Spent a long time at Intel. Only 34 years. Probably one of the greatest American companies ever, and then absolutely went off the rails and got absolutely demolished by Nvidia, TSMC, and Apple to a certain extent. So let's talk about how things went wrong, what went right, and what were the critical mistakes we can learn from.
Having spent so much of my life there, I joined when I was 18. I went through puberty at Intel. Grove, Noyce, Barrett โ they were my mentors, they were deeply technical. When I joined the executive staff for the first time, probably 15 of the 20 people in the room were PhDs. One of the things that went off the rail was when it started to be run by business people as opposed to technical, the bean counters, the finance people.
And you know when I became CEO in 2001 that was the first technical leader in essentially 15 years. And if you have a business leader, who does he promote? Business leaders. As you look at the great technology companies today, they're deeply technical and founder-led typically. And even if they're not โ Satya is not a founder, Sundar is not a founder โ but they're deeply technical individuals. And when you're making these hardcore technical decisions that affect billions of dollars, you don't do that through a spreadsheet. That's a lousy investment, unless the technology trends make it the right investment.
And obviously in the five, six years before I came back, Intel gave $100 billion to shareholders โ dividends and stock buybacks. What I wouldn't have done for another hundred billion dollars. You probably would have made chips for iPhone, which Intel passed on. It hadn't built a new factory in a decade when I got there. How could you not buy EUV machines? There's just all of these things that you would only do as a technologist because the economics behind them by themselves were not good.
So it's getting back to the core of technology โ that was the fundamental thing. You make good decisions, you make bad decisions as leaders, every business does that. But fundamentally this is a technology business and you need technologists running technology, that then hires technologists that are sitting at the staff, that then hire the best technologist โ and take big swings at categories that could matter in the future, like skating to where the puck's going.
If you look at Apple, they did the same thing for the past 15 years, buying back the stock, tremendous amount of dividends. They're the largest holder of capital of any company. And what companies do they buy? Little tiny acquisitions on the margins โ the largest was Beats, because they wanted inroads into certain demographic segments in the Android space they couldn't get into. But my god, what a colossal waste of time. Like you said, they could have done so many amazing things.
Tell me about Steve Jobs in 2008, 2009 deciding "I think we're going to make our own silicon" โ was that a covert product project or did you guys know he was doing that?
I remember when we had the first conversation with Steve about porting the operating system to the Intel chip. We were quite proud of our silicon software competencies. So Steve, we'll help you port the operating system to x86. And I remember Steve said, "I've been working on that for the last four releases." He had been preparing the core technologies inside Apple for something that might happen in the future. I was just shocked.
Well, that seemed to be another one of those forks in the road. Steve was an incredible leader, also a ruthless leader, very difficult. When they moved to Intel and the Centrino chip, it was a big deal โ they were putting extraordinary demands on Intel, make the chip smaller, drive lower power. When he was no longer convinced that we could continue to do that, he started the project โ P.A. Semi, acquired some small companies, started to build some competency, a few little chips internally, and the little chips got a little bit bigger. That's how they got into doing their own semiconductor. "I'm not sure I can rely on Intel to be that much ahead of the industry and I can start optimizing the system design with the silicon design." They sort of said, you failed as a supplier, no, I can supply myself better.
And Jensen decides he's going to go all in making these video cards โ talk about incredible serendipity that these happen to be also very applicable for cryptocurrency and running AI jobs. Was that luck or skill or a combination of both?
When we were at the height of our strength on CPUs at Intel, we sort of scoffed at his machines โ "that's a graphic machine, some gamers who want to use that kind of stuff." But when they started to build a real software stack with it โ this CUDA thing, SIMT as a technology โ it kept getting a little bit better and a little bit better. And all of a sudden the crazy Japanese HPC guys said, "Hey, we could take those graphics cards and maybe start using them in HPC."
That was sort of the defining moment where it wasn't just about doing graphics anymore โ a more computationally dense platform to start attacking some of the world's most interesting workloads. AI had gone through what its fifth nuclear winter by that point โ we're just like, this is never going to matter. But the community around it was continuing to develop and the CUDA software kept getting better generation by generation. I had a project at Intel, Larrabee, where we were trying to take the x86 and essentially do the same thing. In my first departure from Intel, the project was killed a week after I left. The world would have been so much different. As William Gibson said, the street finds its own use for technology โ Nvidia did not predict that this Bitcoin project would take over, nor did they anticipate that AI would take off, but because it was the best solution, the hacker community could figure that out.
