The Story Behind Cerebras’ $63 Billion IPO with Founder and CEO Andrew Feldman
In a Nutshell
Cerebras built a wafer-scale AI chip 100-200x larger than GPUs to deliver 15-20x faster inference, betting on a fundamentally different architecture rather than incremental GPU improvements. The company struggled for years with minimal adoption until 2025 when AI became useful enough for daily work, triggering explosive demand from OpenAI, AWS, and others that led to a $20B+ deal and $63B valuation. CEO Andrew Feldman emphasizes that radical speed doesn't just accelerate existing processes—it enables entirely new business models and productivity leaps once infrastructure catches up with capability.
These notes were generated by AI and may contain inaccuracies.
Netflix used to deliver DVDs and envelopes and when the internet got fast they became a movie studio right it opened up an entirely new business something fundamentally different that's what happens with speed and I think that's what fast AI does right now we're replacing things that everybody can see like coding design the SAS tools but once we start sort of fundamentally reorganizing around this you're going to see this sort of new business models and fundamental jumps in productivity and I'm eager for that.
Today on No Prize, we have Andrew Feldman, the co-founder and CEO of Cerebras. Cerebras was founded in the mid 2010s to focus on new workloads for AI, particularly the machine learning world, and then has made the transition into very fast inference for the foundation model world that we live in today. Cerebras recently went public and is currently worth about $63 billion in the stock market.
Andrew Feldman stated: "We build AI computers, right? Computers designed to and optimized to accelerate AI workloads."
Right now Cerebras is the fastest at inference, not by little, but by a lot, 15, 18, 20x faster than GPUs. Starting in about 2025 AI models got smart enough to be useful. People began using them and we make AI with training and we use it with inference. As people began to use it, it began to be integrated into their day-to-day work. Speed became fundamentally important and Cerebras was crushed with demand.
The speed advantage is faster across the board: big models, small models, US models, Chinese models, trillion parameter models, one billion parameter models.
At the end of the year, Cerebras signed a deal with OpenAI, one of the biggest deals ever in Silicon Valley, north of $20 billion. In March, Cerebras signed an agreement with AWS where they will be deployed in their data centers going forward. It was a whirlwind year and a half of chasing supply and trying to meet the demand.
Andrew Feldman explained that they built a really fast machine and for a long time nobody cared. Many people objected and said this is just a weird architecture. They called it wrong. To be radically better, you can't build something that is a similar architecture. You're not going to get 15 or 20 times better than the GPU with a minor modification to their architecture. If you're going to aspire to a radical improvement, your design has to be different.
From the beginning, Cerebras chose wafer scale, which means they build a 46,000 square millimeter chip, a chip the size of a dinner plate, whereas everybody else is building chips the size of postage stamps. They told them they were out of their mind. It would never work. They listed reasons why it was impossible. But in 2019, they proved it was possible. They began delivering it and they improved on it.
"But we were fast when AI was a novelty and when it's a novelty, nobody cares that you're fast because it's not being used."
From about 2023 to the beginning of 2025 people pointed at AI, but nobody used it every day in their work. Once you use something every day in your work, it can't be slow. How long will you wait for a website to resolve? The market for slow search is zero. The market for dialup internet is zero. That's how big the market for slow inference will be.
Cerebras had to wait until AI was smart enough to be useful. That happened in 2025. That's why there was an explosion of demand from companies like Cognition, Cursor, Lovable, OpenAI, and others. Cerebras was right there with the right product.
Andrew Feldman first met the interviewer back in 2016 when people weren't even saying AI, it sounded weird. They were talking about machine learning. The models of the time were convolutional neural networks and RNNs and the emergence of GANs. They were trying to tell the difference between a chair and a cat.
What gave the team conviction back then was a combination of vision, the right co-founders, and a little bit of arrogance, a little bit of luck. They saw AI on the horizon as a new workload and as computer architects, new workloads are opportunity. It's very hard to enter in the x86 world where nothing new is happening. When graphics emerged, you got the discrete GPU and Nvidia. When mobile compute hit, you got ARM. Intel, AMD, and others who would have been well positioned got no share.
The team knew that this new workload would eat a lot of compute. It would require a new architecture, a dedicated architecture, and it ought to be very different. The architecture could not be a derivative of what's existing. Those were their big bets and they were 100% contrarian.
There were moments of doubt. They were solving a problem that had never been solved before. There had been efforts across the entire 70-year history of the computer industry to build a wafer scale product. Gene Amdahl, one of the fathers of the field, failed miserably to do it.
Cerebras had a period between about middle of 2017 and middle of 2019 where they couldn't build it. They were spending about $8 million a month. They were having board meetings every six weeks saying they couldn't build it. Each time they did a failure analysis, they got a little bit better at it. In the summer of 2019, they yielded it and it began to work. The first time they were sitting in a little makeshift office in downtown Los Altos in a building that was not designed for hardware guys and they were staring at a computer which is about as exciting as watching paint dry and it's working and they couldn't speak for half an hour.
