Pinecone has broken its proprietary cloud lock: the vector database supplier now runs its managed service inside customer accounts on AWS, Azure and Google Cloud, and it plans a fully on-premises edition next.

The move matters because vector databases hold the numerical representations of natural language queries sent to large language models. Enterprises have been reluctant to let that data leave their boundary. Pinecone's bring-your-own-cloud (BYOC) model answers the objection directly.

How does the split architecture work?

Under BYOC, the data plane — vectors, documents, metadata and request payloads — runs inside the customer's selected cloud account and region, including regions beyond Pinecone's standard service footprint. AI agents and applications connect to that data plane directly for reads and writes.

The Pinecone control plane handles resource lifecycle, authentication and service health. It does not store or process customer content. Pinecone receives only anonymized operational metrics and traces for monitoring and support.

The design has one notable property: no standing door. Pinecone does not require SSH, VPN, inbound network access or a persistent cross-account IAM role. Upgrades, scaling and maintenance arrive through an outbound call from the control plane and execute locally.

BYOC works alongside the Enterprise plan's existing controls:

• SSO and RBAC
• SCIM + SAML
• Audit logging and encryption
• Private-networking controls

Admin teams use the same Pinecone APIs, SDKs and control plane workflows across BYOC and standard deployments, so a workload can pick its deployment model without a separate development path.

Who is running it in production?

Toyota Motor North America (TMNA) was one of the first BYOC customers to deploy in production. It keeps decades of proprietary manufacturing knowledge inside Toyota's environment rather than on Pinecone's managed service.

Kordel France, Head of AI Engineering at Toyota Motor North America, said: "Decades of manufacturing know-how live in our documentation, and that institutional knowledge is one of the most valuable assets we have. It also happens to be complex — highly structured engineering data sitting alongside unstructured process documents, across a lot of formats and a lot of different access patterns. Pinecone BYOC runs inside our own environment, so that knowledge never leaves our boundary and is served only to models we've already vetted."

That quote is the whole commercial argument in one sentence. Data governance, not performance, decides where vector workloads land.

What comes next?

Pinecone is developing a fully self-managed option for air-gapped and highly restricted networks, where both the control plane and the data plane would run inside the customer environment. That would take it from cloud-adjacent to genuinely on-premises.

Is Pinecone first here?

It is not. Zilliz Cloud documents regions on all three hyperscalers and offers both SaaS and BYOC on AWS, Google Cloud and Azure, though feature availability varies by region and plan.

Weaviate can be self-hosted on Kubernetes across AWS (EKS), Azure (AKS) and GCP (GKE). Its Weaviate Shared Cloud runs on GCP and AWS, with the AWS version generally available in US East and Europe Frankfurt, and an Azure Marketplace listing covers running it in a customer tenant. Qdrant supports AWS, Azure, GCP and customer-owned hybrid infrastructure.

Pinecone is therefore catching up with its rivals rather than leading them. Still, Toyota's production deployment gives the company a reference customer with a genuinely sensitive corpus, and the promised air-gapped edition would extend the same boundary logic to networks with no cloud egress at all.