Foundation capital It bet on large markets that do not yet exist. In March, the 30-year-old Venture company closed a latest fund of $ 600 million to support technical founders in the earliest stages-often before what the product is-a strategy based on what it calls “markets worth a billion dollars”.
Crunchbase News recently talked to the general partner Steve Vassallowho says that the conviction of his company results from the history of early perception of types – from the incubation of the AI chipmaker Brain In 2016, to support over 100 AI startups in time.
The Palo Alto Foundation in California is looking for technical founders squeezing into unknown spaces, in which infrastructure and the needs of individuals intersect, often long before the type of consensus. Idea: Invest early, double in breakthrough corporations and help them scale by dangerous revenues from the first thousands and thousands of dollars.
Vassallo is an engineer in training with a strong design background. He said he believes that corporations that are distinguished in the latest AI AI will probably be necessary.
“You really have to see the intersection of technology and humanity to move at the moment,” he said. “It’s recognition not only for physics, but psychology. And where will people still matter?”
This is very true when learning to strengthen AI with human feedback, which keeps people in the loop. “Ai gets smarter because people see where the gaps are,” said Vassallo. “And then we become smarter in terms of interaction with these tools.”
Vassallo joined the Foundation 18 years ago. Earlier, he used his education in the field of electromechanical engineering and robotics for product design, including early profession time in the Nineties. IdeaBuilding hardware and software systems that reported his interest in design and quick prototyping.
Nowadays, his team in the Foundation spends a lot of time pondering about “tools that are involved in us, and are careful about all the ways of hacking,” he said.
Investing early with high conviction
Wassallo said that the pace at which Ai is now developing differs from earlier technology cycles. “We are now at a moment that I feel like nothing else that I have experienced north of 30 years.”
The investment approach of the foundations is not caused by consensus. Rather, he is looking for “markets worth a billion dollars”-ribs, which do not yet exist until the founder enters the violation and creates a product.
“A thing that enlivens so many investments that we do here in a foundation – you anticipate where the markets will go if this thing was to exist,” said Vassallo.
The company’s partners are based on “nerdy page” and track Girub Insoles, in addition to individuals who build attractive open source projects in large corporations that might be thrown out.
Vassallo said that the company tends to invest very early, with great conviction, and very technical founders. He said that this was aimed at helping them from their earliest days to the first million dollars of revenues.
In the entire portfolio of 80% of his investment is in kindergarten enterprises-often preliminary and sometimes preliminary.
New computer infrastructure
The Foundation was incubed by the startup chip AI Cerebras and collaborated with series A in 2016 Benchmark AND Eclipseeds. For context, Nvidia At the time of investment in the Foundation’s cerebraces, it was price lower than $ 25 billion. Today, a semiconductor giant is price $ 4 trillion, and Cerebras himself signaled the plans of publication.
Vassallo met the co -founder of the brain Andrew Feldman In 2009, when the latter worked on the energy performance of information centers. Vassallo didn’t invested then, but they each remained in contact.
“Investing in semiconductors in mid-2010 was a recipe for losing a lot of money,” said Vassallo.
But when the previous company Feldman, Seamicrowas purchased by AMD In 2012 and after Feldman’s departure, they each wondered for several years about what might be next.
“You can see that you need to build a different approach to optimization of intelligence per unit of energy,” said Vassallo. How to build something that “not only faster and more efficient, but also will be much more energy -saving?”
He said that the nationwide founded in 2016 decided to build a latest computing infrastructure, questioning some assumptions of six many years of semiconductor physics. Cerebras has built a principally different architecture with semiconductors on a fee scale, optimized for data movement, not calculating.
“AI loads in 2012–2016 have already grown north of 300,000 times,” said Vassallo, which gave him a conviction for investing.
“We were able to gather an unusual band around a completely different architecture. What we build on is not something they focus on,” he said.
100+ AI investments
The Foundation has remodeled 100 investments in artificial intelligence in infrastructure, data infrastructure, programmers tools, security and applications. However, the company didn’t invested the Labs model in Frontier.
Vassallo invested for the first time in artificial intelligence around 2008 and 2009 Towards total knowledge. This technology was also used to filter spam and cyber safety and in e-commerce.
Joanne ChenThe general partner in the foundation led the seed round in 2021. JasperAn application that helps in writing on the top Openai“S GPT-3. Jasper had to turn to a marketing solution when GPT-3.5 was launched in November 2022 with its own consumer chatbot, which he achieved 100 million lively per 30 days Users in just two months.
The company also ran a round of seeds in 2023 IgnitionAutomation company authorization of labor flow in the field of labor with often complicated and paper processes.
In the coding sector, Ashu GargThe general partner, he ran the company’s funds and in July 2025. Playero. The company’s CODESIM product provides for software and failures and debugs the code written by AI before its implementation, a potentially huge market when you are taking into account the latest estimates About 25% of huge technology corporations“The code is written by artificial intelligence.
“The code is such a good use of large language models, because in fact there is a way to close the loop and determine if you have something that works better or not,” said Vassallo.
The company normally invests from $ 15 to 20 million per company, which is a sum designed to perform a startup in the rounds of series A and B. It has a separate leadership fund for corporations that explode.
In addition to investing up and down, AI Stack, Foundation also invests in crypto and “Fintech technology”. The company runs 4 general partners: Vassallo, Chen, Garg and Rodolfo Gonzalez.
Vassallo focuses on infrastructure and “AI intersecting with something else”, in addition to on cryptographic practice. He expects the infrastructure to precede applications in artificial intelligence, like other investment cycles.
In general, it predicts an explosive increase in application in artificial intelligence, but notes that this may require more refined investment evaluation.
“We love working with the founders who live on this edge,” he said.