Tech investors build discounted GPU supercomputers to support AI startups struggling with chip shortage
-
Nvidia CEO Jensen Huang and other tech investors are setting up massive GPU clusters to provide discounted compute access to AI startups who otherwise can't afford tens of thousands of chips.
-
GitHub founders Nat Friedman and Daniel Gross spent 9 figures upfront to build a 4,000+ GPU supercomputer called Andromeda that they rent out to their portfolio companies at near cost.
-
The GPU shortage continues to bottleneck progress in AI research. Other firms like Index Ventures, Y Combinator, and M12 have also set up free limited access compute clusters.
-
Running an industrial-scale GPU cluster requires infrastructure investments and dedicated engineering teams to manage, which most VC firms don't have the appetite for.
-
Friedman and Gross are exploring building an even bigger Andromeda cluster as demand continues to outstrip supply, even for clusters with thousands of chips.