Cryptocurrencies Will 'Devour' AI, Says Cardano Founder

• Hoskinson identifies payments, alignment, and provenance as challenges for AI.
• Compares the rise of data centers to the overbuilding of fiber optics.

Charles Hoskinson, founder of Cardano, stated in an interview on September 16, 2026, that cryptocurrencies will eventually 'devour' artificial intelligence (AI) over the next 5 to 10 years, becoming an infrastructure for payments, coordination, data provenance, and distributed computing.

Hoskinson's thesis stems from a critique of the current AI development model. The entrepreneur considers it unsustainable for spending on training and infrastructure to maintain exponential growth, especially since the available electrical capacity does not allow for a tenfold increase in spending on data centers each year. It is worth noting that in 2025, data centers consumed 485 TWh of electricity, and the International Energy Agency projects that their demand will reach about 950 TWh by 2030, as reported by CriptoNoticias.

Hoskinson also questioned whether companies like OpenAI and Anthropic can indefinitely sustain their losses while seeking to monetize models whose main cost, as he explained, lies in pre-training.

For Hoskinson, cryptocurrency networks could also solve problems related to ownership and provenance of the data used by AI systems. A registry network would allow documentation of when content was created, who owns it, and how it was transferred, while digital assets would enable the automation of royalty payments when a model uses third-party intellectual property.

The founder of Cardano also proposed a transformation of computing infrastructure. Instead of concentrating the training of increasingly larger models in data centers, he suggested connecting GPUs, phones, and other devices to create distributed computing networks. In his view, this model would allow for the gathering of processing capacity without building centralized facilities for each new training cycle.

Cryptocurrencies will eat AI because we solve all the hard problems that AI cannot solve. We solve the payment problem. We solve the alignment problem. What is a blockchain? It is a system of alignment: you take the plurality of understandings from people around the world and turn them into a set of rules that everyone has to follow.

Charles Hoskinson, founder of Cardano.

Additionally, Hoskinson anticipates that the current rise of data centers could end in excess capacity. He compared the scenario to the expansion of fiber optics in the United States during the 1990s, when, as he recalled, nearly 90% of the installed infrastructure initially went without customers. In his opinion, a similar phenomenon could occur with AI data centers if demand turns out to be lower than expected.

The entrepreneur then expects a shift towards local models and more powerful devices. He cited as a reference equipment capable of running models with between 400 billion and 500 billion parameters with 512 GB of memory, which currently costs around USD 20,000 and, according to his projection, could drop to USD 5,000 or USD 3,000 in the coming years.

The combination of distributed devices and shared processing capacity would require, according to Hoskinson, an infrastructure capable of coordinating resources and incentives among participants. Therefore, he believes that cryptocurrencies could fill that space by providing payment and coordination mechanisms already utilized by decentralized networks. On the left side, Charles Hoskinson during the interview for the media outlet Deeptech Insights. Source: Youtube

Finally, Hoskinson anticipated an adjustment period of between 24 and 36 months before a reconstruction of AI infrastructure and estimated that by 2030, the convergence between both industries could help bring the market capitalization of cryptocurrencies to USD 10 trillion and attract billions of users.

Rather than suggesting the disappearance of AI, his forecast points to a change in the layers that support its functioning. If the cost of concentrating computing continues to rise, the provenance of data becomes more relevant, and models require automated payments on a global scale, competition could shift from who builds the largest model to who efficiently coordinates the resources that allow its use.