This morning, Intel is set to provide updates on its foundry business (IFS) and process roadmap at its IFS Direct Connect event in Santa Clara. Intel is expected to unveil plans for how the company will transform the foundry industry and how it is set to become the world’s first and only fully integrated systems foundry in the AI space. Led by Intel CEO Pat Gelsinger and Stuart Pann, the Senior Vice President and General Manager of Intel Foundry Services, both will deliver the event’s opening keynote. Expected guests throughout the keynote include Sam Altman, the co-founder and CEO of OpenAI, Gina Raimondo, the US Secretary of Commerce, and Satya Nadella, the Chairman and CEO of Microsoft.
08:26AM PST – It’s a busy and surprisingly crowded morning at Intel’s first dedicated foundry event for the modern era
08:26AM PST – Everyone is quickly taking their seats for a scheduled 8:30am start
08:27AM PST – According to Intel, attendance is already exceeding their expectations. They had over 1000 registered attendees as of last Thursday, and that’s continued to grow since then
08:29AM PST – The hardest part with doing your first event is knowing just how many people to plan for. Despite all of their history, Intel is a newcommer for the contract foundry space, so the question of how many customers they can attract in their first years remains a big one
08:29AM PST – IFS Direct Connect is for all practical purposes the coming-out party for Intel’s foundry group
08:30AM PST – Up until now, it’s been roadmaps and doing the hard work of catching up on node development, all the while also working to develop a new-to-Intel external foundry ecosystem
08:31AM PST – Intel is still in the process of doing all of this – and will be for years to come – but the first fruits are finally coming to bear
08:31AM PST – As evidenced by a session on EDA tools later this morning that is expected to take an hour
08:31AM PST – And here we go
08:32AM PST – Intel Foundry: We were made for this
08:32AM PST – Surprise! IFS is no longer IFS. The entire fab, testing, and packaging group is now simply Intel Foundry. Or as Intel writes it in all-lowercase, “intel foundry”
08:32AM PST – And here’s Pat Gelsinger
08:33AM PST – Welcoming everyone to Direct Connect 2024
08:33AM PST – Last week marked his 3 year anniversary as the CEO of Intel
08:33AM PST – Recapping his three big goals for Intel, rebuilding Intel and righting the ship after Intel’s foundry foibles in the last decade
08:34AM PST – As well as Pat’s first big event, Intel Unleashed, which outlined Intel’s IDF 2.0 foundry strategy, which has morphed into Intel Foundry today
08:34AM PST – Gelsinger is all about reestablishing Intel’s process leadership position
08:35AM PST – “Today, this vision of Unleashed becomes real”
08:35AM PST – Intel’s first announcement of the day: Intel Foundry
08:35AM PST – A renaming and a reorg of Intel’s foundry groups. Technology development, global manufacturing and supply chain, and Intel’s foundry services and ecosystem operation
08:36AM PST – Meanwhile the other half of Intel will be the Intel Products group
08:36AM PST – This is where the client, datacenter, and other products that Intel develops
08:37AM PST – And these two groups are independent to meet the needs of each – though obviously still interdependent to an extent as Foundry needs Products’ chips, and Products will need a world-class foundry
08:37AM PST – And now: AI
08:38AM PST – AI is the big buzzword of the industry right now, and even though Intel doesn’t have a complete product stack here, Intel Foundry is all about making chips for the AI era
08:39AM PST – Intel is terming their foundry model the “Systems Foundry”
08:39AM PST – Gelsinger: “Moore’s Law is alive and well!”
