Artificial intelligence news.
Straight from the original announcements.

Pinecone BYOC adds on-demand mode so teams can skip index sizing

October 15, 2026By the AI News Bites Editorial TeamSource: Pinecone
Pinecone BYOC adds on-demand mode so teams can skip index sizing
Image: Pinecone

Pinecone has added an on-demand capacity mode to its Bring Your Own Cloud (BYOC) offering, letting customers create indexes without sizing each one. According to the company, on-demand indexes share a single capacity pool that can hold millions of namespaces.

Key points
  • Pinecone BYOC now supports on-demand capacity mode, available to every BYOC customer.
  • On-demand indexes share one capacity pool that can hold millions of namespaces.
  • Dynamic sharding lets customers add shards without rebuilding indexes or moving existing data.
  • A heavily queried namespace can be moved to its own index on dedicated read nodes in the same cluster.
  • Pinecone says the zero-access model remains: data stays in the customer's cloud account.

What on-demand mode changes

Pinecone BYOC runs Pinecone inside a customer's own AWS, Google Cloud, or Azure account. The data stays under the customer's control while Pinecone operates the database. The company says teams in legal, healthcare, and financial services often choose it to keep their data in their own cloud.

With the new mode, announced in a blog post by Manish Talreja on Oct 15, 2026, all on-demand indexes draw from one shared pool. Each namespace can represent a tenant of an app or a user of an AI agent. Customers start with a small pool and add capacity as their data grows.

The sharding problem it targets

Pinecone describes a common issue in search systems, whether semantic or lexical: users must pick the number of shards for every index at creation. Choosing too few means rebuilding the index once it outgrows them. Choosing too many means paying for idle hardware. Adding shards later forces existing data to move, and some shards still end up much fuller than others.

On-demand mode uses dynamic sharding. Instead of tying an index to a fixed set of shards, the system tracks where each part of every index lives. According to Pinecone, customers can add shards without rebuilding anything or moving existing data. New data goes to the shards with the most room, which keeps data evenly spread across the pool.

If one namespace receives far more queries than the others, it can be moved to its own index on dedicated read nodes in the same BYOC cluster. The remaining namespaces continue to share the on-demand pool.

Security and support model

Pinecone says scaling the pool preserves the zero-access BYOC model. Vectors, metadata, and queries stay in the customer's cloud account. Pinecone adds capacity without SSH, VPN, inbound network access, or a standing cross-account role.

The company monitors cluster health through operational metrics such as CPU and latency, without seeing customer data. When help is needed, customers share logs with Pinecone support so both sides can troubleshoot together.

Availability and getting started

On-demand mode is available now for every BYOC customer. New customers use Pinecone's setup wizard, which generates a deployment project to run in their own cloud account to create the cluster. Existing BYOC customers already have a cluster ready.

To create an on-demand index, users pick their BYOC cluster as the location in the Pinecone console, API, or SDKs and choose on-demand as the capacity mode. For cluster planning, Pinecone points customers to their account team. The company also refers readers to its BYOC guide, which covers the architecture and how data is secured, and to the BYOC FAQ.

For multi-tenant applications and AI agents with many per-user namespaces, the change means capacity planning no longer has to happen index by index, according to the company's description of the feature.

Related news

AWS details four ways to speed up Amazon Bedrock AgentCore sessions
For developersNovember 13, 2026, 23:02

AWS details four ways to speed up Amazon Bedrock AgentCore sessions

AWS has outlined four architecture patterns for reducing session startup delays in Amazon Bedrock AgentCore runtime, including snapshot-based initialization, build hooks…

AWS adds three-tier Data Capture to SageMaker HyperPod
Products and toolsNovember 13, 2026, 19:01

AWS adds three-tier Data Capture to SageMaker HyperPod

Amazon SageMaker HyperPod now includes Data Capture, a built-in capability that records live inference inputs and outputs to Amazon S3. AWS says the feature is designed…

Pinecone Adds Google’s Gemini Embedding 2 to Integrated Embedding
Products and toolsNovember 12, 2026

Pinecone Adds Google’s Gemini Embedding 2 to Integrated Embedding

Pinecone has added Google’s Gemini Embedding 2 to its integrated embedding feature for the Documents API. Teams can now send plain text to Pinecone for semantic search…