00:00You said the investment bodes well for Intel's turnaround story.
00:03Does this really strengthen the balance sheet? How does this look to be boding well?
00:08Yeah, so Intel's been busy in its turnaround. It made a really big step in repurchasing 49%
00:16of its Ireland fab back from Apollo. And then you have this morning's announcement with Tesla.
00:25Intel has all the most important pieces right now. It's a U.S.-based company with U.S.-based
00:32production that makes CPUs and has advanced packaging. These are all things that we need
00:39for the future of AI. These are three things that if I said three years ago, they wouldn't have
00:44mattered that much. And right now, each one of those things is absolutely critical. It's not
00:50trivial, though, that Intel will be successful. I think that came up in your conversation with
00:55Mandeep. There's a lot of execution ahead. They need to successfully build plants in the U.S.,
01:02build fabrication for their next waves of technology, including this big joint venture
01:07with Tesla, and then get the yields that make them profitable. As you pointed out, TSMC has been doing
01:13this for years with great levels of success. And Intel still loses money on the fab side and doesn't
01:19have enough core customers to have that volume quite yet. It's lining up the customers, but it's
01:27not quite there yet. And so there's a lot of execution ahead. But in terms of executing on
01:33the turnaround, Libouton is doing a lot to make Intel an important player in the future of AI.
01:40Gil, do you like the idea? Because Libouton came under some, well, cynicism or indeed was beaten up a
01:47little bit, even by the administration of the United States, because they didn't really like
01:51the way in which he was proposing it as build it and then they will come. He wanted to go
01:55more like,
01:56I need the customers first, and then I'm going to go out and build it. Is this enough of a
02:00customer
02:00that they will therefore start to put back onto the R&D, start to refuel the actual manufacturing and
02:06build out that perhaps it pours on?
02:10If we believe Elon Musk and the numbers he's been talking about, that's as much scale as you're
02:15going to get. The numbers that he talks about for XAI, for SpaceX. I mean, we're talking about
02:23building in the U.S. now. Elon's talked about having fabrication plants on the moon. So if we are to
02:29fill all of those and Elon gets what he wants, that's plenty of volume. Let's not forget NVIDIA
02:35as an investor in Intel. That's volume. There's still a relationship with Apple. That's a lot of
02:40volume. So Intel has positioned itself to have the volumes necessary to bring the customers to the
02:47table, to have the volumes necessary to be profitable on the fabrication side. It really
02:54is a matter of execution over the next several years to get those customers what they want
02:59with the high expectations those customers have. Let's not forget those customers are being served
03:05very well by TSMC. Yeah. This is volume beyond TSMC, but their expectations are still anchored
03:12to what they get from TSMC, which is very high yield. Can we go nitty gritty? Because Tesla has
03:18relationships for its own chip building at the moment, and they lie with Samsung and they lie with TSMC.
03:24So how divert? What chips in particular are expecting to be made by Intel? Or indeed,
03:29where do we see the supply reshift? Well, I'm not sure we got that detail today. But broadly speaking
03:35these days, if you look at the broad patterns, this conversation, the conversation yesterday about
03:41Broadcom, Google and Anthropic, Meta signing up with AMD, OpenAI signing up with AMD and with Broadcom
03:48and continuing to use NVIDIA, the approach of all the big players, so the hyperscalers and the big
03:55frontier labs, which includes Elon, Inc., is to diversify around all of these large chip makers
04:04in order to be able to meet this very significant demand that they all have for chips. So they all
04:13are still, most of these companies are still mostly reliant on NVIDIA, but they want to rely more on
04:20Broadcom, AMD and Intel, so they're not beholden to NVIDIA, so they have more capacity. What that looks
04:27like in terms of mix, I'm not sure they know. I think they're going to move forward and expand as
04:32aggressively as they can. And the chip companies that help them the most, that execute the best,
04:39will get the volumes. That's why execution is so important right now.
Comments