08:39AM PST – Outlining all of the foundry technologies Intel is bringing to market over the next several years. High-NA, glass substrates, die stacking, and more
08:40AM PST – And now outlining a bit more of what to expect from today’s show. Intel has a bevy of guest speakers, from government officials to OpenAI’s Sam Altman
08:41AM PST – As well as numerous Intel executives, including foundry VP Stu Penn and Dr. Ann Kelleher
08:42AM PST – Intel also has numerous partners here at the show. EDA and IP providers to flesh out their ecosystem so that external customers have the tools needed to design chips for Intel’s foundries, and the IP needed so that clients have all the parts and libraries they need to build a complete chip
08:43AM PST – Gelsinger is thanking the EDA partners
08:43AM PST – Moving on, Gelsinger wants to become the world’s #2 foundry by 2020
08:44AM PST – Even if Intel retakes process leadership in 2025 as they intend, it will be many, many years of growth and shifting customers to overtake the current #1 of TSMC. Overtaking Samsung will be an easier task
08:46AM PST – And there is a definite degree of geopolitics in Intel Foundry’s plans as well. Intel wants their foundry business to be the answer to business and government customers around the globe who want foundries on friendly and safe soil – a resiliant foundry that isn’t faced with geopolitical risks
08:46AM PST – Silicon is to modern geopolitics what oil has been for the last 50 years
08:47AM PST – In 1990, 80% of semiconductors were built in the US and Europe. Now it’s 80% in just Taiwan and South Korea
08:48AM PST – Intel’s moonshot goal: rebalancing things to get to a 50/50 split between the US/Europe and Asia
08:48AM PST – And now for Pat’s first guest speaker: Gina Raimondo, the US Secretary of Commerce, and a key figure in the CHIPS Act that is helping to fund Intel’s foundry ambitions
08:49AM PST – Raimondo is comparing the current moment to the space race, as that was the last time the US government made such a significant investment in a single industry
08:50AM PST – COVID changed a lot of things. It laid out bare the US’s reliance on external fabs; the inability to produce all of the chips it needs domestically
08:51AM PST – “You’ve been a fantastic partner”
08:52AM PST – Why is building chips on the US so important? The Feds don’t want to make every chip in America, but they want to diversify their supply chain and get much more leading-edge chip manufacturing in the US
08:52AM PST – Not quite complete self-sufficiency, but a self-propelling engine of production
08:56AM PST – “If we want to complete globally, we’re going to have to continue to invest”
08:57AM PST – CHIPS is about more than just incentivizing 10 new fabs in the US. The US government wants to kickstart a much larger domestic ecosystem
08:57AM PST – And that’s Secretary Raimondo
08:57AM PST – Now back to Pat and his Siliconomy
08:58AM PST – Expecting 25% of the world GDP being fueled by the “digital economy”. AI has the potential to make this grow quickly
08:58AM PST – Now on to process nodes and roadmaps
08:58AM PST – Recapping Intel’s 5 Nodes in 4 Years promise
08:59AM PST – Intel 4 is shipping, Intel 3 is HVM ready
08:59AM PST – And the new Angstrom nodes, 20A and 18A, with their GAAFETs/RibbonFETs and PowerVia backside power delivery, remain on track
08:59AM PST – Intel needs to ship 18A by 2025 to make their milestones
09:00AM PST – Intel’s first 18A product has taped out. Clearwater Forest, their lead product for 18A
09:00AM PST – Clearwater is Intel’s second-generation E-core Xeon, to succeed Sierra Forest
09:01AM PST – Clearwater Forst usiong an Intel 3 base die, along with EMIB, and most surprisingly, Foveros Direct, Intel’s ultra-fine die-to-die copper bonding technology
09:01AM PST – Intel has also landed a new 18A customer: Microsoft
09:02AM PST – Now rolling a clip of a message from Microsoft’s CEO Satya Nadella
09:02AM PST – MS looks forward to sharing more details on their 18A chip in the future
09:02AM PST – 18A is the light at the end of the tunnel
09:02AM PST – But what comes next if Moore’s Law is supposed to be alive and well?
09:03AM PST – What comes after 18A is 14A, as well as several new minor nodes
09:03AM PST – For more details, please see our article here: https://www.anandtech.com/show/21271/intel-foundry-future-14a-foveros-direct-beyond
09:03AM PST – And Pat has a 14A wafer!
