0:00:01
The Wall Street Journal has just reported that Apple has reached a preliminary agreement to get Intel to start manufacturing some of Apple's self-designed chips. A separate report from Bloomberg essentially claimed the same thing. So, these reports are likely true. That's a pretty big deal given that Apple is probably the most lucrative chip client after Nvidia today. These reports followed an announcement from a few weeks ago where Elon Musk's companies like SpaceX and Tesla would start building chips using
00:00:25
Intel's next generation 14 process as well. Not much earlier than that, Nvidia invested $5 billion into Intel. And of course, we can't forget that the US government itself became Intel's largest shareholder, acquiring a 10% stake last year, too. In other words, Intel, who until recently was the sick man of the chip industry, has gone through a remarkable turnaround in the last year or so. Unsurprisingly, their stock price is up by almost 500% over the last year, which has outpaced even many of the
00:00:52
other chip champions who themselves have all had tremendous runs during the AI boom. So, in this video, let's talk about how Intel suddenly turned everything around and how this, in my opinion, proves that the previous CEO, Pat Gellzinger, was probably right about pretty much everything and got fired unfairly. A year ago, I made a video saying that Pat Galsinger, the former CEO, took Intel across what I then called the valley of death. This is the process where a company makes a huge and very
00:01:23
expensive bet that requires them to invest a ton of money. They hope that the bet eventually pays off and brings them big returns. But before that happens, as they cross the valley, there's a risk of them just running out of money, investors getting cold feet, or any number of things going wrong that could cause the company to never make it to the other side. Now, chipmaking, for our simplified purposes, consists of two separate jobs, designing and manufacturing. And Pat's big bet was spending unprecedented amounts of money
00:01:47
on making Intel a worldclass chip manufacturer. began. While almost all other chip companies like AMD, Nvidia, and Qualcomm only focused on design and outsourced their manufacturing to specialized fabs like TSMC, Intel had historically done both in-house. But their manufacturing had over time fallen behind TSMC's, which had dragged its whole business down and left Intel with basically two choices. One, just give up and start outsourcing manufacturing to TSMC as well. Or two, make what is probably the most expensive bet in the
00:02:16
chip industry ever to try to catch up with TSMC. Pat, who was a hardcore engineering type, was specifically hired because he chose the second option. And he declared that he would try to deliver five manufacturing generations or nodes in four years to catch up with the SMC. No company had ever done something like this successfully before. And so his firm went on an incredible spending spree, which pushed them into the valley of death. In short, he spent essentially all the cash that his firm produced. And
00:02:41
because this wasn't enough, he also ruthlessly sold off or spun out parts of the firm that weren't central to his mission. He sold the firm's own fabs to private equity investors and then leased those back to free up cash temporarily. He laid off tens of thousands of people who weren't working on his turnaround. And of course, he spent a ton of his time convincing governments to co-inance his manufacturing expansion on top of that as well. In other words, this was an allout bet. Pat Gellzinger punched
00:03:07
Intel deep into the valley of death and then he got fired right as the company reached the absolute bottom of the valley and everything looked absolutely terrible. The investors and the board of directors got cold feet and they pushed him out. A new guy called Lip Bhutan was put into his place and his initial instinct was to scale back on Pat's ambitious manufacturing spent. He said that there were no more blank checks and that he would even fire many of the people who were supposed to make the
00:03:30
manufacturing turnaround happen. This was kind of understandable at the time given that his predecessor was literally fired for spending too much money on manufacturing, but it also would have been catastrophic for Intel becoming a worldclass manufacturer again. And scaling back on manufacturing would have not only been a questionable move for Intel, but it would have also been a huge blow to American efforts to win back chip manufacturing from Taiwan and South Korea again. And so, none other than Trump stepped in to stop this from
00:03:54
happening. In a classic Trump move, the president made a series of personal threats against Lip Bhutan, calling him essentially a national security threat and calling for him to step down from his role. And this resulted in Lip Bhutan striking a deal with him. He sold the US government a 10% stake in Intel, which made them the largest shareholder of the company. And he also agreed to a clause where the government could step in if Intel ever wanted to sell off its fabs. And in exchange, Trump simply
00:04:19
stopped talking about him as a national security threat and started treating him like an ally. In other words, Trump essentially forced Intel to stick with Pat's plans. And luckily for everyone, the timing on this was perfect because not only was Intel starting to exit its own valley of death, but also the entire rest of the industry gave them a ton of backwind. Remember those five nodes in four-year graphics? Well, with a bit of delay, 18A, the last node in the sprint, actually dropped soon after Pat was
00:04:44
gone. And as it turns out, it was pretty good. Probably not quite as magical as Pat hyped it up to be, but it put Intel solidly back on track to being a competent manufacturer. Again, we know this to be true because Intel's latest generation of processors, cenamed Panther Lake and Clear Water Forest, made in large part using Intel's 18A process, are both actually pretty competitive. Fun fact, I got to tour Intel's 18A fab in Arizona where these chips are actually made. And while I wasn't allowed to film inside, seeing
00:05:09
