Building the AI factories of the future

Collaborative Fund leads Phaidra’s $50M Series B alongside NVIDIA.

We’re thrilled to share that Collaborative Fund has led Phaidra’s $50M Series B alongside NVIDIA, Index Ventures, Helena, and others.

We’ve always believed the most transformative companies are the ones that marry profit with progress. AI is no exception. For all the excitement around what AI can do, the reality is that the infrastructure powering it is massively bottlenecked by energy, cost, and complexity. And that’s exactly what Phaidra is solving.

Phaidra builds autonomous agents that make AI data centers — or what we increasingly call AI factories — run smarter, cheaper, and cleaner. Their work is critical not just for advancing AI, but for making sure this wave of technology is sustainable in the long run.

I’ve been fortunate to sit on Phaidra’s board for the past two years, and to work closely with Jim Gao, Katie Hoffman, and Vedavyas Panneershelvam as they’ve built this vision. Few people know AI infrastructure as deeply as Jim, who started as one of Google’s earliest data center engineers and went on to lead industrial AI efforts at DeepMind.

I sat down with Jim to ask him about Phaidra’s origin story, the “aha” moments along the way, and how he sees the future of AI factories:

Some highlights from our conversation (edited and condensed for clarity):

On AI’s counterintuitive insights:

“One of the first things the AI did when we turned it on was bring a lot more chillers online. The load per chiller went down, so per-unit efficiency looked terrible. Everyone thought the AI had made a mistake. Then we looked at the system-level numbers: the PUE [Power usage effectiveness] dropped by 30%. By spreading load, it massively reduced energy use in the pumps and cooling towers. That was not obvious to us because these are very complex systems. But this is precisely where AI shines.

On where AI’s true value lies:

“People today think that AI is about automation primarily. I believe the real value of AI is creativity — a tool for discovering new knowledge in extraordinarily complex systems. We’re seeing this in things like protein folding with AlphaFold, novel materials, new chemicals, and grid balancing.”

On the rise of AI factories:

“Traditional data centers look nothing like modern AI factories. Today’s facilities are essentially one massive computer, with tens to now hundreds of thousands of synchronized GPUs acting as a single unit of compute, with exponentially higher power density and orders of magnitude higher complexity.”

Listen to the whole thing, or scrub to 18:36 for the part where I talk about holding Jim’s baby girl at a board meeting.

⁂

With this new capital, Phaidra will expand its agent ecosystem across more of the AI factory stack, deepen its collaboration with NVIDIA, and bring these capabilities to hyperscale and enterprise data centers worldwide .

And for Collaborative, this raise extends our thesis of profit and progress into AI and the infrastructure that will shape technology’s future.

We’re proud to have been part of Jim, Katie, and Veda’s journey so far — and even more excited for what comes next as they tackle AI’s toughest bottlenecks head-on.

If you’d like to dive deeper:

• Read the full press release and coverage in Axios

• Share the news on Twitter/X and LinkedIn

• Watch Jim’s appearance on TBPN

Thanks for following along and helping us spread the word.

—Sophie

---