Intel's Mid-Quarter Outlook: Notes and Analysis

My notes of interesting things Intel said at it's mid-quarter financial analyst meeting on May 5, 2005. My comments in [red brackets].

• Last on-stage for Craig Barrett, a gentleman and technology leader, in my humble opinion.
• The U.S. was passed by four other countries in digital buildout last year.
• Paul Ottellini: Intel is driven by long-term investments in capacity to meet WW 207 MM PC market (up 36% in only 3 years or sum 55M units), mobility (% of processors above 30% now for the first time and growing quickly and over 50% in revenue dollars-- a fundamental shift), and emerging markets (20 years in China, channel has grown from 6K to 160K members WW in last 10 years, emerging markets represent 38% of WW demand -- up from 23% 10 years ago).
• New metric is $/unit of performance or capability/watt consumed. Was performance.
• Platform inner ring is on dual- and multi-core, 64 bits, power management, "T"'s: virtualization, security, active management; chipsets business is growing, particularly in mobile; communications ring; outer ring is software: compilers, reference drivers, silicon (1M lines of code in Centrino), ISV support (i.e., multi-threading)
• Multi-core: Pentium EE will be followed by Pentium D this month. Mobile "Yonah" ships this year. 15 other multi-core projects are active. There are 3 new multi-core projects since IDF in March. 70% dual core shipments exiting 2006.
• Success in platform initiatives is measured by market growth, segment share, average selling price growth and share of wallet growth. Example is mobile: 6% market share gains, growing ASP, wallet up 5% (worth $5B in chipsets, processors, and WiFi)
• Continued double-digit growth rate this year and next.
• Sean Maloney: mobile CPU sales have doubled in last 2 years: secular change in US, Japan, and Europe to more notebooks mirrors rapid WiFi deployments.
• Expects > 30M unit run rate in smart phone silicon by year end 2005
• Intel is winning back market share from AMD in NOR memory market with better technology (90 nm this year, 65 nm next year improves performance considerably). NOR remains the cell phone memory of choice.
• Common manufacturing convergence will happen in 65 nm timeframe, beginning next year. This will lower costs.
• Centrino has matured with Sonoma platform to more than 150 design wins. Even Intel WiFi chips is getting 80% acceptance
• Next generation mobile will have better battery life and be thinner and lighter due to better heat dissipation.
• Intel is driving industry and component manufacturers towards (much) better interchangeability of components such as hard drives, opticals, and LCD panels. This will lower everybody's costs [while making innovation by an OEM harder].
• 2006 cell phone/PDA convergence is really happening. Much more memory per phone. Intel is going after all elements of this device.
• Why WiMax? Notebooks demanding bandwidth in "always on" lifestyle. GPRS is too small a pipe and WiFi coverage is too limited/cell footprints are too small. There are plenty of comms engineers rolling off 3G projects onto WiMax. But a 3-5 year timeframe to buildout and mass market.
• Abhi Talwalkar Digital Enterprise Group GM: client, server and storage, network, comms infrastructure. Drivers are workforce productivity, grow business, and IT efficiency,
• Seamless Collaboration requires unified comms,voice + video + data conferencing, across locations and companies (and partners)
• Information Processing & Delivery: get the right data at the right time to decision makers. Knowledge Workers will need dual processors for work load and multitasking
• Cross-platform Comms for staying connected anytime, anywhere, across platforms (phone, notebook, desktop, home)
• Lower TCO, provide agility and security to end users
• Embedded management engine will fit in LAN and clients chipsets: OS independent. Increase efficiency in diagnosis and resolution; manage client regardless of state [out of band]
• I/O Acceleration Over IP requires CPU, NIC, storage, and LAN ecosystem changes. Goal is improved response times over the network, increased CPU efficiency [through offload to smarter comms engines] and lowered data transfers.
