its affiliate, Bryan Garnier & Co Limited may own more than 1% of the securities of the company(ies) which is (are) the subject matter of this Report, may act as a market maker in the securities of the company(ies) discussed herein, may manage or co-manage a public offering of securities for the subject company(ies), may sell such securities to or buy them from customers on a principal basis and may also perform or seek to perform investment banking services for the company(ies). Bryan Garnier Securities, LLC and/or Bryan Garnier & Co Limited are unaware of any actual, material conflict of interest of the research analyst who prepared this Report and are also not aware that the research analyst knew or had reason to know of any actual, material conflict of interest at the time this Report is distributed or made available..