Charles Hoskinson Says Crypto Could “Eat” AI Over the Next 5 to 10 Years

Cardano founder Charles Hoskinson believes cryptocurrency could absorb a significant part of the artificial intelligence industry over the next five to 10 years. Speaking on the Deeptech Insights podcast, Hoskinson argued that blockchains can address several problems that centralized AI companies struggle to solve, including payments, alignment and data provenance.

“Cryptocurrencies are going to eat AI,” Hoskinson said, describing a future in which crypto becomes infrastructure around AI rather than simply another technology sector. His argument is based on the idea that AI systems will increasingly need open payment networks, shared rules and verifiable records for the data and computing resources they use.

Hoskinson Challenges the AI Data Center Model

One of Hoskinson’s main arguments concerns the enormous infrastructure spending required to train and operate increasingly capable AI models. He questioned whether companies can continue increasing data-center investment at the current pace when electricity supply and the economics of model training impose physical and financial limits.

Hoskinson specifically pointed to companies such as OpenAI and Anthropic, arguing that operating individual AI models can generate revenue while large-scale pre-training remains particularly expensive. His comments come as the AI industry continues investing heavily in computing infrastructure, even as questions about long-term capital requirements and returns become more prominent.

Related: Cardano’s Charles Hoskinson Warns Academics Could Lose Control of Ideas Shared With AI

The Cardano founder compared the current AI infrastructure cycle with the fiber-optic expansion of the 1990s. He said roughly 90% of newly installed fiber once remained unused while demand caught up, and suggested that AI could experience a similar period of excess capacity.

That scenario could produce what Hoskinson calls “dark data centers,” where large amounts of installed computing capacity remain underutilized. He expects the industry to eventually shift toward smaller models running closer to users, including local devices and distributed computing networks.

Hoskinson pointed to increasingly powerful consumer hardware as evidence that local AI could become more practical. He cited Apple’s M5 Mac Studio and Nvidia’s compact Spark systems as examples of computing platforms that could bring substantial model capabilities outside traditional hyperscale data centers.

Crypto as the Coordination Layer

The bigger part of Hoskinson’s argument is what happens after computing becomes more distributed. He believes cryptocurrency could provide a coordination layer capable of connecting GPUs, phones and other devices into a broader marketplace for computing resources.

Payments are one part of that model. Blockchain networks can transfer value between participants without requiring every device owner, developer or AI agent to establish a traditional financial relationship with a centralized intermediary.

Hoskinson also highlighted alignment as another potential role for blockchain. His argument is that decentralized networks can establish shared rules between participants, while individual AI companies currently make their own decisions about how their systems operate and what behavior they permit.

Data provenance is the third piece. Blockchain-based records could potentially help establish where data originated, how it was modified and which parties interacted with it. That could become increasingly relevant as AI systems consume enormous amounts of content and questions around ownership, attribution and automated payments grow.

Related: Charles Hoskinson Says Pogun Could Bring Bitcoin Liquidity to Cardano DeFi

The thesis does not mean blockchain technology has already solved these problems at the scale required by the AI industry. Distributed computing, blockchain throughput, privacy, verification costs and coordination mechanisms all remain technical challenges. Hoskinson is presenting a long-term direction rather than announcing an existing crypto-powered AI grid.

His broader forecasts are similarly ambitious. He said he does not expect a U.S. CLARITY Act to pass before 2029 and projected that public blockchains could reach roughly $10 trillion in assets and one billion new users by the end of 2030. Those figures are Hoskinson’s forecasts, not established market expectations.

For Cardano and the wider crypto sector, the significance of his argument is less about replacing AI companies outright and more about where blockchain infrastructure could sit around AI. If computing becomes more distributed, AI agents conduct more automated transactions and data provenance becomes more important, payment and coordination networks could become part of the underlying infrastructure.

Hoskinson’s “crypto will eat AI” thesis therefore describes a possible convergence between two industries that are currently developing largely in parallel. Whether that convergence occurs on the five-to-10-year timeline he outlined will depend on economics, energy availability, hardware, regulation and whether decentralized networks can deliver practical advantages over centralized infrastructure.

About The Author