• Virtualization hardware enables software (e.g., VMware) to operate much more efficiently. Expect IT-managed user images running next to management and diagnosis partitions on desktops. Server consolidation will accelerate with more efficient virtualization. Ditto storage consolidation through virtualization.
• 3,000 of 10,000 Intel software engineers are used to pollinate the Independent Software Vendor ecosystem [a huge market advantage]. New services to large customers (i.e., Wall Street) by these engineers.
• Server and chipset growth is robust [vs RISC]. Overall server $ revenue crossed above RISC in 2003, so IA gets the bulk of revenues now.
• Xeon 64-bit Nocona/Lindenhurst have fastest ramp-up ever. Very competitive versus AMD Opteron since launch in August 2004. MP Truland platform launched this March is also grabbing market share in the 4-way space with over 100 design wins.
• Over 80% of Xeons are shipping with 64 bits today. Remaining 20% by year-end.
• Dual-core DP Bensley chipset supports 8 threads in hardware on two, dual-core chips. [This is a price/perofrmance winner]. Thousands of trial units will ship in 2H-2005. The Paxville MP platform supports 16 hardware threads on a 4-way MP. Samples to OEMs soon. Demoed on stage.
• Expect more products aimed at ultra-dense, dense [i.e., blade], and performance optimized environments with better performance per watt.
• Itanium had best quarter ever in Q1-05. A $2.5B CPU opportunity. Large scale, mission-critical. 40% of F100 deployed Itaniums. 83% of wins are versus RISC [remainder is 32-bit IA and mainframes]. IBM Itanium business represents about 30% of IBMs Power RISC business -- a material business.
• Desktop demand driver with more decision support and analysis in Longhorn timeframe (2007) Office. New form-factors will emerge for home.
• Comms growth in $4.7B ATCA market in 2008
• Anand Chandrasekar - measured on design wins in all platforms, Intel has compelling platforms and can expect strong demand
• Emerging markets will soon represent 50% of Intel's WW business. Continuing investments in China, India, Russia, Latin
  America, Middle East, etc. Developing channel in 112 countires. Result: 100M unit growth in last 4 years.
• Digital Government: 70% of growth is in new PCs (i.e., new users, not replacements)
• Bob Baker, Technology & Manufacturing - 32nm pathfinding, 45nm development, 65nm ramp-up, 90nm maturity
• 65nm manufacturing defect ramp is well on the way to Intel's goals, and has encountered no bumps along the way. 2nd generation strained silicon with 35nm gate lengths. Transition crossover in units with 90nm will occur in about a year (but wafer conversion occurs in Q4-05)
• Andy Bryant, CFO -- 65nm dual-core will not cost more than today's 90nm single core microprocessors
• Replacing process technology every 2 years adds $16B revenue potential to Intel over a decade or so versus holding pat and decreasing costs. The 65nm transition actually saves $2B over its life. But there is a risk in Intel's philosophy in that production factories must be started before the process is proved out.
• $5B investment required to build a 300mm fab and get to production ($3B fab; $1-2B pilot line)
• Wafer capacity is tight -- implying demand is high and not much excess capacity at the moment
• 300mm depreciation costs are declining -- the kit is paid for and useful life is reaching maturity
• Microprocessors still 77% of revenue, with chipsets another 7-8%
• $13B free cash flow
• Stock repurchase in 2004 was 3x stock options issued
• Key measure is Return on Invested Capital. At ~19% last two years, Intel is in top 20% of S&P 500.
• Questions and Answers
• Expect a major new branding campaign later this year. [You can assume it is about dual core].
• Intel does not expect a slowdown in buying even though the new technology roadmap is clear. "People expect improved technology. They buy what is available now."
• Why is Intel revenue growth now diverging from U.S. GDP growth? A: Intel is now much more global economy GDP-growth driven than mere U.S. economy.
• The semiconductor ecosystem is unable to accelerate Intel's desire to move from an 8 quarter to a 6 quarter technology refresh cycle time.

- Peter S. Kastner