Earlier this year, Bushuyeu discovered a startup called B3 Labs that offers dedicated GPU hardware on a rent-to-own basis under the brand B3IQ. Under the company’s program, customers can pay 30% down on a variety of machines, including a frontier level Nvidia H200 rig that starts at $54,516, and spread out remaining payments over five years.

The option is an attractive one, says Bushuyeu, because the grant system that supports many university researchers typically requires that projects have predictable costs—a goal that is hard to meet when relying on rented compute, where prices can sway significantly.

A second advantage comes in the form of privacy. Owning a dedicated machine ensures researchers can comply with health-related laws like HIPAA, and avoid triggering the regulatory or national security concerns that go with relying on overseas compute providers. For practical purposes, B3IQ customers can arrange for the machine to be assembled on their own premises or, more practically, ask the company to host it at its 27,000-square-foot facility in Oregon.

While New York-based B3IQ only started offering its rent-to-own compute services this year, the startup is notching a growing number of customers. Those include researchers at New York University who are using B3IQ-supplied compute to build simulations based on war-zone evacuation data, and to model nation-level diplomatic negotiations. The startup says that academics at Stanford and Dartmouth are also among its customers, and that it is in talks with university procurement offices to provide its services on a more wholesale basis.

An unexpected pivot

As is the case with many startups, B3IQ’s current business model is not the one it started out with. Instead, co-founders Daryl Xu, Viktoriya Hying and Sean Geng—who met while working at crypto giant Coinbase—started the company in 2024 with a focus on video games.

According to Xu, who is B3IQ’s CEO, the trio are “nerds at heart” and saw gaming as a natural segue from crypto. But in the course of trying to get their startup off the ground in early 2025, they discovered that they and companies of all sorts were jostling to get their hands on GPU chips, which power gaming consoles as well as AI machines.

“AI compute was exponentially increasing. I think we had a lot of gaming customers just kind of yell at us because GPU prices were so expensive, and we thought there was something there. And so we did what we’ve always done: build vertical, meaning we built both software and hardware in the AI space,” Xu recalls.

This led the founders to develop expertise in building and operating GPU powered machines, which they honed in running the Oregon data center. This means B3IQ is able not only to assemble GPU rigs on demand, but also to help customers upgrade to newer chips or add additional compute to the machines they buy.

B3IQ is hardly the only firm to offer alternatives to buying GPU machines outright. Large providers like AWS, of course, provide compute on an hourly metered basis, while the likes of Vast.AI offer various long-term financing options. But for now, B3IQ appears to be carving out a niche by offering an arrangement that is conducive to the privacy and budgeting concerns of academics. As an added sweetener, the startup says machine owners can deploy a dashboard to have B3IQ sell any spare compute capacity on their behalf to approved third parties.

The startup, which currently has around 30 employees, is on track to bring in $50 million in revenue this year, according to Xu. The company has raised $21 million from Pantera and Coinbase Ventures, among others, and is making plans to complete a larger Series A round in the coming year.