Blaxel is joining Baseten | Infrastructure for agents

[TL;DR]

Today we’re announcing that Blaxel has been acquired by Baseten.
_The TL;DR: the Blaxel platform keeps running. Features keep shipping. Your access, your credits, and the people you talk to are unchanged. No action is required from you._
### Why Baseten
We started Blaxel to build a cloud for AI. We've been building it one primitive at a time.
Sandboxes came first because nothing else works without them. Agents write and run code, call tools, and carry state across long tasks, and you can't do that safely in a shared process. Every agent gets its own microVM. Then we made them fast: suspend and resume in 25 milliseconds, up to 5x faster than other sandbox products. A sandbox can sit idle for hours at close to zero cost and come back before the model finishes its next sentence, so state persists without paying for compute you aren't using. Sapiom ran 2.5 million of them on Blaxel in three months. As David Zhang, founding engineer there, put it: "Blaxel handled that kind of growth in a way I'd normally expect to take a year or more to reach."
Agents also produce a lot of stuff. Files, dependencies, working context, half-finished outputs. That has to outlive a single run and be reachable by the next sandbox or the next agent. Agent Drive is our distributed filesystem for that. Mount it into any sandbox and whatever an agent writes is durable, versioned, and there when something else needs it.
And agents have to talk to things. Tools, MCP servers, APIs, other agents. We built the networking layer that connects sandboxes to all of that with the isolation and access controls a production team will actually sign off on.
Our customers describe the result in terms we wouldn't have dared to use ourselves. Adi Singh, founder of Ara: "In our stack, alongside our auth, database, and frontend providers, the sandbox provider is the fourth pillar, and it's a new one. Blaxel is our first and most essential."
The primitive we didn't own was inference. We've seen open-weight and custom models become core to how autonomous agents get deployed in production. Inference is no longer a service you call from another data center. It has to sit next to compute, storage, and network, or latency compounds on every loop.
Joining Baseten is how we colocate all four. Baseten's inference stack runs at scale across dozens of regions, with more than $2B raised to build it. Together we're building the same thing from both ends: the future of agentic infrastructure.
### What this means for you
Blaxel is continuing. The product you use today is the same product with more support behind it. The team you’ve been reaching out to for support will continue to assist you just the same. Features will keep shipping. Your API keys and credits remain valid. If anything about your setup should need to move, you’ll hear it from us first with plenty of notice.
Over time, Baseten will add new products based on Blaxel’s primitives, starting with Sandboxes. Together with inference and training, these primitives will be the foundation for the agentic infrastructure we’re building.
### What’s shipping next
The roadmap is full and we only plan on increasing our shipping velocity.
On the compute side, we’re rolling out Carbon (aka Mk 3.1), our new generation of compute runtime, which is our most important upgrade we’ve done on the product since we launched sandboxes. The Carbon runtime unlocks many long-awaited features: sandbox checkpointing (so you can rollback a sandbox state at a point in time), forking (so that your agents can explore parallel universes in sandboxes – or simply to create already running sandboxes faster), and dedicated IPv6 for every sandbox.
On the storage side, Blaxel Agent Drive now has an S3 API, so you can read and write files directly from the outside world. We’re rolling it out in Europe, and improving performances every week.
On the networking side, we recently added custom functions and method matching to Blaxel Proxy. Next, we’re unlocking full extensibility with custom plugins and extensions.
Across the products, reliability is our north star. We intend to serve the next trillion agents.
### The people stay the people
Our co-founders Paul, Charles, Mathis, Chris, Nico, and the rest of the team you talk to daily remain the same people. They will be doing the same work, with more support behind them. Feel free to reach out to us if you have any questions about this acquisition.
Most importantly: thank you. You made a bet on our infrastructure early, and you’ve shaped the product with your feedback. We don’t take that lightly.
If you have a question about any of this, contact us directly on our Discord or your traditional point of contact. We’re happy to talk.

Today we’re announcing that Blaxel has been acquired by Baseten.

The TL;DR: the Blaxel platform keeps running. Features keep shipping. Your access, your credits, and the people you talk to are unchanged. No action is required from you.

Why Baseten

We started Blaxel to build a cloud for AI. We've been building it one primitive at a time.

Sandboxes came first because nothing else works without them. Agents write and run code, call tools, and carry state across long tasks, and you can't do that safely in a shared process. Every agent gets its own microVM. Then we made them fast: suspend and resume in 25 milliseconds, up to 5x faster than other sandbox products. A sandbox can sit idle for hours at close to zero cost and come back before the model finishes its next sentence, so state persists without paying for compute you aren't using. Sapiom ran 2.5 million of them on Blaxel in three months. As David Zhang, founding engineer there, put it: "Blaxel handled that kind of growth in a way I'd normally expect to take a year or more to reach."

Agents also produce a lot of stuff. Files, dependencies, working context, half-finished outputs. That has to outlive a single run and be reachable by the next sandbox or the next agent. Agent Drive is our distributed filesystem for that. Mount it into any sandbox and whatever an agent writes is durable, versioned, and there when something else needs it.

And agents have to talk to things. Tools, MCP servers, APIs, other agents. We built the networking layer that connects sandboxes to all of that with the isolation and access controls a production team will actually sign off on.

Our customers describe the result in terms we wouldn't have dared to use ourselves. Adi Singh, founder of Ara: "In our stack, alongside our auth, database, and frontend providers, the sandbox provider is the fourth pillar, and it's a new one. Blaxel is our first and most essential."

The primitive we didn't own was inference. We've seen open-weight and custom models become core to how autonomous agents get deployed in production. Inference is no longer a service you call from another data center. It has to sit next to compute, storage, and network, or latency compounds on every loop.

Joining Baseten is how we colocate all four. Baseten's inference stack runs at scale across dozens of regions, with more than $2B raised to build it. Together we're building the same thing from both ends: the future of agentic infrastructure.

What this means for you

Blaxel is continuing. The product you use today is the same product with more support behind it. The team you’ve been reaching out to for support will continue to assist you just the same. Features will keep shipping. Your API keys and credits remain valid. If anything about your setup should need to move, you’ll hear it from us first with plenty of notice.

Over time, Baseten will add new products based on Blaxel’s primitives, starting with Sandboxes. Together with inference and training, these primitives will be the foundation for the agentic infrastructure we’re building.

What’s shipping next

The roadmap is full and we only plan on increasing our shipping velocity.

On the compute side, we’re rolling out Carbon (aka Mk 3.1), our new generation of compute runtime, which is our most important upgrade we’ve done on the product since we launched sandboxes. The Carbon runtime unlocks many long-awaited features: sandbox checkpointing (so you can rollback a sandbox state at a point in time), forking (so that your agents can explore parallel universes in sandboxes – or simply to create already running sandboxes faster), and dedicated IPv6 for every sandbox.

On the storage side, Blaxel Agent Drive now has an S3 API, so you can read and write files directly from the outside world. We’re rolling it out in Europe, and improving performances every week.

On the networking side, we recently added custom functions and method matching to Blaxel Proxy. Next, we’re unlocking full extensibility with custom plugins and extensions.

Across the products, reliability is our north star. We intend to serve the next trillion agents.

The people stay the people

My co-founders Charles, Mathis, Chris, Nico, and the rest of the team you talk to daily remain the same people. They will be doing the same work, with more support behind them. Feel free to reach out to us if you have any questions about this acquisition.

Most importantly: thank you. You made a bet on our infrastructure early, and you’ve shaped the product with your feedback. We don’t take that lightly.

If you have a question about any of this, contact us directly on our Discord or your traditional point of contact. We’re happy to talk.