all of the equipment, including the EUV machines, inside was one of the most unforgettable experiences I've had so far. Chip manufacturing is truly an awe inspiring thing. Anyway, and meanwhile, 14A, the next note can't be too bad either since that's what Elon Musk's companies are adopting now. Of course, chip fabs take years to build and chips take years to design. So, we know for a fact that things like Intel 18A, Pantoake, etc. are all primarily the result of massive engineering efforts
00:05:35
that happened under their previous leadership. Especially given that we have seen these exact products being previewed by Pat before he was fired. In other words, we now know that while Pat was at the company, they laid some pretty solid technical foundations and after he was gone, the company got really lucky. I'm talking about the AI boom, of course, which is the mega trend that has lifted all boats in the chip industry lately. I think by now everyone is sick and tired about hearing how many
00:05:58
data centers are being built, but I recently saw this chart on Asimo's blog which really put things into context for me. Again, the post itself is titled Google's free cash flow will hit zero next year. But the chart is even more interesting than that. It shows the free cash flows of all so-called US hyperscalers. So basically all the large cloud giants that have data centers, Microsoft, Amazon, Alphabet, Meta, and Oracle. Note how until about 2024, these firms had massive cash surpluses which
00:06:25
were also growing quickly until they almost reached an insane $300 billion. These companies were cash generating machines. But then poof, all of that is gone in a year and a half with the combined free cash flows reaching basically zero. These firms are now spending every dollar they have on data centers. And in fact, many are going cash negative, taking on huge loans as well. This incredible spending frenzy has of course been great for all the chip companies that touched the data centers. And as we've just discussed,
00:06:53
Intel now does have a competitive data center chip again in the form of clear water forest. But recently, they've also been two more things going for Intel. First, data centers in the beginning were mostly filled with GPUs as those were the primary chips used for training, which was bad for CPU makers like Intel. But lately, especially with agentic AI being the hot new thing, more generalpurpose CPUs are needed in the data center again to orchestrate a lot of tasks. This means that Intel has
00:07:19
recently been getting a much bigger share of the pie. In their latest earnings call, they even said that they were able to sell some older chips that they had previously already written off as too old and too outdated, but there's so much demand for CPUs again that they could just sell them. And second, there's such a demand for chips of all types now, be it CPUs, GPUs, memory chips, and more that the traditional fabs like TSMC and Samsung just can't keep up with the demand. Suddenly, this
00:07:42
makes Intel's manufacturing capacity extra interesting. See, historically, Intel has built factories for its own use, but PET explicitly rebuilt Intel Foundry to try to attract third party customers as well. He never got to announce any major customers before he was pushed out, but now all the stars are starting to align. Intel 18A and 14A have matured and proven themselves. Everyone is looking for more capacity. And credit where credit is due, the new CEO has proven himself to be a pretty shrewd dealmaker. Le Bhutan is a chip
00:08:11
industry veteran who by all accounts is very wellconed personally and has been instrumental for convincing potential clients to give them a shot. And perhaps most importantly, while he probably wasn't happy about Trump's threats, he managed to turn lemons into lemonade by giving the US government a 10% stake in his company. He successfully aligned the interests of Intel with those of Trump. And in its current state, Trump is probably the best ally Intel could possibly hope for. Because while he
00:08:36
might have flip-flopped on some of his other political ideas, he's been remarkably consistent about two things. One, wanting more manufacturing in the US, and two, loving to make deals. Trump explicitly said that he would help Intel win customers when he made the purchase. And back then, I made a whole video titled How Trump will force Intel to win. You can watch the video in full in the description. I think it holds up pretty well. But this is pretty much exactly what happened. As Trump has
00:09:01
since said, quote, "As soon as I went in, Apple went in, Nvidia went in, a lot of smart people went in." We know that Trump's explicitly expects big American firms like Apple, for example, to make large public commitments to US manufacturing. And while I think this never really made sense for, for example, iPhones, it absolutely can make sense for chips. And for that, the most obvious choice would be Intel. Maybe Intel's FABs are a little more expensive and a little less good than competing
00:09:27
FAPS from TSMC and Samsung for now. And I doubt that Apple or Nvidia will move their most high-end chips away from those anytime soon. But due to the big push on their patch, the gap is now not all that big anymore. And given all of the circumstances, an advanced US fab that is owned by a US company and co-owned by the US government, one that has available capacity, one that is kind of just good enough, and one that is directly aligned with the US president is a pretty damn attractive thing right
00:09:52
about now, especially since not all chips have to be made to the absolute highest specs at all times. And so this, in my opinion, is what Intel climbing out of the valley of death looks like. Pat never made it to the other side, but his company probably did. All right. Did you know that I have a Patreon? If you want to support the channel, you could do that over there, or you can also do it over on Nebula. If you choose Nebula, you also get a ton of extra videos, bonus documentaries, everything, Q&A
00:10:16
sessions with me, and so on. But the choice is yours. Links are in the description. Thank you for supporting the channel, and I'll see you in the next one. Bye.