About
Steve invests at the intersection of technology, design, and business: at the sweet spot…
Activity
16K followers
-
Steve Vassallo shared thisI just reread the original Cerebras Series A deck. 10 years ago, I gave Andrew Feldman his first term sheet off the back of this. The cover says "Cerebras Systems (working name)." They were so early the name wasn't even final! It was Andrew, Gary Lauterbach, Sean Lie, Michael James and Jean-Philippe Fricker and an idea most people thought was nuts. Even the language is from a different era. The deck talks about neural networks, CNNs and RNNs. No transformers, LLMs or foundation models. The vocabulary we use today didn't exist yet. But the team's two bold bets proved exactly right. Bet 1: AI workloads were about to explode. In 2016, this was not obvious. The smartest people I knew were divided on whether AI would even be useful. AlexNet had arrived in 2012 and a few corners of the research community were doing interesting things with image recognition, but in the broader software industry, AI was somewhere between a marketing buzzword and a science project. Andrew and his co-founders saw the demand curve going vertical. The deck makes this case page after page. Google had just used 1,000 servers to train on a single 224-pixel image. Meanwhile 350 million photos were being uploaded to Facebook every day. Training a neural network still took months. The infrastructure was not going to keep up. Bet 2: The GPU was the wrong machine for the job. There's a slide in the deck titled "Why Would an Engine Built for Pushing Images to a Monitor Be Ideal for ML?" At this point, everyone else was iterating on the GPU. Andrew kept coming back to a different idea. He thought the hard part of deep learning was going to be moving data around, not doing the math. Only 4% of a GPU's die area was being used for ML compute. Yes, only 4%! The rest was carrying the burden of graphics. It was a battlefield promotion - not what anyone would design starting from a blank sheet of paper. So while the rest of the industry kept optimizing the GPU, Cerebras designed something from scratch that was purpose-built for neural networks. Optimized for the movement of data across a fabric rather than the multiplication of matrices in isolated cores. They proposed a different kind of computer altogether - one that broke with how computers had been built for 75 years. Pulling it off meant inventing new technology across semiconductors, systems, data fabric, and software all at once. Any one of those could have been a company on its own. They signed up for all four. Inference doesn't come up in the deck because it couldn't have. In 2016, nobody was running models at scale. Inference wasn't a job anyone needed done yet. What came next was years and years of wrestling with problems that no-one in the history of computing had ever solved. 10 years later, Andrew and team were proven right. Proud of what they’ve built. And proud of the journey we’ve been on together.
-
Steve Vassallo shared thisProud to be on this year’s Forbes Midas List 100. But the list isn't really measuring what people think it's measuring. - It measures Andrew Feldman, building wafer-scale silicon when the industry said it couldn't be done. - It measures Anatoly Yakovenko, betting on a new way to build blockchains. - It measures Pierre-Damien Vaujour and Alex Greenberg, going all-in on satellite buses before they had clarity on the future. - It measures Lynn Jurich and Edward Fenster, rebuilding Sunrun overnight when the 2008 GFC broke the model. These are the people who move the world from what is to what ought to be. And I feel lucky to play a supporting role. I love what I do and I love who I get to do it with too.
-
Steve Vassallo reposted thisSteve Vassallo reposted thisWhen the chip on your shoulder is the largest chip the world has ever seen.
-
Steve Vassallo shared thisI love to back founders with a chip on their shoulder.
-
Steve Vassallo shared thisSocial media says design is dead. The data says otherwise. In partnership with Designer Fund, we surveyed 900 designers in tech and interviewed design leaders at companies like Anthropic, Stripe, Linear, and more. Here's what we found. Designers are doing more, owning more, and shipping more than they were a year ago. Half of all designers - ICs, founders, and design leaders - have shipped AI-generated code to production. The question is no longer should you ship product. Designers who have leaned into AI are 2x more likely to feel creative, confident, and capable. The bar is now higher and those who are experimenting are the ones clearing it. The AI toolkit more than doubled in a year. Designers used 3 AI tools in 2025. They are using 7 now and some are even building their own. Inside the report you’ll find: - The top 7 AI tools designers are using in 2026 - The 2 new skills design leaders are screening for this year - How to build a culture that makes designers 2x as creative and capable Read it here: stateofaidesign.com Huge thanks to Ben Blumenrose, Robyn Park and the Designer Fund team for driving this.
