Ai in 2026

My thoughts for the year ahead

The sky exploding while people watch seems apropos for 2026. Photo by Arthur Chauvineau on Unsplash

This past year kinda sucked. It wasn’t entirely horrible, but there is always so much noise that it’s difficult to find something good at any given moment. Anyway, whether you think 2025 was fantastic or terrible, it doesn’t matter. It’s over. It’s time to look ahead. And today, we are looking at a specific thing—artificial intelligence.

* * *

Ai is often in the news, but the focus always seems to be on its function (or, perhaps more appropriately, its disfunction).

There are those who claim generative Ai models like ChatGPT 5 or Claude 4 have improved their “reasoning, task execution, and multimodal understanding,” but the areas where there is measurable improvement in their application are severely limited. This is reflected in declining sales of almost every major Ai product.

Much of the hype about Ai comes from those who stand to profit, like Sam Altman or Dario Amodei. They overinflate the technology’s abilities because they have to for their financial survival. The way Ai has been funded looks far more like a Ponzi scheme than a flourishing new—and sustainable—business sector. Some are characterizing it as a bubble.

Whichever it is—a bubble or something akin to a scheme to defraud—the outcome is looking bad for us. So, here is what we might see in the Ai sector in 2026 (and beyond).

Bubble Talk

Generally, an economic bubble describes a situation when a particular sector’s value becomes overinflated by speculation driven by factors that have little substance in reality. The ‘Ai bubble’ refers to the vast amounts of money being invested into Ai that seem impossible to recoup as profit.

Some time after a bubble forms, it inevitably pops when investors realize there is nowhere near the value in the sector they expected, which leads to an abrupt market correction/contraction. That’s what happened at the turn of the 21st century, what we now call the dot-com bubble.

Shortly after the dot-com burst came the dot-com recession, which saw a loss of almost 80% of the value of the NASDAQ. The Ai bubble makes the dot-com situation look miniscule by comparison.

Spending on Ai hit an astounding $375 billion in 2025, and is expected to reach $500 billion in 2026. From 2013 to the end of 2024, total investment in Ai exceeded $1.6 trillion. Jared Bernstein, senior fellow at the Center on Budget and Policy Priorities, stated:

You’ve got to account for future earnings. But to us and many others, that divergence between credibly, plausible, expected future earnings and this level of investment certainly looks bubbly.

He pointed to OpenAI, which has made nearly $1 trillion in Ai deals, but is slated to generate a mere $13 billion in revenue by 2030—maybe less. Brian Phillips at The Ringer posited that these kinds of huge discrepancies exist because Ai companies don’t really know what they’re selling. He wrote:

The hype around AI insists that it’s a world-transforming technology that will revolutionize every aspect of human society. The reality… is that AI companies are burning through staggering amounts of money (and fossil fuels) with no clear path to profitability, that the companies themselves aren’t super clear about what their products are for, and that many of those products have failed to perform in the applications they’ve been assigned to.

This year, 60% of stock market gains were tied to Ai, but that volume is deceptive. Three companies that are heavily invested in Ai—Nvidia, Microsoft, and Apple — make up 20% of global investable equity. By comparison, China’s is just 3%. Put simply, a significant amount of alleged earnings in the stock market this year went to just three entities, all of whom are invested in each other. And that has raised eyebrows among experts.

The arrangement of these deals is so circular that dollars spent by one player often return as revenue for another. This gives the impression of breakneck growth. (Source)

Luciano Floridi, founding director of the Digital Ethics Center at Yale University, who analyzes tech bubbles, concluded:

The release of ChatGPT in November 2022 accelerated this trend [of excitement], creating a perfect storm… of inflated expectations and speculative investment. This phenomenon bears striking resemblances to the tech bubbles of the past, suggesting that we may be witnessing the formation of an AI bubble.

The “incredible claims, revolutionary statements, and extraordinary promises” made about Ai are of the same vein as those made just before the dot-com bubble burst, he argued.

