Build your product.
Keep your options.
For AI startups, inference providers and research teams. Choose graded capacity across independent clouds without a new long-term commitment at every stage of growth.
Your next stage of growth shouldn't come with a cloud lock-in. Graded GPU hours. A choice of providers. One invoice.
Meet VivaldiStart with the GPU class your workload needs. Buy hours without buying into a single provider's future. Pricing follows a published market benchmark.
Redeem your hours across eligible providers with capacity in your class. The next place you run is still your decision.
Direct customers pay monthly on net-30 terms, subject to approval. Keep your capital for the company you're building.
A GPU's name doesn't tell you how it has been treated. A current Voltry condition record is the foundation of Vivaldi's network standard.
Thermal history gives context that a model number alone cannot provide.
For AI startups, inference providers and research teams. Choose graded capacity across independent clouds without a new long-term commitment at every stage of growth.
Qualify your fleet and reach buyers beyond your own sales pipeline. Vivaldi brings independent GPU clouds into a shared channel for compute demand.
For VC platforms, accelerators and rewards providers. Give your customers GPU hours they can redeem across an eligible network, with the freedom to choose where they run.
Your eligible hours can be redeemed with different network providers that have capacity in the same GPU class. Workloads still need compatible software, storage and networking. Portability is a buying option, not automatic live migration.
The Voltry record describes measured hardware condition and its trend. It supports network eligibility; it does not guarantee uptime, application performance or remaining hardware life.
Hours are class-specific. An H100 hour is not a B200 hour. Cross-class conversion is not part of the launch offering.
Vivaldi is forming its founding network, initially focused on US compute buyers, qualifying cloud operators and distribution partners. Provider availability, commercial terms and launch timing will be confirmed as the network comes together.
We're bringing together the first buyers, cloud operators and partners for a more flexible way to buy compute.
Explore the founding network