Solana Confidential Compute Launches in Beta for Private AI Apps

The Solana Foundation launched Solana Confidential Compute on November 13, 2026, an open-source toolkit for building private AI applications with hardware attestations checked on the Solana (SOL) blockchain. The toolkit is in beta on Solana devnet, built with Nito, the team behind TryNitoAI.

How Solana Confidential Compute Checks the Hardware

AI workloads run inside trusted execution environments, parts of a chip that keep data and models away from the operator. Attestation is how that hardware proves what it is running. The toolkit moves that check on chain. A hardware-signed report binds a signing key to a recent Solana slot. An on-chain program then verifies the report and records the signing key, hardware type and expiration, the foundation wrote in the Confidential Compute announcement.

What the Beta Supports Today

The first beta covers two attestation formats, Intel TDX and AMD SEV-SNP. NVIDIA GPU attestation is still in development, according to the Solana Foundation's launch thread. The release ships as an open-source SDK plus the on-chain verification program, so developers can run private AI workloads and let the chain verify the hardware behind them instead of trusting a provider's promise. For AI teams, the setup covers the usual chain of custody: a model runs inside the enclave, the enclave signs a report, and the Solana program checks it.

What Comes Next

The devnet beta is unaudited and meant for development use, and no mainnet date has been announced. Developers can start building now through the foundation's Solana AI portal. Everything runs on devnet for now, so the attestation flow can be tested end to end before any mainnet deployment. Solana's market history over the last month is on the SOL 30-day chart. Confidential compute gives the Solana ecosystem a way to run sensitive AI workloads while keeping the hardware behind them verifiable on chain.