Another telltale sign of trouble, Floridi noted, is that market excitement and investment in Ai have severely outpaced actual development. This has led investors to ignore traditional financial metrics, and replace them with what he calls “new, unorthodox, and sometimes irrelevant measures of value.”

He wrapped up with a warning:

Unfortunately, it is in the nature of a bubble that, once it starts growing, it usually does not deflate gently. While it may still fizzle out gradually, a sudden burst is more likely.

Will Lockett, a journalist who writes prolifically on tech matters, described two specific occurrences that suggests a bubble burst is imminent:

It isn’t just that a bubble exists, but that it is being powered by what is arguably more than $1.2 trillion in mis-sold debt. I even called this revelation the AI bubble’s “Big Short” moment. Well, the actual “Big Short” guy, Michael Burry himself, has just revealed he is heavily shorting the very core of the AI bubble.

But his other example is much more alarming than Burry’s big bet—over a billion dollars!—on the burst happening. Locket noted that Deutsche Bank itself is also planning to short Ai-related stocks.

Why is that such a big deal? Let Lockett explain:

A major global investment bank is actively shorting its own gargantuan investments [in Ai] to protect itself from losses. That is insane! To take a short position large enough to stem this potential multi-billion-dollar hole would also mean that if the AI bubble didn’t pop, this short position would be so expensive it would almost certainly negate any gains from these investments. They are effectively putting themselves in a lose-lose situation.

Deutsche Bank is basically saying — through its actions — that it believes, with enough certainty to burn billions of dollars, that the Ai bubble is about to burst. It would rather lose a little (relatively speaking) than a lot. The losses will number in the billions in either case.

The International Monetary Fund and Bank of England have issued their own warnings of a “sharp market correction” based on “valuations [that] appear stretched.”

So what happens to everyone else if the bubble bursts?

Deutsche Bank’s head of FX Research, George Saravelos, told his clients:

AI machines — in quite a literal sense — appear to be saving the US economy right now. In the absence of tech-related spending, the US would be close to, or in, recession this year.

GDP growth in the United States in the first half of 2025 was a paltry 0.1%, if you exclude Ai-related investment. That’s where the US economy will be left if the Ai bubble bursts. Not a healthy situation, and possibly over-optimistic. After all, the burst would likely “put tens of thousands of people out of work, vaporize trillions of investment dollars, torpedo retirement and education funds, obliterate life savings, and ruin lives.” The remaining bottom of GDP would fall out.

This would have reverberations everywhere because what happens to the US economy spreads to the global economy. Moreover, the Ai sector exists in or affects other countries.

There are about 12,000 data centers (a number that includes colocation centers) across the globe, with nearly half of them in the US. The rate of new-builds now is primarily based on fulfilling the needs of Ai services.

If public-consumption-Ai does not become profitable, wouldn’t these data centers become liquefiable, reusable assets?

Shane Greenstein, Professor of Business Administration at Harvard Business School, doesn’t think so, at least not for many of them. According to him, data centers built for specific Ai tasks, including but not limited to training, “are being built with the expectation of a shorter shelf life even than usual data centers, perhaps even just 3-5 years. They are based on GPUs rather than CPUs, which sharply limits the ranges of tasks for which they’re useful.”

Furthermore, many of these facilities are being built in places that make reuse difficult or impossible. Jerry Kaplan, a longtime investor in Ai, told the BBC:

We're creating a new man-made ecological disaster: enormous data centers in remote places like deserts, that will be rusting away and leaching bad things into the environment, with no one left to hold accountable because the builders and investors will be long gone.