They solved the hardest problem in the computer industry and nobody cared. The first gen they might have sold a dozen. The second gen they probably sold 300. The third gen, they had a 2 or 3 year period where they were ahead of the market and absolutely nobody cared that they were blisteringly fast.
They found some pioneering customers that were atypical. There were some sovereigns who really bought ahead. They won at Argonne National Labs, Lawrence Livermore, Sandia, and in Europe at European Parallel Computing Center at LRZ. They won some customers in the oil and gas space and some in pharma, all of whom have long histories of using extraordinary amounts of compute. But then historically there's this giant chasm because none of them provide the volume to get to mainstream.
Cerebras won a sovereign G42 and they became a strategic partner and close friends. Pang is CEO of G42 and his chairman is Sheikh Tahnoon. G42 placed a billion dollar order on Cerebras and with that they were able to transform the company. They were able to change their supply chain. They were able to deploy equipment in big enough clusters that they could battle test at scale. One of the challenges in hardware is your QA lab can't be as big as some of the customers you want to deploy to. You can't put $100 million in your QA lab worth of your own gear. G42 worked with them and they began training models for them and doing inference for them. They've been an extraordinary partner.
When OpenAI came along and when AWS came along, they had the capacity. They were ready. They'd battle tested. They'd gotten over the chasm. They'd had a bridge and so they could meet the demand.
It's years of work. Often in the software world you can scale so fast, but when you're building things, you have to double, you have to call your manufacturing partner, your CM. They have to find power. They have to rent a building. They have to add more lines. You have to make test fixtures. Each step takes real time and effort to grow. Cerebras is going to try to increase manufacturing 10x this year. That's about as fast as anybody in the history of hardware.
One of the co-founders said it was going to take about 10 years to build a compiler. Andrew Feldman said no, that's crazy, that's big company talk, we can do it in five. It takes about 10 years. It takes a long time to build a compiler. It's an extraordinarily difficult piece of software. Now they've got a good software stack.
AI generated coding is hugely relevant for Cerebras internally. Eight months ago they weren't spending $1,000 an engineer on tokens and they're probably at $25,000 or $30,000 right now and it's ripping. It's not useful for everybody. There are some people who have the perfect mindset for it. They are running eight or 10 agents 24/7. They've moved their coding style to being one in which they govern agents. They think about how to QA, so they've got a QA agent running. They think about how to remedy some of the weaknesses in the coding models. They're often verbose, they often cut out comments. They've really thought about it and it's a type of puzzle that is the perfect fit for their mind and they've gone from being 10x guys to being 100x guys.
The rest of the team is limping along. They're trying to figure out how they can make it work for their different jobs, for being the CEO, for being the CFO, for being accountants, for being in marketing. But for a small number, it is such a tool. The rest try and show them what others are doing, what best practices are.
Cerebras is about 800 to 850 people now. It's a lot of market cap per person. When thinking about where to go from here, with a backlog that's north of $20 billion, delivery is pretty important every day. They have to continue to be fearless. One of the malaise of companies as they get to 1,000 to 2,000 to 3,000 people is they stop taking the type of risks that they were taking before. You move from being a fearless engineering culture to being what can we get in the time frame of the next rev and that's extraordinarily damaging.
Cerebras takes pride in doing fearless work. They want to hire people who do fearless work. They want to guard that culture that says they would much rather fail in pursuit of the extraordinary than succeed in the ordinary. Recruiting is a worry. You have so many openings and it's so easy to settle and it's so easy to just try and put a butt in a seat. That is death. Andrew Feldman spends a meaningful part of every day talking to candidates.
Being CEO is an extraordinarily lonely thing. You're building a business. Being a leader is lonely and it's not easy and people don't like to say that especially for those who like to solve problems specifically the problems everyone else says can't be solved. You gain fire from that chip on your shoulder. When they say it can't be solved you say in your head you can't solve it.
You have to love the journey. The things they do are too hard if you don't like the building. Doing this for the money is a horrible thing. There are way easier ways to make money than trying to create something extraordinary and compete with somebody as strong as Nvidia. That is not the easiest path. You got to love being a David. Andrew Feldman is a professional David. This is his fifth startup. He competes against Goliath. That's what he does for a living. Every dollar, every million dollars, every billion dollars they sell, if it wasn't for their brains, their muscle would have taken it in a heartbeat. You got to love that. If you don't love that, it's a very long road.
The right time to give up is when you've laid out a set of hypotheses about what it's going to take to win and they all come back negative. People often do this sequentially. They say they just need to test one more thing. They test and it doesn't work and they say they need to test one more. The slippery slope is a beast. The slippery slope in all things, in ethical situations, in your life, is something you have to guard against.