09:04AM PST – As for the minor nodes, Intel is planning feature and performance enhancements of existing nodes, as well as nodes with specific through-silicon-via capabilities
09:04AM PST – Intel 3-T, Intel 3-E, Intel 3-PT, Intel 16-E, and towards the end of the roadmap, Intel 14A-E
09:05AM PST – E= Feature enhancements. P= Performance enhancements. T= Through silicon vias
09:05AM PST – As Intel Foundry, the comapny is also throwing the doors wide-open on their advanced packaging technologies. Everything from EMIB to Foveros Direct
09:06AM PST – All of this is in development or available to customers. And is a cornerstone of Intel Foundry’s offerings. Intel will catch up on lithography; they thinkn they’re already in the lead in advanced packaging
09:07AM PST – Intel Foundry has secured over $15 billion in customer business
09:07AM PST – “Today is a day 3 years in the making”
09:08AM PST – And Pat is now handing things off to Stu Pann, Intel’s SVP and GM for Foundry Services
09:08AM PST – Pann is here to talk about just what a “systems foundry” is. And why the AI era is driving so much demand
09:09AM PST – (Pann has been at Intel for 43 years)
09:10AM PST – Intel needs to earn its chance to be a customer’s chip fab
09:11AM PST – Recapping the history of Moore’s Law, and Gordon Moore’s original paper that lead to his law
09:12AM PST – Moore’s Paper discussed the possibility of building larger systems out of separately packaged components
09:13AM PST – And this is where Intel reaches the system foundry. A foundry ecosystem that follows open standards to allow multiple components/chiplets to be used together
09:13AM PST – Intel is a systems company turning into a foundry
09:14AM PST – Comparing Intel’s new foundry model to TSMC’s
09:15AM PST – TSMC has been very succesful as a classic foundry provider that processes wafers, does a bit of advanced packaging, and leaves the rest to other ecosystem partners
09:15AM PST – Intel wants to be more and do more, in order to keep Moore’s Law going
09:15AM PST – Offering not only traditional chipmaking resources, but also the company’s expertise in system architecture and software
09:16AM PST – Intel visualizes this as a 3 layer model
09:16AM PST – The base layer is being a world-class foundry and complete ecosystem
09:17AM PST – The second layer is offering manufacturing resiliancy
09:17AM PST – And finally, at the top, the idea of a system of chips. Working with foundry partners to create systems of chips
09:17AM PST – 18A is being well-received by the hyperscaler community
09:18AM PST – The high-power nature of AI accelerators makes backside power delivery particularly interesting to these groups
09:19AM PST – Meanwhile, Intel is going to be doing more than just cutting-edge chips. Using its existing tooling and expertise to develop cost-effective nodes like their Intel 12 collaboration with UMC
09:20AM PST – Intel wants to cover as many paths as possible with customers. Give them access to whatever they need. EDA tooling, IP, design services
09:21AM PST – As for Intel’s resiliant fab plans, this covers both the geopolitical aspects, as well as the environmentak sustainability aspects
09:22AM PST – Security is also a major platform for Intel, especially as they chase military business
09:23AM PST – And finally, the system of chips is all about taking advantage of chiplets and tiling
09:23AM PST – Both to improve chip yields, and to let customers mix and match chiplets to build exactly the kind of chips they’d like
09:24AM PST – Pann says that Intel Foundry has learned a lot about chiplets through Ponte Vecchio (aka Intel Data Center GPU Max)
09:24AM PST – The assembly test yield on Ponte is 95%+
09:25AM PST – Which is what made building the Aurora supercomputer possible
09:26AM PST – And recapping all the uses for Aurora. And how Intel wants to make something like that even more cost efficient
09:26AM PST – Systems of Chips requires not only advanced chip packing technologies, but also standards within those ecosystems for how those chiplets will connect an interact
09:27AM PST – e.g. UCIe, co-packaged optical interconnects, and more
09:27AM PST – Providing standards and connectivity
09:28AM PST – Moving on: “What does it take to bring AI everywhere?”
