For funders

Contribute to a safer AI future.

Pacific Compute operates as a safety-only cluster for all open-weight labs on both sides of the Pacific. We use this compute to run evaluations on new model releases, and to work with the labs to troubleshoot safety issues as they arise. The first node in Singapore, NVIDIA H200 GPUs, is rented and live.

Per launch

$200k estimate

Targeting about a $200k spend per major model launch, with smaller iterative model launches having a more modest budget. A clean model may finish well below $200k and a difficult model may run well above it. Unused compute rolls into the next launch, so this number is an approximation rather than a hard cap. What the $200k buys

Overheads

Compute first

We rent rather than buy, so nothing is stranded if the project ends. The team stays small and salaries modest until the work calls for more. No office, events or communications spend.

Sponsor

US 501(c)(3) fiscal sponsorship

An established US 501(c)(3) has offered fiscal sponsorship and the agreement is being drafted.

Why your money cannot become capability
Gradient updates default-blocked

Gradient updates to the model are blocked by default. Every exception is approved by our technical team, only for a problem we are solving with the lab, and logged on the ledger where the other participants can see it.

The record

Every job on a public ledger. Anyone can audit what the compute did.

Nothing stranded

No owned hardware, so nothing is stranded if the project stops. Unused compute rolls into the next launch.

Write to us: info@pacificcompute.org