Sometimes having other former CEOs or other really seasoned entrepreneurs who are on your side and who can share with you, remember a year ago you said if you got to this point you didn't have this and to remind you so they pull you back off that slippery slope. They said the old frog in the warm water thing is like you said if it got this hot you were going to get out and it slowly kept getting warm. It's basically can other people keep you effectively accountable towards both directions, accountable to your own thinking. If you understand why it's not working, if there are some things that you can articulate that have to change in order for it to work and you can put some sort of time frame on it. Lots of efforts ought to be truncated and those people sort of redeploy their efforts to new and different ideas that they have.
Going public is exchanging some professional investors, venture capitalists who specialize in technology investing for a different class of investors and in so doing reducing your cost of capital a little bit. Suddenly you go from pros to my dad. That's the trade-off. In return for that, you have to agree to be governed by a set of extraordinarily stringent rules.
There have been for the first time in history four or five companies that can raise huge amounts of money without going public. This was never a thing before OpenAI and Anthropic. The four-year option package timeline from Silicon Valley comes from how long it would take you to get public. It used to be four years and that was the way you got a valuation in the hundreds of millions. Now people have a tender cycle at a certain scale. It took Cerebras 10 years.
Cerebras opened up the secondary market and let people sell. If you're going to bet big chunks of your career with them, they thought it would be perfectly reasonable for you to find modest liquidity as you went along. You have to think very differently if it's going to take you a decade.
For a very small number of companies, they've been able to raise public market money at public market valuations in the private market. For the rest of the world if you want super high valuations, if you want the legitimacy that comes with it, historically large companies like doing business with other public companies in the US and you get a credibility and a legitimacy from having your books audited, from them being able to see who you are. That's different than when you're private.
Cerebras could offer the public market something unique. They would be the first and only for a period of time AI pure play. They are the only company where 100% of the revenue is this exact market. There's no gaming, there's no graphics, there's no PC. This is it. That was an opportunity, a differentiator that they thought was interesting. For them it was an opportunity to graduate from corporate adolescence to corporate adulthood.
Andrew Feldman spoke to Sam in middle of the summer in 2025 and he said for the first time they've been trying so hard just to keep up with demand. They now see the importance of fast inference. That produced a set of trials and some testing that was done. Cerebras was so much faster than the competition. It felt really good.
Andrew Feldman loves talking to super smart customers. He has a rule that if his mother buys it or uses it, he doesn't want to make it or sell it because he really wants super smart customers who are doing really interesting things with their stuff. They got in with some of their guys and they were like whoa, this is we understand now. At Thanksgiving, the night before Thanksgiving, they signed a term sheet and four weeks later on the 24th of December, they signed a big master agreement. They can fly and they were working seven days a week. They had several law firms.
A $20+ billion deal completed in four and a half weeks demonstrates exceptional market velocity. This pace reflects a new characteristic where participants across the industry are working to keep up with demand. Deals that previously seemed impossible at traditional speeds are now achievable—such as acquiring a $300 million company in 24 days when working eight to ten hours daily on the transaction.
The art of the possible has expanded significantly. Ambition for speed becomes a major advantage when paired with the belief that faster execution is achievable.
Several operators in the current market have built remarkable companies with growth rates previously considered unattainable. Examples include Cognition and Cursor, where development velocity and scaling have exceeded historical limits. Data center construction and acquisition timelines that were once viewed as fixed constraints are now being compressed.
The growth of the open source ecosystem has enabled a new generation of companies to achieve impressive results. Open source solutions kept interest and momentum alive during periods when closed source options were prohibitively expensive. This competitive dynamic pushed established players to remain ahead of rapid advances, including techniques demonstrated by Chinese developers.
The result is a vibrant, creative ecosystem where ideas from external developers can run effectively on specialized hardware. Infrastructure that supports other people's innovations provides a meaningful advantage in this environment.
Speed does more than incrementally improve existing processes—it fundamentally changes what is possible. Historical precedent shows this pattern: Netflix transitioned from DVD delivery to becoming a major film studio once internet speeds enabled streaming, rather than simply optimizing its original model.
Fast AI is expected to create entirely new business models rather than only replacing current functions. The PC initially replaced typewriters and manual accounting ledgers, but the larger productivity gains came from reorganizing work processes. Cloud computing then enabled SaaS, which in turn made previously expensive tools accessible to smaller teams and individual companies.
Current AI applications are visibly replacing tasks such as coding, design work, and certain SaaS functions. The larger opportunity lies in reorganizing operations around AI capabilities, which is anticipated to unlock new business models and substantial productivity increases. These transformations are expected to emerge over the coming years.
Keep No Priors: AI, Machine Learning, Tech, & Startups in your library
Save the videos and channels worth coming back to, and find them again in one place.