TSMC started with a vision of foundry โ they were going to become the factory for the industry. Intel was IDM โ we never worked to make our process and factories available for third parties. TSMC basically said, "I don't care whose chip it is, I don't care what you're designing, I'll be your manufacturing partner." When I came back to Intel in 2001, TSMC was producing 5x the wafers of Intel. Not 10% more โ 5x.
These factories are so expensive, 20, 30 billion dollars, and at the time that foundry model was such a trivial piece of the business Intel didn't even care. Over steady progress and Apple as a customer driving them, the model of foundry became the model of the semiconductor industry โ with two exceptions, Intel and memory. Now it's more like seven to one in terms of wafers to TSMC. That was one of the core thesis of the new strategy โ we must become a foundry as well.
Are we going to be able to onshore that? We had the chips act โ some people believe it's going to happen the year after Trump's out, others believe as early as '27 or '28. Are we going to replicate that here in America, or could this be a cataclysmic event if China decides to blockade Taiwan?
The chips act is having benefit. When we started the chips act in 2001, the US was building about 12% of leading edge. Today that number is more like 18%. Now let's make it ugly for a second โ the island of Taiwan has less than 3 weeks of energy reserves. When you turn off a fab, it doesn't come back on for 90 days. The economic impact of a brownout of Taiwan is greater than the Great Depression.
Never do you need to do anything, a shot to be fired โ you just need to say, "great, no energy for 3 weeks." No oil, no LG, that's how the island runs. That is scary. We need more resilient supply chains. China has blockaded the Taiwan Straits seven times over the last four years. This isn't a theory โ they're running exercises, being pernicious and pretty provocative. Their intentions have been clear over a sustained period of time. We need more resilient supply chains โ something I put a lot of my time and energy into, and we're making progress, but we need to go faster.
Do you think it's a bubble? What worries you โ that we build too much, or that the technology doesn't solve enough problems and we are swimming in tokens? The valuations of these companies have gotten quite extraordinary.
There is a silver lining here that guarantees we don't get too far ahead of ourselves in terms of bubble โ and that is energy capacity. Essentially, nobody's going to build and buy GPUs and build data centers if they don't have energy. So you have an upper bound on how aggressive and how hyped and bubbled that we get. I take a lot of solace in that.
Energy capacity in the world is expanding four, five percent. In the US we had a decade at 1%. It's hideous what we did to our energy grid over about a decade and a half. But now that's getting built out. What then is the incremental value of a token, and if it's a measure of intelligence, it's somewhat infinite โ better supply chain, better finance, more efficient logistics. Particularly with labor shortages, I am an optimist that we're in a couple of decade buildout. Not a couple of years, a couple of decades.
One of the big objectives I've said is that I have to make AI 10,000x better. It's way too expensive today. We want to drop by five orders of magnitude the cost per token, the energy per token, so that we really do have Jevons law โ we just explode the access to AI in much more economic ways.
There has not been a time in human history where it's been better to be a technologist than the one we're in right now. We will solve chemistry, we will solve language, we will invent new materials, new forms of interaction, killing cancer, lifting people out of poverty. There is not a better time to be alive than the one we're in right now, and as technologists we get to sit in the driver's seat of it.
Anytime the multiples get too high, some corrections โ periodic corrections that keep the earnings multiples in reasonable things is good, because this will not be a smooth curve. I'm predicting two decades of goodness and there's going to be lots of disruptions along the way. Every time we have one of those corrections, say thank you โ we're not letting the bubble get ahead of itself. We had the SaaS apocalypse, there's going to be other apocalypses along that journey โ and that's even before it gets exciting, what I call the trinity of computing: classical computing, AI computing, and quantum computing. When those three come together, that's when things get really exciting.
This decade โ this decade โ we will see quantum supremacy results across multiple industries. We know how to build qubits, we know how to error correct qubits, we now have algorithmics against quantum, and now it's just about engineering scale. My prediction is meaningful results before 2030.
You're going to be able to start doing things that cannot be computed today โ chemistry, biology. Some of the easy things will be logistics, the traveling salesman problem. It's probably going to be 2032, 2033 when we solve things like encryption โ the fundamental Q-day implications. You now have four, five, six modalities of quantum demonstrating pretty good results โ trapped ions, photonic approaches, spin approaches. Modality is not an issue, error correction's been proven across them, and the race will be on.
Oika is one of my favorite founders. He's built a product that people are addicted to โ primarily Anton, the people who work for me โ and I love talking to you because as a founder you have a north star, incredibly laser focused on enabling anyone to build great software.
The first gap is to build a product. The second gap is to build a business around that product. The first one, we got very far โ we're seeing a million new projects built every single week. On the second one, we're investing a lot in making it easier to run your business. We're seeing more than 700 million visits to the applications every month. More than 50 million apps built on the platform to date.
