The money is concentrated at the top

Major tech companies like Nvidia, Google, OpenAI, Microsoft, Meta, Amazon, and Oracle have all seen their stock prices explode over the past few years as they’ve consolidated their AI efforts, pulled in gargantuan amounts of funding based on AI hype, and announced incredible infrastructure projects that have upset almost every industry, from smartphone production to water management.

They, in turn, are funding a few select data companies. (How do you think companies like ScaleAI grew so big, so fast?) But as for everyone else, they’re not doing so good. Indeed, if it weren’t for the big tech companies, the US economy would almost certainly be in a recession right now, according to Deutsche Bank research.

That might be OK if these companies were developing something genuinely innovative, profitable, and/or showing real potential to deliver on their investments. But the biggest reason for their sky-high valuations is simple: they’re all just investing in each other.

Circular investments don’t go anywhere

When Oracle announced its $300 billion Stargate Project with OpenAI, Nvidia was set to be the major hardware supplier. Nvidia also invested $100 billion of its own in OpenAI, and Nvidia is also a major investor in CoreWeave (another Oracle supplier). CoreWeave works closely with Microsoft, which is itself a major OpenAI investor. Microsoft and Nvidia have also invested in OpenAI rival Anthropic, which has done major infrastructure deals with Amazon and Google.

The list goes on. And while these enormous investments might have sent company valuations skyward, many of them are based on multi-year plans with estimates for future hardware delivery and assumptions around costs and scalability. For it all to succeed, it’s got to be done before whatever AI bubble there is bursts.

In totality, these investments are completely unprecedented, too. By 2030, McKinsey estimates that AI investment could reach almost $7 trillion. For comparison, the entire Manhattan Project cost only $30 billion (adjusted for inflation). Yikes.

Companies aren’t making more money

Nvidia might be making a killing on AI as the seller of GPU-shaped shovels in this proverbial gold rush, but everyone else is struggling. Microsoft has revised AI sales targets due to poor uptake of its paid-for services, and OpenAI has churned through over $150 billion in investment dollars just to make $15 billion or so in revenue in 2025.

If OpenAI can’t figure out how to turn a profit with the backing of all its AI friends, with close to a billion active users, and with the most dominant mindshare in the AI chatbot space, who else will?

For institutionalized tech companies like Google, Meta, Amazon, and Microsoft—the ones with diversified businesses, long-tail revenue options, and expansive cash stockpiles—this may not be a problem in the short or medium terms. But even major companies can’t burn through money forever. Microsoft and many others have enacted huge layoffs over the past year to help maintain healthier balance sheets, and investors are going to come calling for their returns sooner or later.

Smaller AI companies will be affected first, but as we saw with Meta’s disastrously unprofitable endeavors to develop the Metaverse, even massive tech companies can run dry on interest and momentum. What happens then to the trillions in AI investment?

Local AI is getting much better

From Nvidia’s DGI Spark system to home hackers running large language models on their gaming GPUs, it’s easier than ever to run local AI on your own machine. They aren’t the top models with their trillions of parameters, mind you, but the latest large language models designed for home hardware are becoming increasingly capable.

OpenAI, Anthropic, Microsoft, and others would love a future where you run all your AI services through their cloud platforms that are gated with subscription fees, but the latest local LLMs are capable enough to handle basic text generation, editing, summarizing, and image generation.

With the added benefits of improved privacy, security, and response time for local LLMs, more and more individuals and companies are going to pivot this way in the months and years to come. That’s not going to do any favors for AI companies seeking profitability.

It’s already outlasted most tech bubbles

This one might be more of a meta point on economic bubbles than a specific point for the AI industry, but massive market rallies only tend to last a few years. Yahoo Finance highlights how the dot-com bubble lasted just over two years, the Japanese stock bubble of the 1980s lasted three years, and the big tech-disrupted rally after COVID just under a year.

In each of these cases, their respective stock markets saw enormous growth of several hundred percent in just a few years. The AI boom hasn’t quite managed that—it has experienced a gain of around 130 percent over the past three years—but three years is already longer than most of these historic booms. If AI is a bubble that’s going to bust, then it’s almost past due going by historic trends.

AI has a power problem

All the major AI companies have announced their deals and are starting to make them reality. We aren’t going to see new rounds of trillion-dollar deals from even the major tech firms, as even they have limits on the capital they have available to them.

But scaling up to these grandiose goals of AI data centers all over the world is proving difficult. After buying up all the GPUs and memory they can, some of these companies are still struggling to bring them online.

Microsoft CEO Satya Nadella said in November that the company now had a power problem, meaning it had GPUs that couldn’t physically plug in because it lacked the power to run them. And then you have Elon Musk’s xAI trying to import a power station (no, really) and supersonic jet company Boom Supersonic converting its jet engines into gas generators (no, really!). AI needs power, but power is at a premium.

Power stations and their associated grid infrastructures take years or even decades to build. When power supply fails to catch up to their ambitions, it’s going to slam the brakes on expansion. A slowdown like that is the last thing the AI industry needs to keep its hype train rolling.

Consumer AI fatigue is very real

No one likes Grok’s deepfakes and child exploitation images. Fake frames in Nvidia-run games are making gamers feel like they’re not really playing their games. AI-driven memory shortages and associated price spikes for everyday tech products are driving people crazy.

Companies are noticing and already pivoting, too. The most stark example so far this year is Dell relaunching its XPS brand at CES 2026. Sure, it’s still a “Copilot+ PC,” but you’d never know it from the marketing. AI is gone from the forefront and back is the focus on longevity, everyday performance, and lightweight design—you know, the things that consumers actually care about.

If no one’s even interested in buying AI, how are these companies ever going to make it profitable? That’s not something investors are going to want to see or hear this year.

Global trade issues could derail everything

On top of all the internal AI industry factors that could bring about its fall, there’s also global instability that could just as likely to deal a death blow. The US administration’s volatile leadership keeps throwing up (and tearing down) trade barriers. Hardware nationalization and nationalism are driving secular investments rather than global branch-outs.

And as the superpowers keep eyeing up what their neighbors have, there’s the risk of war stalling out the world economy. Just China closing off Taiwan’s access to global markets would be enough to collapse just about everything in the tech sphere.

Hopefully nothing like that comes to pass, but the prospect of it is one more Sword of Damocles hanging over the AI industry.

Even if it bursts, Nvidia will keep pushing

Google, Amazon, and Microsoft won’t collapse. OpenAI probably won’t collapse. But the AI hype train of today could be heavily derailed. The smaller and medium AI companies, and all the firms promising agentic revolutions for your business? They’ll be gone. Stock prices will collapse, and global recession could be the medium-term fallout that eventually corrects everything back to some semblance of normalcy.

Long term, though, even the companies that remain will have to contend with hardware deprecation that’ll see them scurrying back to Nvidia every 2 to 3 years. Even with all that hardware power, none of these companies are going to reach AGI (or even super-intelligence) with large language models that can’t functionally understand anything.

AI is here to stay, but the industry as it is can’t last much longer. The signs are there that 2026 could be the year it all changes—again.