-
Steve Vassallo shared thisA year ago, Ben and I went out on a limb. Now it’s a tree.Steve Vassallo shared thisWhen we released the AI in Design Report last year, the response took us by surprise. In retrospect, it shouldn’t have. Everyone was feeling a wave of change coming for every role and process while we all clearly lacked best practices or guidance to manage that change. While the 2025 report provided some guidance and data, it also left many questions unanswered (frankly we simply didn’t have the answers yet!). So earlier this year we began wondering: should we run it back? And this time try to find the answers that were so elusive last year? We went back and forth on it but ultimately the decision to do a report this year came down to two big considerations - 1. More data – another report would give us an additional data set so we could spot trends. One report gives you dots, two gives you lines. 2. Best practices – another report would give us an opportunity to package best practices from top design teams. This would begin to answer the HOW question we kept getting again and again. So we started to pitch this idea of a bigger, deeper report + case studies to the very best companies we could think of, our dream partners basically: Anthropic, Stripe, Notion, Linear, Shopify, Sierra and Framer. We told them to partner on this you’d have to support the work AND open up how you work internally to the broader design community. The response was a resounding “count us in!��. With that momentum we then brought in incredible freelancers and agencies to help with research, branding, video production, writing, and website design and development. We went all in. The result is the culmination of thousands of people hours and many late nights to create what we believe is the most comprehensive, well-researched report capturing and synthesizing the state of Design + AI today. While we used AI in many areas (and we'll share some of the process in the near future), doing this kind of work still required deep thinking, grit, and humans coming together to do what they do best. The report spans nearly 20k words covering the survey results of over 900 people paired with dozens of qualitative interviews. Over the coming months we will also release 7 beautiful case studies of how the best teams are working on the ground. It is a true labor of love to help guide a design community we hold so dear. You can read it now at – https://lnkd.in/dyaHFNrf Lastly, let us know what you think: what speaks true to you? What don't agree with? How are you using this report to help your team? This feedback is a gift and will help shape our work for years to come...
-
Steve Vassallo shared thisCerebras just IPO'd in the largest semiconductor IPO of all time. 10 years ago, their first office was on the 2nd floor of Foundation Capital at 250 Middlefield in Menlo Park. I still have the floor plan our office manager drew up at the time. Cerebras sat next to Bill Elmore, one of Foundation Capital’s founders and a few doors down from Bhavin Shah, who would go on to start Moveworks. I can still picture Andrew, Gary, Sean, JP and Michael whiteboarding dataflow architectures. Back then, they set out with a big dream: To build the first wafer scale chip in the history of computing. To put it in perspective: the largest chip prior to that was about 820 square millimeters - roughly the size of a postage stamp. Cerebras wanted to build one 56x larger. 46,000 square millimeters. The size of a dinner plate, drawing nearly 20 kilowatts of power, when the previous record was 600 watts. Nobody had ever done it, which meant no playbook existed for any of it. Our first prototype went up in smoke on power-up. The team called it a "thermal event" - which is I guess what you can call a fire when you don’t want to scare your board or your landlord. No existing supply chain could produce a chip this size, so we had to convince TSMC to build entirely new processes alongside them. Then we had to convince the rest of the supply chain to do the same - manufacturing partners who had never handled a chip this size, cooling and heat sink vendors who had to invent new thermal systems, power delivery suppliers who had never pushed 20 kilowatts through a single piece of silicon. And then we had to go sell it - in a world that wasn't quite ready to believe in dedicated AI compute. This is where Andrew's particular strengths kicked in. Andrew is always willing to get on a plane and go meet customers, no matter how far the destination. He was in the Middle East talking to early UAE customers years before it was cool. At the time, NVIDIA was too big to think the trip was worth the bother. Today, that UAE customer is one of the largest buyers of AI compute in the world. And 10 years on, Cerebras is one of the most important AI infrastructure companies in the world. But it started as 5 people in small offices above mine with a whiteboard and a dream most people thought was impossible. To every founder right now unpacking boxes in your VC’s office: Use this as proof that you can dream bigger. You can try to build things that have never been done before. And you will stumble in spectacular ways along the way. Your first prototypes might literally catch fire. Suppliers will tell you it can't be made and most VCs will tell you it can't be funded. Keep going.
