Blaxel raises $7.3M seed round to build ‘AWS for AI agents’ after processing billions of agent requests

The company announced on Tuesday that Blaxel, a basic cloud structure based on an emerging construction that was specially designed for artificial intelligence agents, raised $ 7.3 million of seed financing led by Capital First Round. This financing comes just three months after the six -year -old team graduated from the Y Combinator Batch in the Spring of 2025, which confirms the investor’s appetite for plays in infrastructure in the market of the expanding artificial intelligence agent.

The San Francisco-based company is betting that the current generation of cloud service providers-Amazon, Google Cloud and Microsoft Azure- is mainly not identical to the new wave of independent artificial intelligence systems that can take action without human intervention. Artificial intelligence agents, who deal with everything from the management of the calendars to generate software instructions, require a significantly different infrastructure from the traditional web applications created for human users.

“The current Web 2.0 cloud service providers have been designed as software as a service era,” Paul Sinai, co -founder and CEO of Blaxel, said in an exclusive interview with Venturebeat. “But with this new AI AI, we believe a new type of infrastructure for artificial intelligence factors is needed.”

Why is AWS and Google Cloud not designed for independent artificial intelligence agents

The timing reflects a wider shift in computing institutions, as companies are increasingly publishing artificial intelligence agents for customer service, data processing and automating workflow. Unlike the traditional applications in which databases sit along with web servers in the predictable patterns, artificial intelligence agents create unique networks challenges by communicating with language models in a region, applications facades in another cloud, and the rules of knowledge elsewhere – while users expect immediate responses.

Blaxel has already shown a large traction, as it has processed millions of agent’s requests daily across 16 global regions by the end of their Yombinator boost. A customer works more than one billion seconds from the agent to operate to treat millions of videos, and represents a measure that shows the requirements of the infrastructure of the first companies.

“One of our customers is to address the re -equipment of the session to enable product managers to better understand how the user’s behavior for their product,” Cenae explained. “They need to treat millions of re -session operations every month. So it represents millions of minutes of sessions. They use our agent’s infrastructure to address these re -session operations and provide visions to product managers.”

The company’s approach focuses on providing the infrastructure that artificial intelligence agents can work themselves, instead of asking human officials. This includes virtual devices with sand that extends in less than 25 milliliters, automatic expansion based on the patterns of the worker, and the programming programming facades designed for consumption directly by artificial intelligence systems instead of human developers.

How six founders have a successful exit plan to take over the large technology

The unusual Blaxel Blaxel Street, consisting of six sections of the joint team experience and selling a previous company to Ovhcloud, the largest cloud provider in Europe. This company has become a collection of entire analysis products from OVH, giving the team direct experience with both the challenges of the successful and outdoor infrastructure.

“I know this looks unusual, a very big team. We were not completely suitable on the stage for the pilot day,” said Sinai, who indicates the Y Combinator Signature event. “But we did so already. My previous company, which I sold to OVH Cloud, was also six participating founders.”

Lieutenant -General Charles Dippier, who has known Senis for more than 30 years, along with the co -founders Christophe Blogo, Nicolas Licome, Thomas Crochet, and Mattis Jovri. Their collective experience extends to the infrastructure, developers tools and platform engineering – critical experience of competition against technology giants with almost unlimited resources.

“I think it is important to be six years old now, because we have a lot of ambition,” Senis said. “What we do is building this generation of cloud computing for this new agent era.”

What distinguishes Blaxel in the competitive cloud infrastructure market

The cloud infrastructure market is able to compete, as AWS receives about a third of the market share and the latest players such as the burden of Modal, Repalate and Runpod that target artificial intelligence. Blaxel itself is characterized by focusing specifically on artificial intelligence agents instead of inference the form or training.

“Most of the competitors you mentioned are a very difficult problem, which is about inference – how you can host your model, and how you can make these models as quickly as possible in terms of the number of symbols,” said Sinai. “But there are not many people who work on agents, and this is exactly what we do.”

The company platform includes three main components: hosting agents to publish artificial intelligence systems such as Application facades of application programming without a server, and MCP context protocol (the form of the context of the context of the model) for agents of the agents with external tools, and a unified gate to reach multiple AI models. The infrastructure is designed to deal with the requirements of the changing resources of artificial intelligence factors, which may require minimal computing strength while waiting for responses but it needs large resources during active treatment.

Security and compliance features of the targeted institutions organized industries

Although younger companies targeted artificial intelligence, Blaxel has implemented security measures at the level of institutions, including compliance with SOC2 and HIPAA. The platform provides controls in accommodation in data that allow customers to restrict work burdens on specific geographical areas – and from the Duchess of Corporates in the organized industries.

“We offer a political framework where you can attach, for example, on the burden of work, this agent cannot run outside these sub -groups of regions,” Cenae explained. “You can attach a policy to say that this agent cannot run outside the United States, so you are sure that this agent will only process data in the areas you have chosen.”

This approach reflects the company’s belief that even artificial intelligence companies in the early stage need strong infrastructure practices because they build tomorrow’s institutions. “We believe it is extremely important to have, even for young companies, the best infrastructure with best practices, because they will become institutions,” Senice said.

The AS-Y -G-G pay

Blaxel adopted a model of AS-YOG-G paying as well as the well-known cloud providers, and avoiding the initial subscription approach after checking the market demand during their YMBINATOR boost. The model receives customers only when their agents actively treat tasks and close infrastructure during inactivity periods to improve costs.

“We offer the infrastructure that takes place in just a few milliliters and closed it in just one second,” said Sinai. “So, you are only paying the time when your agent treats something. When your agent wait another, you do not have to pay for it because we are closing it.”

This approach has already led to saving costs for customers, as one customer has achieved a 50 % cost reduction compared to typical solutions without a server while processing terabytes from data monthly.

Gartner expects 75 % of artificial intelligence applications to use by 2028

The investment comes at a time when industry analysts predict explosive growth in adopting artificial intelligence agent. Gartner expects that 75 % of the development of applications will include artificial intelligence agents by 2028, although Sinaï believes that the adoption of current institutions is still largely experimental.

“At the present time, most of the companies that are operating in production are mostly smaller companies, and not yet institutional companies,” he said. “So we are really focused on their service just as the large cloud service provider has done in the past.”

The strategy reflects how Amazon Web Services initially focused on startups and developers friendly companies before expanding institutional customers. Blaxel plans to follow a similar path, using $ 7.3 million to expand their software platform before moving to improvement of devices and dedicated data center.

“Seven million are not enough to build data centers, it is clear, but I think it is important to go step by step,” Sinai said. “Ensure that we now have the best facades we can provide to our customers, the best services for their agents, then go to improve the deepest infrastructure.”

The company’s road map includes features such as Snapshot Forming to try the agent, automatic failure capabilities, and the deeper improvement of the huge scale they expect. With the expectations of hundreds of billions of artificial intelligence agents in the coming decades, Blaxel sees an opportunity to build the infrastructure designed for this new computing model from A to Z.

“We believe that there is a huge economy that begins throughout the agents,” said Sinai. “There will be hundreds of billions of artificial intelligence agents, and the infrastructure we have today is not designed for this new wave.”

The financing round included a participation from Yombinator investors, SILED2, Transpose and Angel who participate in the company’s vision of the agent’s infrastructure. With artificial intelligence agents moving from experimental tools to production systems that deal with important business operations, Blaxel’s approach to obtaining a large market share may be placed while it may become the next main category of cloud computing.