We're excited to announce that Blaxel has raised $7.3M in seed funding from First Round, Y Combinator, Liquid 2, Multimodal and several others. This investment will accelerate our mission to provide the infrastructure layer that AI agents need to operate reliably at scale. Current cloud offerings were designed for the SaaS era of human users. Now, there's an urgent need for a new cloud platform that understands the unique requirements of autonomous agents.
The Problem: AI Agents Need Different Infrastructure
The rise of AI agents has fundamentally changed how software is built and deployed. Agents are now managing calendars, responding to emails, generating code, and even creating entire applications autonomously. Yet they're running on infrastructure designed for human developers and traditional applications.
Current cloud providers offer the same services they've had for years, adding features without removing complexity. The result? Infrastructure that's increasingly misaligned with how AI agents actually work.
Consider what happens when an AI coding agent needs to test generated code:
- Agents are making several LLM and tools calls, requiring a very variable amount of resources to fulfill one request
- A request to an agent can last for dozens of seconds or even minutes
- Traditional containers aren't secure enough—agents can be prompted to escape sandboxes
- Virtual Machine boot times of 30+ seconds are unacceptable when agents make hundreds of iterations
- Authentication, auditing, and observability systems weren't designed for autonomous actors
- Network architectures can't handle the distributed, high-latency connections agents require
- Existing APIs weren't designed for LLM-friendly interfaces
Unlike traditional applications where databases sit next to web servers, AI agents face unique networking challenges. An agent might connect to an LLM in one region, APIs in another cloud, memory systems in a third location, and knowledge bases elsewhere. Each connection adds latency, yet users expect instant responses—especially with voice interfaces becoming the primary way humans interact with agents.
This mismatch creates real problems. Companies building AI agents spend more time wrestling with infrastructure than improving their products. Costs spiral as agents require dedicated resources just to achieve acceptable performance and users leave their product because of performance issue (not only the accuracy).
Enter Blaxel: Infrastructure Built for Agents, Not Humans
Blaxel provides a complete infrastructure platform designed specifically for AI agents. Instead of adapting legacy services, we built from first principles around agent requirements.
Secure Sandboxes Our sandbox environments spin up in under 3 seconds, with cold starts from hibernation measured in milliseconds. They provide:
- True isolation that prevents container escapes
- Automatic hibernation after 1s inactivity
- Snapshot and forking* capabilities for rapid experimentation
- File systems optimized for high-frequency modifications
- LLM-friendly OS interfaces that agents can directly consume (using MCP)
Intelligent Gateway We provide unified access to all major LLMs through a single API, with:
- Built-in caching and rate limiting
- Observability designed for agent workflows
- Support for private model deployments
- Automatic failover between providers*
Serverless Agent Hosting Deploy agents as a serverless API from a simple GitHub connection. We handle:
- Automatic scaling based on agent activity patterns
- Integrated observability for agent-specific metrics
- MCP (Model Context Protocol) server hosting
- Letting the agent schedule and execute thousands of background subtasks in parallel
- Pay-per-use pricing that scales with agent efficiency
Early Success: from YC to Production
During our YC batch, we worked closely with many of the fastest growing AI agent startups – we helped companies like Human Behavior, Vybe, and Jazzberry deploy and scale their agents. These early-stage teams needed infrastructure that could scale with them—from prototype to production. For startups moving fast, having infrastructure that doesn't slow them down is critical.
By the end of the batch, we had deployed infrastructure across 16 global regions and processed millions of agent requests daily. One of our customers is running over 1 billion seconds of agent runtime processing for millions of videos without managing their infrastructure—all for 50% less cost than typical serverless infrastructure. Every day, we provide terabytes of compute memory, enabling AI agent companies to automate their infrastructure. This saves them from hiring expensive DevOps teams and allows them to focus on acquiring customers.
The Road Ahead
The shift to agent-based computing is accelerating. Gartner predicts that by 2028, 75% of application development will involve AI agents. These billions of agents will need infrastructure that understands their unique requirements.
Current cloud offerings were designed for the SaaS / Web 2.0 era, where the main users of software were humans. We are building a cloud offering designed for agents. The difference matters when every millisecond of latency and every security vulnerability directly impacts agent performance and reliability.
Who we are
Paul Sinai was the Co-Founder and CEO of ForePaaS, which was later acquired by OVHcloud, Europe’s largest cloud provider. When Paul noticed the large gap in the industry that inspired Blaxel, he teamed up with his former colleagues from OVH and ForePaaS: Christophe Ploujoux, Nicolas Lecomte, Thomas Crochet, Charles Drappier, and Mathis Joffre. It might seem unusual to have six cofounders, but to create what we think of as "AWS for AI Agent Companies", we knew we needed a robust and world class team, and we all joined forces to build Blaxel.
Join the Agent Revolution
We're grateful to our investors—First Round Capital, Y Combinator, Liquid2, Transpose, Multimodal and an exceptional group of angels—who share our vision of purpose-built agent infrastructure.
But most of all, we're grateful to our customers who trust us with their agents every day. You've shown us what's possible when infrastructure gets out of the way.
Building AI agents? Stop fighting infrastructure. Start shipping products. Try Blaxel at blaxel.ai or reach out to us.
The future of software is autonomous. The infrastructure powering it should be too.
- Coming soon.