-
Steve Vassallo shared thisIn April 2016, I threatened to climb over Andrew Feldman's fence to give him his first term sheet for Cerebras. It was April Fool’s day, but I wasn’t fooling around. The story started in October 2007, when Andrew and his co-founder Gary Lauterbach had just started SeaMicro. Even then, Andrew was a force of nature. He was extremely intense and miswired in all the right ways. You could feel the sparks flying off him. We didn't invest in SeaMicro, but we stayed in touch. Andrew and the team built SeaMicro then sold it to AMD in 2012. When AMD acquired SeaMicro, I had a hunch Andrew wouldn't last long inside a big company. He has, as I've said many times, immense ambition and a heart full of disobedience. By early 2014, he was looking for an exit. Over the next year and a half, Andrew and I met 6 or 7 times. Sometimes in our office. Sometimes at a coffee shop in Portola Valley. Sometimes at our local tennis club. We kept coming back to one thing: deep learning workloads were growing exponentially, and traditional compute architectures couldn't keep up. GPUs had become the default for neural network training, mainly because researchers had accidentally discovered they were less terrible than CPUs. Andrew, Gary and Sean saw the GPU for what it was: a battlefield promotion of a chip optimized for graphics. Better than a CPU, but not what anyone would design starting from a blank sheet of paper. Their key insight was that memory bandwidth, not raw compute, was the real constraint on what neural networks could achieve. So Andrew, Sean, Gary, Jean-Philippe and Michael set out to do something nobody had pulled off in the 75-year history of semiconductors: Build a wafer-scale chip the size of a dinner plate. In April 2016, I asked Andrew if we could be his first term sheet. Eric at Benchmark and I co-led the round along with Pierre Lamond from Eclipse. Then the hard work began. In the 75-year history of computing, no one had made wafer scale work. Which meant no one had ever had to solve the problems that came from trying. How do you power a chip that large? How do you cool one? How do you maintain electrical continuity across tens of thousands of connection points on a single piece of silicon? To get there, Cerebras had to invent in nearly every modern computing discipline at once: semiconductors, systems, data fabric, software, algorithms. Each was a startup in its own right. Their first wafer self-destructed on initial power-up and Andrew and the team were back in the lab the next morning, identifying what didn’t work and coming up with approaches to solving it. Yesterday, Cerebras went public. 19 years after our first meeting, 10 years after that April Fool's term sheet, they’ve built a generational AI company. From a coffee shop in Portola Valley to ringing the bell at the NASDAQ. What a journey. Proud to have been Andrew's first partner in Cerebras. Even prouder to call him my friend.
-
Steve Vassallo shared thisToday, Cerebras IPO’d. Andrew rang the bell at NASDAQ. I was proud to stand alongside him - just as I was 10 years ago when he, Gary, Sean, Jean-Philippe and Michael were unpacking their boxes on the 2nd floor of our old office. The bell at Nasdaq is the public version of their success. But the moment that made today possible looked very different. It happened in a small lab in August 2019. The Cerebras team had just powered on the first wafer-scale system ever built. Andrew, Sean, Gary, JP, Michael and a handful of engineers stood watching a computer do nothing visibly interesting. About as engaging, in Andrew's telling, as watching paint dry. The difference here was that this paint had never dried before. They had just powered on the first wafer scale chip in the 75-year history of the semiconductor industry. They stood there and watched it for 30 minutes, and then got back to work. This rings true of most frontier breakthroughs. They happen in quiet labs, in front of a handful of engineers, and only get recognized for what they were years later. For the 3 years before that morning, our board met every six to eight weeks. The team walked us through what they'd tried - a new variant of packaging, a new approach to power delivery, an adjustment to thermal management. What they thought had gone wrong, and what they planned to try next. Then we’d spend the next 6-8 weeks, on the phone, in the lab, on planes - bringing whatever people, resources and relationships we could to help wrestle the next problem to the ground. Each solution revealed the next thing that needed solving. There's no transactional version of work like this. There's only the long version - staying in the room through the partial solutions and patient explanations, so that when it does work, you're there to see it. Today is the public version of that August morning. I couldn't be prouder to be standing alongside Andrew, Sean, Gary, JP, Michael and the whole Cerebras team. What looks like an overnight success from the outside was, from the inside, a thousand problems no one had ever solved - solved one at a time. They earned every inch of today.
