Deal of the Week
Got to love a good IPO — and especially a Berkeley-founded company IPO. Michael James.
Cerberas had a blockbuster debut, raising $5.5B. After already increasing the IPO price to $185, the stock promptly soared to $350, giving the company a market cap of roughly $64B.
Cerebras is building AI chips optimized for inference, alongside an inference cloud business with major Middle Eastern customers and a large OpenAI deal. I have spoken a lot about the coming explosion in inference demand, and Cerebras is one of the biggest beneficiaries of that shift. You can test the speed here.
So many fun elements to this story:
The founders had conviction around AI long before it entered the mainstream zeitgeist, starting back in 2016, and kept building through years when there was far less excitement around the space.
They even attempted to IPO in 2024, but ultimately hit the market in 2026 just as expectations around inference demand exploded.
Foundation Capital effectively incubated the company and ended up with a stake reportedly worth around $5B. (Due to lock-ups etc., we won’t know realized gains for a while, but still an incredible outcome.) Benchmark and Eclipse also did extremely well.
This is one of those stories that reminds you just how fun startups and venture capital can be! Early to a thesis, strong conviction, long timelines and BOOM massive IPO. Lets have a few more!
The Wild West Era of AI Adoption
It feels like we are currently in the Wild West Era of AI.
The gunslingers chose their own weapons. In many cases teams are picking their own tools outside of company structures between Cursor, Codex and Claude. Each has its own quirks and may end up being hard to harmonise. (See OpenAI & Anthropic both standing up consulting companies to try to fix this by creating ecosystems!)
There's no sheriff in town. Governance is improvised. Prompts live in Slack threads. Plugins run scripts nobody audits. Most orgs can't tell you who's using what, where the data goes, or who's accountable when it goes sideways.
Nobody's cares on the cost (yet!). When the frontier is open, you grab land and care less on balancing the books. Teams are burning tokens on low-value work because being left behind feels like a failure. FOMO > ROI.
Finally in the Wild West there are the prospectors willing to push the limit and in many cases find untold riches and in other cases peddling snake oil. It is currently messy and vibrant. But so alive!
But every Wild West must end. The railroads came and the government laid down the law. Whoever figures out how to capitalise on this next era make the next category defining company.
Intelligence per Token
A debate gaining steam is the idea of Intelligence per Token. The idea that not all tokens are created equal. For a given unit of inference cost, how much useful intelligence are you actually getting back? People are seeing this vary massively and trying to measure the variance.
A few factors driving the variance:
Time of day - Peak-hour traffic means degraded routing and slower, lower-quality responses.
GPU cluster routing. - Whether your query hits the latest hardware (see Cerberas above!) or an older cluster. This even goes down to the data center geography. This impacts latency and output.
Context window utilization. Performance degrades as you fill the window, often well below the stated limit.
Silent model degradation. Providers need to optimize for cost, and the model you tested last quarter may not be the model you're calling today behind the API.
Query type. Some prompts are just harder to serve consistently than others.
This creates pain in the enterprise adoption as model behavior has drifted making it hard to create consistency.
The implication is that single-model lock-in starts trick. The winners will be the teams that route dynamically, matching query to model to context. This builds on the idea of the atomisation of data pipelines where some models will be used for certain tasks.
Expect routing layers, model arbitrage, and intelligence benchmarking to become real categories. The question stops being "which model is best?" and becomes "which model is best right now, for this query, at this cost?"
Quick Takes
Love this image - this would probably be 2026 - Year of the Agents. Most valuable company started ???
Summary by the #️⃣ & 💰:
4 Berkeley-founded companies funded
$121M of capital raised from the 11th April to 17th May
💡 Got any ideas or feedback on how to improve this weekly digest? Just hit reply.
IPOs
🧠 Cerebras Systems. $5.6B IPO 🇺🇸 AI infrastructure company. 💰 Benchmark Capital, Fidelity, Foundation Capital
🐻 Michael James, Founder & Chief Architect. BA Molecular Biology Article
Closed Rounds
⚡ GridCARE. $64.0M Series A 🇺🇸 AI grid optimization platform. 💰 Clocktower Ventures, Emerson Collective, Sutter Hill Ventures
🐻 Ram Rajagopal, Co-Founder & CTO. PhD & MA EECS Article
🤖 Config. $35.0M Seed 🇺🇸 Robotics data infrastructure platform. 💰 Kakao Ventures, Mirae Asset Global Investments
🐻 Minjoon Seo, Co-Founder & CEO. BA EECS Article
🏥 Knit Health. $11.6M Seed 🇺🇸 Healthcare intelligence platform. 💰 Frist Cressey Ventures, Moxxie Ventures, Uncork Capital
🐻 Jonathan Kolstad, Co-Founder & CTO. PhD & MA EECS Article
🏡 HINT. $10.0M Seed 🇺🇸 AI home management platform. 💰 Energy Impact Partners, Slow Ventures, Tusk Venture Partners
🐻 Yih-Han Ma, Co-Founder & CTO. BA Business Admin Article
Date Built By Berkeley Started | Companies Funded | Total Raised ($M) |
7/8/24 | 751 | 212,193 |
Our goal is to document the startup ecosystem of Berkeley-founded companies. Please share this newsletter with any Cal Bears in your network so we can crowdsource information about all investment rounds and job opportunities.
Did we miss a company or want to announce a round? Add it here, and we will post it next week.
Do you have a job you want to post from a Berkeley Company? Add here.
Is there someone in the Berkeley ecosystem that you would want us to do a profile on? A founder, funder, or general startup person? Add here.
Built By Berkeley is not affiliated with UC Berkeley, but maybe we will be one day if we get enough subscribers….
Built By Berkeley, where we announce all the funding rounds by Berkeley-founded companies. This is a community effort, so please let us know if we missed a company here. 🐻