09:28AM PST – AI use is growing rapidly. AI power consumption is also growing rapidly
09:28AM PST – “AI could soon need as much electricity as an entire country”
09:29AM PST – Expanding the role of AI means finding ways to develop these chips cost-effecitvely. And then making them power-efficient so that they’re cost-effective to run
09:30AM PST – Spotted: Foveros 3D Direct with a bump pitch less than or equal to 4 micron, less than half that of first-gen 9 micron Foveros Direct
09:31AM PST – Getting to that point is going to require all of Intel’s foundry capabilties. Leading-edge nodes, leading-edge chip packing, glass substrates, and the ability to cool 2000W+ devices
09:32AM PST – New memory technologies is also going to be a major component of improving AI. Not just more bandwidth, but keeping that power consumption in check
09:32AM PST – And Intel is going to continue pushing new interconnect and networking technologies as well, such as silicon photonics and co-packaged optics
09:33AM PST – “It’s another 5-year roadmap”
09:33AM PST – And to top this all off with Intel’s software expertise
09:34AM PST – And to do all of this, Intel can’t do it alone. They have to have partners
09:34AM PST – Reference designs to speed development. EDA tools to actually make these chip designs
09:35AM PST – And Intel’s investment arm, Intel Capital, to further bootstrap the ecosystem and funding the next wave of innovators
09:36AM PST – Intel is even partnering with universities to make 18A avaialble to their students, so that those students come out of their schooling with experience on Intel’s leadijng-edge node
09:36AM PST – Intel is also announcing a new partnership with Arm. The Emerging Business Initiative. Arm provides the IP, Intel provides the 18A shuttles
09:37AM PST – Intel has 30+ partners overall, several of which will be speaking today at Direct Connect
09:37AM PST – With EDA tool makers providing hard engineering data on their Intel tools
09:38AM PST – But first, Intel’s most important partner: Arm
09:38AM PST – Now on stage, Arm’s CEO Rene Haas
09:38AM PST – Pann: “Arm is my most important business partner”
09:38AM PST – Which just goes to show how wildly the world of chipmaking has changed in the last decade
09:39AM PST – Intel cannot be in the foundry business without a partnership with Arm
09:39AM PST – Haas: “This is a bit of strange bedfellows”
09:40AM PST – Intel and Arm started their conversations shortly after Haas became CEO in early 2022
09:40AM PST – Haas is recapping today’s Neoverse product announcements. Neoverse V3 and Neoverse N3
09:40AM PST – More details here: https://www.anandtech.com/show/21270/arm-announces-neoverse-v3-and-n3-cpu-cores
09:40AM PST – Intel for their part is super-excited about the Neoverse announcement
09:41AM PST – Haas: “It has been a joy to work with [Intel’s] groups”
09:42AM PST – “The results so far are terrific”
09:42AM PST – Arm is ready to get to the next phase and have customers use the product
09:43AM PST – Also quickly recapping Arm’s Compute Subsysten (CSS) initiative, which delivers fully designed and validated CPU cluster designs
09:43AM PST – Which can save customers significant time by cutting down on the amount of remaining design and integration work they need to do
09:44AM PST – Arm and Intel are looking forward to seeing what they can do together
09:44AM PST – And that’s a wrap for Arm
09:45AM PST – And that will be a wrap for our live blog coverage as well. Intel’s public keynote will be continuing for another 20 minutes or so, with a fireside chat with Intel’s customers
09:46AM PST – Thank you for joining us, and please be sure to check out our companion article on Intel’s foundry announcement: https://www.anandtech.com/show/21271/intel-foundry-future-14a-foveros-direct-beyond