Author: Sea

In the past few years, if one were to say what the most important resource in the AI era is, it would not be models and talent, but rather the lower-level GPU.

The consecutive years of the U.S. government's restrictions on the export of high-end GPUs from Nvidia to China are the best proof of this.

Most businesses, research institutions, and professional AI users face two core issues:

1) How to acquire GPUs;

2) After acquiring them, where to place them, how to maintain them, and what to do with idle computing power?

The startup company @B3Labs from the Coinbase ecosystem recently launched B3IQ, which aims to solve these two major issues.

In one sentence, B3IQ allows enterprises, institutions/universities, and professional users to own their own servers equipped with Nvidia GPUs through an installment payment method.

How to buy?

The American manufacturer Andromeda, invested by B3 Labs, assembles servers equipped with Nvidia GPUs, deployed in a 27,000-square-foot data center in Oregon, USA, with the platform responsible for the operations and maintenance.

There are two options to use these GPU servers: rent-to-own (30% down payment) or a one-time purchase.

GPUs are expensive AI production materials, and the "rent-to-own" model is relatively friendly to small and medium-sized enterprises and professional users: a 30% down payment can lock in the machine and put it into use, and the remaining amount does not require cash payment from you, as the machine will earn it back on its own. During the payback period, computing power income is split 30% to you in cash and 70% to the platform for repaying the remaining principal and 8% interest (management fees are calculated separately).

After paying it off, if you continue to rent out computing power, the profit-sharing structure reverses: 70% goes to you, and 30% goes to the platform. Users can choose to continue renting out computing power to earn income or take the physical server away.

B3IQ currently offers GPU chips/servers such as H200 NVL, Workstation RTX PRO 6000, RTX PRO 6000 Server Edition, RTX 5090, B300, etc., to meet various levels of demands.

How to maintain and utilize idle computing power?

Whether through rent-to-own or one-time purchase, when the GPU is idle, you can rent it out to other users/companies in need through the platform to earn income. The B3IQ platform is responsible for matching demands, and the earnings can offset the purchase cost or directly become income.

This is one area where I find B3IQ's solution quite interesting.

In other words, B3IQ provides a third flexible model that differs from "renting cloud computing power" and "building computing power independently." For many universities, laboratories, and AI startups, this is more predictable than long-term renting of cloud computing power and much lighter than building a data center independently in one go.

I checked the earnings calculator on B3IQ's official website and estimated that with a sample of one 4× H200 NVL, users can fully own this GPU server after 28 months, with a 4-year IRR (pre-tax) of 55%.

B3IQ's clients

It is not just companies and institutions in China that have this demand; some institutions in the United States, while able to purchase Nvidia GPUs, also require B3 Labs' products.

Because not all AI work is suitable for third-party cloud services/AI large models, such as medical research, internal corporate business data, and data involving privacy, security, or intellectual property from government/defense/military; at the same time, there are more flexible usage demands for GPUs.

Researchers from New York University, Stanford University, Dartmouth College, and the University of Hawaii at Manoa are using B3IQ to support their AI work, including cancer research and training specialized AI models.

Some of this work even involves national-level diplomatic negotiation simulations, and such tasks cannot rely on commercial AI services to complete.

About B3 Labs

B3 Labs was established in 2024, with a founding team from Coinbase, receiving over $21 million in investments from institutions like Coinbase and Pantera Capital.

Compared to its financing background, I think what is more noteworthy is its entry point, which is a pressing need in the AI industry.

B3 Labs aims to transform GPUs from a resource that can only be leased into an infrastructure that can be owned, operated, and generate income.

As open-source models become more mature, deployment thresholds lower, and the talent gap between China and the U.S. continues to narrow, the ever-daunting hurdle of GPUs remains a battleground that cannot be bypassed.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。