General Catalyst builds a data center to feed its startups’ AI compute needs

General Catalyst is building a data center to give its portfolio startups access to AI servers. The Information reported the move in an exclusive on October 14, 2026.

Venture firms have long handed founders checks, advice, and introductions. Now at least one of them is handing out GPUs, which in the current market may be worth more than the advice.

What General Catalyst is building

The facility is designed to supply AI compute, mainly servers and GPUs, to companies backed by the firm. These are startups facing severe shortages of the hardware needed to train and run AI models.

A separate report on October 13 indicated the center will serve a broad mix of startups. That includes companies at early stages as well as those further along in their growth.

Part of the setup is reportedly a compute cluster set aside for startups with high token consumption. Tokens are the small chunks of text an AI model reads and writes, so heavy token use means a company is running its models hard and often.

The firm’s stated approach centers on creating shared compute pools. Instead of each founder negotiating separately for scarce capacity, portfolio companies would draw from a common reservoir the firm controls.

General Catalyst has not publicly shared where the facility will be located, how much capacity it will hold, or what it will cost to build.

The people and money behind the push

In May 2026, General Catalyst hired Alexis Black Bjorlin, a former Nvidia executive, as its Chief Strategy Officer.

The firm’s motivation traces back to its own founders. Feedback from portfolio companies pointed to significant compute challenges, and the shared pools are meant to address those pain points directly.

General Catalyst’s $1.5 billion “Creation Strategy” focuses heavily on AI-native companies, meaning businesses built around AI from day one rather than ones adding it later.

General Catalyst is not alone in thinking along these lines. Other venture firms have taken a similar approach to securing AI resources for their portfolios, according to the research on the move.

Why this matters for startups and VCs

Access to computing resources is currently tight enough to limit how quickly AI startups can build and scale. A venture backer that can guarantee some baseline capacity removes one of the biggest variables founders face.

For the venture industry, the move marks a notable shift in what a VC firm actually is. The traditional model is capital plus network, with operations left to the companies themselves.

There are open questions. Data centers are capital intensive and operationally complex, and the firm has not detailed how it plans to price or allocate access across its portfolio.

Allocation could get tricky. If the shared pool is finite and several high-usage startups need it at once, someone has to decide who gets priority, which is a new kind of conversation for a VC to have with founders.

What to watch next: whether General Catalyst discloses the facility’s location and capacity, how many portfolio companies end up drawing on the shared pools, and whether more venture firms follow it into the compute business.