How many years has Lovable been in market? 20 months. We're seeing people who are first-time founders, enterprise leaders move much faster together with their teams on this platform that has a lot of opinionated pieces in how you should create software and how to operate that software.
Who is the customer โ do developers use Lovable, or is it the other 95% of society?
About 20% are technical or some type of engineer, and they love that we're quite opinionated โ we put all the best practices into how the software is architected. Four out of five are nontechnical, and they're building often first to figure out what is the right thing to build. Now we're seeing people running businesses making more than million dollars of revenue on this platform.
Many engineers don't look at the code, they don't write code anymore โ that means you don't need to be an engineer to create software. But Lovable creates a structure for the architecture that makes sure you don't go off a cliff โ payments, emails, getting discovered by other AI chat engines and Google search are kind of taken care of.
For one of our projects, founder university, someone wanted to make an internet. This is something I would have never okayed because it would have cost $500,000 ten years ago. In 4 to 8 hours, they made the whole internet โ and she did it on her own without permission, on Lovable. She said, "I just put it on my corporate card." Lovable is 50 bucks a month.
The economic impact โ what you built to us would have cost me $500,000 two years ago. It was built in 4 hours by an employee, which if you put employees at $50, $60, $70 an hour plus the cost of your software, it got made for less than $2,000 in a year. Then she asked Lovable to build the economic impact of the 50 companies in the program โ how many people work at each company, what taxes they pay, how much they rent their home for, what their average salary is. Built something I would have never been able to afford to build.
NAD works at a pretty large company, Nursa. He came to our platform because he wanted to build out a new product line โ nurse study for educating more nurses. He built out all the admin tools for the program, the scheduling, licenses and certification management. He also took it into the back office internally and they've now replaced more than 10 tools that they had bespoke applications for. In their case they're saving more than a million dollars per year.
What we're already doing โ I've been doing for a very long time โ is to compound from everything we're learning every time Lovable makes a mistake. It goes to an agentic system with our engineers in it improving it. That compounding intelligence is applicable to our customers running their business on our platform as well.
Is software going to become 100% bespoke, even internal tools? How do you think foundational pieces of software โ Salesforce, HubSpot, Slack, Google Suite, Microsoft Office โ will bespoke software start to replace those?
People have announced that Lovable's dead every 6 months since you started, and then every 6 months you go from 100 to 200 to 300 โ I think you're at 400 million in revenue. "We reached 500 in May." So you're dying again by another 100 million in annual revenue.
We've always had this strategy that we do whatever is best for our customers. In terms of intelligence, that means we're using multiple models โ both commercial frontier models from multiple vendors, and increasingly open weight models. We have a really strong research team up in Stockholm working on post training, applying all the best practices, because we also believe it's part of the European ecosystem to have that capability in Europe specifically.
We're looking at the mistakes that any of the models do right now, then we prioritize them by what drives most impact for our customers, and then we create data sets or do reinforcement learning specifically for the problems where the frontier models are making mistakes for us. We have this enormous token distribution from a million new products being built every single week.
We have caps and overages. Our customers definitely hit caps. From the lowest subscription tier, it's something like 60% of our customers. We always monitor our margins, but we're doing what's best for our customers โ we've never had the decision to say let's use a cheaper model if it's measurably worse.
I'm hearing that people are willing to pay the overages because they're getting so much value. If I'm paying $600 and if you top max to $6,000 a year, but this is a $500,000 piece of software, I don't care โ I'm still paying somewhere between 0.1% and 1% of what I would have paid 3 years ago.
What is your take on this latest version from Anthropic, Fable? I know they're a partner.
We use Fable as well as one of the models. Is it a massive step function? Yeah. What I've seen is that it can in the first attempt create very sophisticated things that look really good โ like it creates really beautiful things, 3D games for example. But on figuring out what the right strategic directions or experiments you should run to improve outcomes for your business โ that's not changing as fast as the humans knowing how to use the tool to plug in all the right data to take the right decisions.
Do your customers โ multiple people in the organization try to solve the same software problem and they're competing with each other. Is that the right thing to do?
I worked at a place called CERN where they do particle physics. That's where I was introduced to this concept of co-opetition โ two isolated teams working on the same particle accelerator but different places on it, and they don't share results until they publish. Now that engineering is less of the bottleneck, it's more the question of what is the right thing to build. I think it's a great thing to have sufficiently many humans try to solve the same problem in different ways, then take the best three things from each and bring them together, or run a split test.
Congratulations on being reborn six times, because every 6 months you add 100 million in revenue it seems. And then everybody says Lovable's dead because the new foundation model is so good. But you keep studying your customer and you keep somehow surviving and thriving.