Is the AI Bubble About to Pop?

Happy Saturday, one and all,

While Arthur Hayes recently warned that the “drunken reverie” of the AI bubble would inevitably succumb to the laws of financial gravity, the Maelstrom founder came on the BeInCrypto Expert Council podcast this week, and explained why he has adjusted his view.

Is the multi-billion-dollar AI boom a sustainable economic revolution or a massive bubble on the verge of a historic crash? Wednesday’s blockbuster Nvidia earnings call forced a major reality check, shifting market consensus on the sheer longevity of this buildout. Below, I’ll lay out Arthur Hayes’ post-earnings outlook on where the sector goes from here.

On Wednesday, Nvidia reported a record-breaking $81.62 billion in Q1 2026 revenue (beating expectations of $78.86 billion), driven by a staggering 92% year-over-year growth in its datacenter vertical.

With US tech giants projected to spend $750 billion this year on AI infrastructure, Nvidia’s CEO Jensen Huang was unapologetically triumphant:

“The buildout of AI factories, the largest infrastructure expansion in human history, is accelerating at extraordinary speed. Agentic AI has arrived, doing productive work, generating real value, and scaling rapidly across companies and industries.”

But as tech giants pour near-infinite capital into data centers and hardware, a critical question looms for investors: When do these technologies actually start paying for themselves?

Many analysts look at outlandish valuations for pending tech IPOs like OpenAI and SpaceX, or note that 90% of current large language model (LLM) users are non-paying subscribers, concluding that a crash is imminent.

Arthur Hayes strongly disagrees.

While historical tech bubbles were built on pure speculation and cash-flow-negative companies, the current front-runners are pivoting to profitability faster than the market realises. Hayes points to the massive surge in inference requests, the computing power used when an AI model actually processes and answers queries for active workflows.

Take Anthropic, for example. After burning billions of dollars in development, the company has officially crossed into profitability on a go-forward basis. This comes as businesses are rapidly integrating agentic AI solutions into their daily operations. As Hayes notes:

“The revenues are there... It’s not as if the high flyers are cash-flow negative companies... and as the price of intelligence declines, the usage increases. The more that data centers are built, the more chips that are sold, the cheaper it becomes to run these models.”

The Deflationary Flywheel of Intelligence: Because the cost of running these models is falling, human civilization is using vastly more of it to replace traditional knowledge work, sparking a feedback loop of insatiable demand for compute.

Arthur then explained why tech giants are trapped in a hyper-expensive, endless sprint to stay ahead of the competition. It comes down to evolutionary biology applied to tech: The Red Queen Effect, where you must run as fast as you can just to stay in the exact same place.

On the front-end software layer (like OpenAI, Gemini, or Anthropic), if a model falls even a few percentage points behind in reasoning capability, corporate clients will immediately jump ship to a competitor. To survive, these firms have no choice but to continuously pour money into hardware to train the next-generation model. As Hayes explains:

“If your front-term model drops behind a few percentage points in effectiveness, then the corporate clients switch... and then all of a sudden all that money you spend is worthless. And so this leads to the ‘I need to train the next best front-term model. I need to spend another 100 billion, 150 billion, whatever it costs to train one of these models.’ That continues to drive the capex cycle because you cannot fall behind.”

For investors, trying to pick the ultimate software winner in this environment is incredibly risky because today’s leading LLM can become obsolete tomorrow. Instead, the smartest plays remain anchored to the immutable infrastructure layer, the hardware, the chip supply chains, and the specialized data centers that power the ecosystem.

As Hayes points out, while it isn’t clear which front-end model will ultimately win, “if you move more to the source of intelligence generation right on the hardware side, yes, there are clear ways: Nvidia, Intel, TSMC... it’s pretty clear that these guys are doing things and you can’t replace them.”

The mainstream narrative that workers displaced by automation will seamlessly transition into “magical new AI jobs” is historical fiction. When manufacturing left the West, those workers didn’t find identical employment overnight. AI is poised to trigger a similar, severe dislocation of knowledge work, but the social fallout will look completely different depending on where the worker is located.​

In Western advanced economies, knowledge workers possess significant political and social leverage. If efficiency gains threaten their livelihoods, they won’t go quietly. Arthur points out that these workers will use their systemic influence to force a protective response from the state:

“Knowledge workers are the most politically active people in the United States, whether it’s a Republican or Democrat... they’re going to find a way to get protected. They’re the ones out in the streets, they’re the ones who the mainstream media caters to... And if they’re saying that they’re no longer gainfully employed and it’s these tech overlords that have taken their jobs through efficiency gains, I think there’s going to be a political backlash in the United States. And maybe they’re going to be some sort of AI-to-labor, social contract reform, whatever that means.”

However, the outlook is vastly more grim for emerging economies like India, Bangladesh, and the Philippines, which have spent decades building massive industries acting as low-cost back-office processing and call centers for Western firms.

Because an AI agent can now handle data entry, customer service, and manual coding for pennies on the dollar, Western corporations are erasing these cost centers entirely. And unlike Western knowledge workers, these overseas labor forces have zero political leverage over the multinational corporations cutting them loose. As Arthur bluntly puts it:

“These workers are essentially cost centers for American and Western European companies. Nobody cares if they survive or not. ‘Oh, okay, I’m going to lay off my entire Filipino staff that was doing all this manual work... Well, now an AI could do all that. Goodbye.’ And nobody cares if they lose their jobs... And so the Philippines, a lot of these countries don’t have the wherewithal to make a situation better. And so I think there’s going to be massive social strife in a lot of these third world countries that are essentially back office processing centers for a lot of Western companies.”

The ultimate check on the AI CapEx orgy isn’t just capital availability or software obsolescence; it is the physical limitation of the electrical grid and the environment.

The scale of infrastructure needed for next-generation compute is reaching unfathomable proportions, perfectly illustrated by the controversy surrounding the proposed Stratos mega-scale datacenter campus in Box Elder County, Utah. At full buildout, this single 40,000-acre project would require up to 9 GW of power, a demand equivalent to that of the entire city of New York, or more than the current electrical consumption of the entire state of Utah.

The Stratos campus and its accompanying fossil-fuel power infrastructure could dump 16 gigawatts of thermal load into the local high-desert valley.

This translates to 23 atomic bombs’ worth of thermal energy being injected into the local ecosystem every single day!

The latest BeInCrypto Expert Council podcast with Arthur Hayes will be hitting the BeInCrypto YouTube channel next week; please make sure to subscribe here so you can be notified exactly when it drops!

What does it actually take to bring crypto to the masses without scaring them off? This week, I sat down with fintech heavyweight Michael Rihani, Senior Director of Digital Assets at Nubank, to dissect how Latin America’s banking giant has already onboarded over 7 million crypto users.

Rihani, whose pedigree includes architecting Apple Pay, integrating Bitcoin payments at Tesla with Elon Musk, and leading product teams at Cash App and Coinbase, breaks down Nubank’s masterclass in transparent, zero-fee onboarding.

From launching a 6% native yield on Solana and navigating their historic conditional OCC approval for a US national bank charter, to why they strictly ban speculative meme coins, Rihani explains the exact playbook behind “blockchain banking,” the most critical phrase in global fintech right now.

​Watch the full interview here:

Have a perfect weekend,

​

Brian McGleenon,

Global Head of News, BeInCrypto