-
Steve Vassallo liked thisSteve Vassallo liked thismy smartest friends talk about preserving "option value" so they are always in a highly adaptable position. my most successful friends talk about how "the earlier you commit, the earlier it compounds" and how there's a cost to holding options. the overlap between my smartest and most successful friends is smaller than you'd think. this all feels akin to "concentration to create wealth, diversification to preserve it." the only difference is that group (1) is trying to create wealth while group (2) has already made it.
-
Steve Vassallo liked thisSteve Vassallo liked thisWe’ll pay you $10,000 if you help us find our next Founding Engineer at Hobbes. We build self-improving agents that run autonomous demos + sales for B2B teams. The simple version: replace “book a demo” with “start demo.” The real version: build production agents that deeply understand a customer's solution, adapt to different buyers + context and improve from every interaction. The hard problems span world-models, real-time conversational AI, generative UI, self-improvement loops + evals, sandboxes and agent reliability in very messy customer environments. We’re seed stage, $6M raised, have a really cool customer base and are in person in San Francisco (North Beach). Good fit if they: - have shipped production software to real customers and love working closely with them - obsess about taste and product quality, down to the pixel - would rather own a hard ambiguous problem than inherit a perfectly scoped ticket Top-level comp, founding-caliber equity, gym membership + longevity panels covered, meals comp'd, unlimited coconut waters and more. JD in comments. Just DM with their profile and we'll take it from there.
-
Steve Vassallo reacted on thisSteve Vassallo reacted on thisBack when I was in my twenties, I came across a picture of me when I was crying as a little kid. My parents had been taught to feel uncomfortable about the emotions I was having. So to feel less uncomfortable, they would tease me. It was quite an image: You could see the dumbfounded expression on my face. I couldn’t believe my parents had the camera out. I realized then: “That must be why I haven’t cried in nearly fourteen years...” After I saw that picture, I put it on my desk, and told myself: “I’m going to learn how to cry again.” A whole year went by of that picture sitting on my desk… but I still hadn’t figured out how to cry. I just couldn’t get the tears to come. So I decided to try something different. I went out into the woods to a faraway trail where no one could see or hear me (that’s how much shame I had around it) and I started fake crying. I did that for about three months, just faking it. Until all of a sudden, it started to actually happen: I began to cry. It was one of the biggest reliefs of my life. My body let go of years of tension in just days. I just let myself cry for nearly four days straight — while I was brushing my teeth, while I was eating lunch, whatever I was doing. The most beautiful part I discovered was that underneath all that sadness was a deep, persistent joy and a deeper capacity to feel and love.
