Flatkey Inc. has launched Flatkey Compute, a NeoCloud that builds and operates dedicated NVIDIA B300 clusters for its customers. According to the company, every cluster is reserved for a single customer and shared with no one, and Flatkey runs each cluster as a compliant, trusted inference cloud. The company states it works only on AI, not general-purpose computing.

Why dedicated clusters

Flatkey frames the offering around a contrast with general-purpose clouds, where many tenants share the same physical hardware. The company says that arrangement works for experiments, but production inference is harder: latency has to be predictable, another tenant's workload can slow yours down, and regulated customers have to document who else has access to the machines.

A Flatkey Compute cluster is presented as removing those variables. Because the whole cluster belongs to one customer, no other tenant's workload can affect its performance, and a security review has no co-tenants to account for.

Built and run by Flatkey

Ahead of the launch, Flatkey says it has delivered compute capacity to dozens of major customers. It sizes each cluster to the customer's workload, with GPU count, region and network topology set per deployment instead of being picked from a list of standard instance types. Once the cluster is live, Flatkey operates it, so the customer does not need its own infrastructure team to run it.

A B300 inference cloud, for AI only

Clusters run on the NVIDIA B300 and are built for AI inference. Flatkey does not offer general-purpose computing and says it designs its infrastructure around AI workloads alone. On data protection, the company states it is GDPR compliant, monitors its security controls continuously through Vanta, and does not retain request content.

Pricing and terms

Customers request capacity through Flatkey's website, stating the GPU count, region and term they need. Flatkey says it replies within 24 hours, then sends a formal quote that sets the price, configuration, region and provisioning date. Reserved terms of one, three and five years are available, and longer commitments are priced lower. A reference GPU price index on the same page shows indicative rates before a customer gets in touch.

“Inference is moving from something teams experiment with to something their product depends on,” said Hunter Guo, founder of Flatkey. "At that point, you don't want to be sharing a machine with strangers. You want your own cluster, you want to know where it is, and you want somebody to run it for you. That is what we are building."

Availability

Flatkey Compute is available now by request. Smaller teams that need only a few GPUs can also rent them on demand, billed by the hour, in the Flatkey console.

Company background

Flatkey Inc. describes itself as a NeoCloud that builds and operates dedicated NVIDIA B300 infrastructure for AI inference, with each customer running on its own hardware sized to its workload. The company also gives developers access to more than 100 official AI models through one API key. Flatkey launched in July 2026 and is headquartered in San Jose, California.