These ‘stranded assets’ would levy a great deal of cost onto everyone else beyond just cleaning them up. The reason is because the mass of infrastructure needed for them to function has mostly been funded by bets on future profitability (and concurrent rate hikes to consumers). Ari Peskoe, director of the Electricity Law Initiative at Harvard Law School, stated:

If you believe the projections here, we’re in the early stages of this buildout for industrial-scale computing, and therefore the infrastructure needed to power these facilities is just starting to get built, which means we’re all just starting to pay for it. So, there’s a possibility that the consumer impacts get a lot worse, unless there’s significant reforms of how utilities spread the costs of infrastructure.

Peskoe added that this energy infrastructure “could potentially be repurposed for other energy customers if the demand is there,” but data centers in the US currently use 4.4% of the total electricity output. What industry will arise in time to generate that level of demand?

The entire industrial sector in the US uses only 33%, so a new player would need to require as much as 10% of the total industrial sector energy use to compensate for the loss of data centers. In other countries, the situation is even more bleak. Data centers in Ireland, for example, consumed 22% of that nation’s electricity in 2024. If the energy goes unused, consumer prices will skyrocket.

The longer this hyper-investment in Ai goes on, the higher it will cost everyday people if it collapses. That’s because the infrastructure being built still needs to be paid for, but there will be no companies or investors around to foot any part of that bill. The inability to reuse these properties will exacerbate the harm.

There is another issue if it takes a few years for this bubble to burst. Nvidia is the company with the largest valuation in the world—over $4 trillion, or larger than 97% of the world’s national economies. Ninety percent of that value comes from selling chips to data centers, both standard and Ai centers, but the majority to the latter. If the bubble bursts in a few years, during which time many new Ai data centers are projected to be built, demand for Nvidia’s prime product will fall off a cliff.

Investors would lose trillions and Nvidia would probably have to nuke much of its workforce. In 2025, it employed around 36,000 people. Even if the company somehow salvaged half its valuation (an unlikely prospect), it would probably still have to layoff more than 10,000 people.

In short, if the Ai bubble bursts, it will be a global economic catastrophe. If there is good news, the biggest companies have the funds to crawl out of the ashes of an economic flameout. But make no mistake, they will do everything they can to shift the costs to everyone… anyone! else. It will be a repeat of 2008, only worse.

If the bubble doesn’t burst?

It’s certainly possible the biggest Ai companies will avoid letting it all fall down, but it probably won’t be much better for the public, and people might not notice as readily.

For example, to recoup potential losses from over-investment and under-performance, tech companies could sell their technology or infrastructure to governments hoping to increase their surveillance capabilities. Tech companies could sell tools to criminal dictators or militaries. They could turn their capabilities toward stealing even more from the people than they already have.

Smaller startups will be the ones to suffer complete losses, along with their workers and expendable workers in the big companies. A few businesses may benefit from buyouts by the big guys, but that will only lead to further enshittification for consumers. There’s an abundant record of small startups developing cool new things, only to sell out to the big players who then kill those cool new things (or muck them up with ads or subscription costs or whatever).

One thing is for sure, as evidenced by the fact that it’s already been happening. The big corporations will continue to find ways to extract wealth, probably by raising prices on existing products or services, and by charging for services that were—or should—be free. There’s even talk of shifting to a per-use fee for many services, particularly ChatGPT and its ilk (I’ll expound more on this later). Whether or not the bubble bursts in 2026, this slow fleecing will continue.

It seems clear that two things about Ai are true—the sector is very bubbly, but it is also a kind of Ponzi scheme. This is not an indictment of the technology itself, which undoubtedly has great potential we may not see for a very long time. But right now it is being wildly overhyped to generate investment funds, which has succeeded to the tune of almost two trillion dollars.

Because those investment funds are so far beyond what Ai can realistically produce, companies will play fast and loose with the truth to keep the sector afloat for as long as possible. The largest companies undoubtedly believe that no matter how things proceed, they will come out just fine. They can either milk investors and customers until it crashes, then make the ‘too big to fail’ argument to get government bailouts. Or they can try to keep the hype alive until, by some miracle, Ai finally can do something to justify its valuation. Either way, they will have succeeded in the biggest heist of public money in history.