I had the chance to talk with Intel CEO Pat Gelsinger Monday afternoon to get the latest on strategy disclosures on manufacturing, technologies and product roadmaps. Today, the company streamed an event and took questions on the content. Overall, Gelsinger’s disclosure gave me many reasons to believe Intel is “back” if the company can execute its plans. Let me hit the news and my thoughts on it.
IDM 2.0 Strategy news
Intel today announced in its press release that:
- Two new fabs: Arizona-based, EUV-enabled, beginning with a $20 billion investment. These will be available to both internal Intel and new foundry.
- Intel 7 nm: process development progressing well with tape in of 7nm compute tile for “Meteor Lake” expected in the second quarter of 2021 with high volume manufacturing in 2023
- New foundry business: Re-entering the foundry business with “Intel Foundry Services” with the goal to become a major provider of leading-edge foundry capacity in the U.S. and Europe to serve global customers. Dr. Randhir Thakur will lead this business reporting into Gelsinger. IFS will offer X86, graphics, A.I., display, Arm and RISC-V IP.
- Increased foundry use: Intel uses a lot of external foundry capacity today
- IBM Research collaboration: co-creating next-generation logic and packaging technologies, with the goals to accelerate semiconductor manufacturing innovation across the ecosystem, enhance the competitiveness of the U.S. semiconductor industry and support key U.S. government initiatives. Both IBM and Intel are research leaders in this space.
- Intel On forums: Similar to the Intel Developer Forum (IDF) of yesteryear, the company is committing to bringing together its ecosystems. The first event is called Intel On Innovation in October in San Francisco
Gelsinger, a straight shooter
I have to start by saying I appreciate Gelsinger’s honest self-deprecation related to previously missing 10nm schedules and having to push out 7nm recently. With me, Gelsinger used the words “abysmal” and “embarrassing,” and I appreciate that as it’s the most direct any Intel senior executive has ever been with me on the topic. He explained the 10nm overreach, going for it on too many variables, not hitting on enough, making it five years late. He also talked about the passivity on EUV given how much work Intel had put into it, only to be used first by TSMC. Openness and honesty elicit trust with me. Sure, I understand that Gelsinger wasn’t at the company when it happened, so it makes it easier to explain, but that doesn’t matter. It’s part of what I needed to hear.
Intel 7nm and 2023 risk mitigation
I have become more confident with Intel’s 7nm plans given its sticking with its revised schedules and its increased use of EUV which should simplify the manufacturing process and should improve efficiency when it is released. While 7nm Meteor Lake won’t be in high volume production until 2023, Intel announced it will be tapping into external foundries for 2023. I am expecting Intel to use TSMC as a “gap filler” for some of the highest performance CPU tiles, an area of intense competitive pressure until the company gets its process house more density and power competitive. It will be interesting to see both AMD and Intel CPUs on TSMC 3nm process head-to-head.
$20B is only the start
The $20B doubling down of manufacturing is bold and is intended to serve its internal uses and be used for its own Intel Foundry Services. I don’t believe this is the extent of Intel’s investment, particularly given the need of U.S. and European governments who are asking for more on-shore, leading-edge manufacturing. Obviously, if Gelsinger and the board didn’t have confidence in its future capability and reinforced by its end customers, it would not be making these investments. This isn’t Intel hedging its manufacturing bets- it looks to me as the company is all-in. I was pleased to hear that the $20B investment is not determinant of any government funding, but I believe Intel should get funded by both the U.S. and the E.U. to build more capacity more quickly.
Intel Foundry Services actually sounds credible
As for its new foundry business, Intel Foundry Services, I believe this is the first time Intel is seriously doing what it needs to take to create what I consider a “real” foundry. It’s the first time I’ve seen the company commit to standard PDK models, standard design rules, and standard tools from Cadence and Synopsys. That is a requirement for other foundry customers who already use those models, rules, and tools and enables import to and export designs from TSMC, Samsung, UMC, and GlobalFoundries.
When discussing I.P., the only surprise for me was RISC-V, but given increased interest, I shouldn’t be. This was in addition to the X86, graphics, display, A.I., and Arm I.P. discussed. While I am not privy to anyone’s legal agreements between designer and Arm, I think that if an Arm licensee like Qualcomm wanted to build some Arm-based chips at IFS, it could without Arm’s permission. Oh, and by the way, Intel is an Arm licensee.
For what it’s worth, I could see a world where both the U.S. and E.U. mandate certain products destined for certain government institutions, critical infrastructure, maybe even financial and healthcare industries, to be manufactured on U.S. soil by a U.S. company. Of course, TSMC in Arizona and Samsung in Austin would likely beg to differ, but I think this gives Intel an advantage.
Intel has had a tough few years reputationally based on the domino effects of the 10nm slips. Ironically, the company has been overperforming financially, still has 90% share of the server market, 80% of the P.C. market and growing in ancillary markets like automotive, networking and IoT. While I am certain many of the news disclosed today were in motion under prior CEO Bob Swan, Gelsinger’s mark is clearly on the plans, and he exudes senior-level technical and execution confidence and openness that I haven’t seen at Intel for years. Gelsinger said all the right words and gave me all the reasons to believe that Intel is “back,” but it now needs to execute.