Flatkey will sell you a whole B300 cluster and run it, too
Founder Hunter Guo's San Jose neocloud reserves each B300 cluster for one customer, operates it for them, and sells it on one, three and five year terms.
By Ryan Merket · Published
Primary source: PR Newswire Business Technology
Why it matters
Production inference is where AI costs stop being experimental, and Flatkey is betting that teams hitting latency, compliance and noisy-neighbor limits will trade shared cloud GPUs for reserved B300 clusters they do not have to operate. The move also shows how fast a model-access startup can move into capital-heavy infrastructure, where GPU supply and contract length help determine success.
Flatkey founder Hunter Guo launched Flatkey Compute on November 4th, a service that builds and operates dedicated NVIDIA B300 clusters for one customer at a time. Flatkey said every cluster is reserved for a single customer, operated by Flatkey, and used for AI rather than general-purpose computing.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.