-
Steve Vassallo liked thisExceptional. Just exceptional. At Cerebras we are proud to be Windsurf’s partner.Steve Vassallo liked thisThere’s probably no other screenshot that summarizes the hard work at Cognition than this one Before Windsurf was acquired, it took 2.5 years of grinding, creativity, and luck, to get us closer and closer to $100M… But Cognition just added more revenue than where Windsurf was, all in the last month Right now we are operating at another scale, and all credit goes to an incredibly talent dense team
-
Steve Vassallo liked thisSteve Vassallo liked thisIn 2019, Cerebras solved the hardest problem in the computer industry - and nobody cared. I sat down with Elad Gil and Sarah Guo to talk through the 10 year journey of building Cerebras into a public company: Some highlights: - How we signed a $20B+ deal with OpenAI in 4.5 weeks - The 2-year stretch where we were burning $8m/month and couldn't even turn our chip on - How we solved one of the hardest problems in computing and for years, no one gave a sh*t - What open source is really doing to the AI ecosystem Watch the full interview here: https://lnkd.in/egh4gYrp
-
Steve Vassallo reacted on thisSteve Vassallo reacted on thisTwo weeks ago, I was inside the Nasdaq for the largest tech IPO since Uber. Here’s what it looked like from the inside: After 10 years of incredible work, Cerebras IPO’d. If you have ever built something, you imagine a day like this as the finish line. The day your crazy idea is recognized by the world. Here is how it actually went, hour by hour. 7:30am The day started early for the CEO Andrew. In this case, on Squawk Box (CNBC). You want to tell your story to as many networks as possible - this is the day that everyone wants to talk to you and you can own a news cycle. 8:00am About 50 people from the Cerebras team gather for breakfast inside the Nasdaq. Many of them had devoted the best part of a decade working at Cerebras. Their partners, parents and kids were there too. You could see how much this meant to them. Many of the early investors were there too, like Steve Vassallo, Eric Vishria, Lior Susan and Brad Gerstner. 9:30am The Cerebras team rings the Nasdaq bell to open trading for the day. Ringing the bell opens the market for everyone else, but the company's own stock does not start trading yet. That comes later, after a lot of work behind the scenes. Once you've taken pictures, you all head out to Times Square and take pictures in front of the iconic billboards. 10:30am to 12:30pm The investment banks and the Nasdaq are "building the book." What that means is they are matching buyers with sellers before the first trade. The goal is a deep order book. They want enough orders lined up on both sides so the price does not swing wildly the second it opens. 12:30pm The management team heads over to one of the investment banks (Morgan Stanley), and Andrew, the CEO approves the first trade. Cerebras is a public company! 1pm to 5pm Much of the team goes home and rests. Andrew, the CEO, does press, press, press. Fox, CNN, WSJ, TBPN, everyone you can think of. A new one every 15 minutes. While we were waiting to go into one interview, one of the presenters said, "I'd never heard of Cerebras until earlier this week." They had just IPO'd at >$40 billion! It is a reminder to everyone that no one remembers you exist. You have to remind them, constantly. 6pm. Party time. Sarah Chieng fed me more shots than I’d had in years. It is an iconic moment in a company's history. In Andrew's words, “a graduation from corporate adolescence to corporate adulthood.” Congratulations to the Cerebras team, Julie, Sneha, James. The journey has just begun. 📷 Credit: Vanja Savic
-
Steve Vassallo reacted on thisSteve Vassallo reacted on thisYou can’t be what you can’t see, which is why this piece feels so personal. This week, Fortune looked at how the CFO role is being redefined in the AI era through the experiences of Susan Li, Amy Hood, Anat Ashkenazi, Ruth Porat, Hilary Maxson, Colette Kress, and me. And we see you Amie Thuener (O'Toole) Jean Hu Simona Jankowski, CFA and many others! As I shared in the article: “I don’t think of this as a story about ‘female CFOs.’ I think it’s a story about a generation of leaders helping redefine the CFO role, and many of them happen to be women. The role today is far more than managing numbers. It’s about building companies through complexity and change, staying curious, adaptable, and kind.” AI is changing how companies build, invest, operate, and serve people. For CFOs, that means connecting innovation to capital allocation, governance, trust, and long-term responsibility. At OpenAI, that responsibility is especially meaningful because our mission is to ensure AGI benefits all of humanity. Seeing this group of women in the CFO seat at such a defining moment for technology moved me. Representation matters, and while there is still much work to be done, I hope the next generation of leaders sees this not as surprising, but as expected. Thank you to Sheryl Estrada and Fortune for telling this story with such care. https://lnkd.in/guKevuCPFive giant hyperscalers—and Nvidia—share a surprising trait: female CFOs | FortuneFive giant hyperscalers—and Nvidia—share a surprising trait: female CFOs | Fortune
Experience & Education
-
Foundation Capital
View Steve’s full experience
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Other similar profiles
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content