Artificial Intelligence is consuming trillions of dollars around the world, and Arthur Hayes, former CEO of BitMEX and co-founder of crypto investment firm Maelstrom, has a pretty unique take on where all of this is headed.
For those unfamiliar, Hayes is one of the most respected — and controversial — voices in the crypto universe. He is not the kind of person who sits on the fence, and this time he went straight to the point answering questions from CNBC at the Gamma Prime Investing Conference in Singapore: humanity is wasting trillions of dollars building AI data centers, and it is going to end the way it always ends during major tech waves throughout history. With excess capacity, a sharp downturn and, ultimately, a bailout — and whoever is positioned in cryptocurrency will reap the rewards.
Sound familiar? That is because we have seen this script before. 📈
The Same Old Cycle: Boom, Bust and Bailout
Hayes is not talking about science fiction. He is comparing the current moment with well-documented historical cycles. In his words, if you study financial history and every major technology rollout, there is always an overbuild, there is always a crash and there is always a bailout. It is a pattern that repeats with an almost uncomfortable consistency. The dot-com bubble in the late 90s is the textbook example: billions were poured into internet infrastructure that the market simply could not absorb at that pace. Companies went public, raised fortunes and vanished from the map within months. The infrastructure remained, investors cried — but those who survived the chaos and positioned themselves correctly came out ahead when the internet finally showed its true potential years later. History has a curious memory: it repeats the same mistakes wearing different outfits, and the artificial intelligence market appears to be following that same path at an impressive speed.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
The difference now is scale. When we talk about data centers to support AI workloads, we are talking about investments that easily surpass hundreds of billions of dollars, with tech companies in an all-out race to secure the processing power needed to develop and run increasingly advanced AI models. These are massive physical structures packed with wildly expensive GPUs that need electricity equivalent to small cities just to operate. And all of this is being built at a pace that may be well ahead of the actual demand the market can generate in revenue. According to Hayes, all this expansion will ultimately make powerful computing extremely cheap and extremely abundant — which is great in the long run but devastating for anyone who bet big with the wrong timing.
Hayes argues that this mismatch between supply and demand will inevitably create a massive concentration of idle capacity. When that happens in strategic sectors — and AI clearly is one — governments historically step in. The technical term is bailout, but in practice it means public money will be used to rescue what private capital built in a disorganized way. And that is exactly the point where Hayes’s reasoning starts making a lot of sense for anyone following the cryptocurrency market and understanding the role digital assets can play in a scenario of accelerated monetary expansion.
Who Is Driving the Demand — and Who Still Is Not Profitable
One interesting point Hayes raised is about who the big consumers of this processing power actually are. He names heavyweights like SpaceX, OpenAI and Anthropic among the top end users driving demand for powerful computing. The interesting detail? According to him, none of these companies are profitable yet. That raises an important red flag: as soon as the data centers currently under construction come online, infrastructure providers will want to get paid for the processing these companies committed to using.
And when does that bill come due? Hayes points to a window between late 2027 and 2028, when a significant portion of new data center capacity should be delivered and operational. That is when the pressure for real revenue will knock on the door, and companies that cannot turn technology into profit are going to feel the hit. On the other hand, there is the optimistic scenario that Hayes himself acknowledges: if AI becomes useful enough over the next twelve months to the point where demand grows sufficiently to make these companies profitable, the game changes completely. It is a delicate balance between real utility and inflated expectations.
Why Crypto Enters This Equation
Hayes’s logic is straightforward: when governments need to rescue entire sectors of the economy, they print money or expand credit in ways that historically weaken the value of fiat currencies. This already happened in 2008 with the financial system and in other episodes over the past two decades. Anyone who understands this mechanism knows that assets with a limited supply — like Bitcoin — tend to appreciate precisely in those moments, because they function as a kind of protection against the dilution of purchasing power in traditional currencies. This is not a bet made in a vacuum; it is a thesis built on repeated historical patterns.
In Hayes’s own words, fortunately we have Bitcoin and other cryptocurrencies to absorb that excess liquidity, and that is why we already know which asset will perform best when the bailout arrives. His recommendation for anyone wanting to ride this wave is short and sweet: just be patient. ⏳ The investment in AI infrastructure also has a geopolitical dimension that cannot be ignored. The United States and China are in an open race to lead the development of artificial intelligence systems and secure control over the powerful computing needed to run the most advanced models. That means part of the money being poured into this infrastructure does not come solely from market decisions — it comes from national strategies where the government is the driving force. And when the government is the driving force, the bill at the end is always paid one way or another by society as a whole.
It is worth noting that Hayes is not a fan of betting on falling prices or shorting AI companies. He himself classifies that strategy as not being a good investment opportunity, even though he acknowledges that major technology rollouts are historically oversized. In other words, he believes in the downturn but prefers to position himself on the constructive side of the equation, aiming at the bailout and the appreciation of digital assets after the dust settles.
What This Means in Practice
Translating all of this to the real world: we are at a moment when the volume of investment in artificial intelligence and the infrastructure supporting it is creating enormous pressure on the global market for chips, energy and specialized construction. And it is not all losses for now — Hayes acknowledges that some companies feeding this boom are already making real money, specifically citing memory chip manufacturers and Nvidia. The big question for investors, according to him, is whether they are paying the right multiple for these companies’ future earnings. That is where both the danger and the opportunity live.
But Hayes is not preaching pessimism — quite the opposite. His view is that this cycle will happen, it will hurt whoever is in the wrong places, and it will create historic opportunities for those who are well positioned. The powerful computing being built today will be used, just like the fiber optic cables from the dot-com bubble were used — only years later, by companies that did not even exist when those assets were built. The problem is not the technology itself; it is the pace and volume of capital being allocated before real demand is mature enough to absorb all of it in a sustainable and profitable way.
Flop: Hayes’s Bet on the Future of AI Agents
And it does not stop there. All that abundance of processing power generated by the massive data center buildout is behind Hayes’s newest bet in the crypto world: a project called Flop, focused on payments between AI agents, with a launch planned for the first quarter of 2027. The logic is elegant in its simplicity. If powerful computing is going to become cheaper and more abundant, AI agents will multiply rapidly — and they are going to need a way to pay for what they consume.
Flop aims to create a spot market for processing power, where participants are rewarded with Flop tokens for providing GPUs and running AI inferences. Today, according to Hayes, there is no payment network designed specifically for AI agents, and that is the gap the project wants to fill. In his words, if agents can convert a currency directly into processing — which is what they eat and consume — then they will use that currency. That is the bet. 🤖
For anyone following the cryptocurrency market and thinking about how to position themselves in this scenario, Hayes’s core message is clear: understand the cycles, understand where money goes when systems collapse, and be there before everyone else reaches the same conclusion. 🚀 Artificial intelligence is going to change the world — that is not up for debate. The question is who is going to foot the bill on the road to get there, and who is going to be in the right place when the dust settles.
