The future of AI and Web3 is being built on GPUs—literally

Sam kamani

September 13, 2025

As AI continues to scale and Web3 matures, a quiet but crucial infrastructure war is unfolding—one centered around GPUs.

In Episode 190 of Web3 with Sam Kamani, recorded live at Token2049, we sit down with Nick Filichkin, founder of ComputeLabs, to unpack what it means to treat GPUs not just as hardware, but as a financial asset class. If you're wondering what the future of compute looks like, or how AI and Web3 might share (or fight for) resources, this conversation offers a front-row seat.

What is ComputeLabs, and why do GPUs matter?

ComputeLabs isn’t just another infrastructure company. They're on a mission to decentralize compute power by making GPUs—traditionally centralized in big data centers—accessible through leasing, tokenization, and open networks. Think of it as "DeFi for compute."

Nick explains:

“GPUs are the fuel for AI. But right now, access is limited to a few big players. We want to open that up—just like how Web3 opened up finance.”

Do AI and Web3 need each other?

One of the most engaging parts of the episode is the discussion around whether AI and Web3 are allies or opposites. On one hand, AI thrives in centralized environments where massive datasets and compute power are easily coordinated. Web3, on the other hand, is built on principles of decentralization and user sovereignty.

Nick puts it eloquently:

“Web3 decentralizes. AI centralizes. But both need compute. That’s where the bridge is.”

And ComputeLabs is that bridge—creating infrastructure where decentralized networks can access the same power that fuels today’s AI models.

Tokenizing compute: The rise of GPU-backed assets

One of ComputeLabs' most radical ideas is turning GPUs into tokenized assets. Instead of locking value in hardware, they're creating liquidity and access:

Companies can lease idle GPU resources.
Users can invest in compute without buying physical hardware.
The system is backed by utility tokens that reflect usage and demand.

This isn’t just theory—it's already happening. ComputeLabs is laying the groundwork for what could become a decentralized "NASDAQ for compute."

Breaking NVIDIA's mountain

The GPU market is largely dominated by NVIDIA. While their chips are best-in-class, their market dominance makes GPU access expensive and unpredictable—especially for startups.

Nick shares his thoughts:

“We’re not trying to replace NVIDIA. We’re trying to make access to their hardware—and others’—more equitable and decentralized.”

Philosophy meets infrastructure

Sam and Nick dive into a deeper question: what kind of world are we building?

AI’s centralizing tendencies raise concerns about data ownership and power concentration. Web3 offers an alternative—systems governed by communities, not corporations.

This episode pushes listeners to consider:

What happens when AI models run on decentralized GPUs?
Can the next ChatGPT be trained on community-run hardware?
Will we tokenize compute the way we’ve tokenized art and finance?

Final thoughts: Action over hype

Nick closes with a call to action for founders, builders, and investors:

If you're an AI company looking for compute, try decentralized leasing.
If you're building Web3 infra, start thinking about tokenized assets beyond crypto.
And if you're curious about where the future is headed—get involved early.