The most consequential argument in technology right now is not about which model is smartest or which chip is fastest. It is about whether the entire enterprise is a generational revolution or the largest financial bubble ever inflated — and the unsettling answer, increasingly, is that it may be both at once.
On one side stand the people writing the checks and the engineers watching AI rewrite their workflows in real time, who see undeniable, fast-growing revenue and productivity. On the other stand economists, fund managers, and even the Federal Reserve, who see capital spending so far ahead of returns that a correction looks less like a risk than a matter of timing. Both sides are looking at the same year — 2026 — and reaching opposite conclusions. Understanding why is the key to understanding the stakes for anyone with a job, a retirement account, or a stake in the companies driving the boom.
The Case That It's a Bubble
Start with the number that frightens people. The world's largest technology companies are on track to spend roughly $725 billion on AI infrastructure in 2026, up from about $162 billion in 2022, and broader estimates of total AI spending this year run into the trillions. That capital expenditure has become macroeconomically significant: by some analyses, AI-related capex accounted for a large share of US GDP growth in the first quarter of 2026 and now represents a meaningful fraction of the entire economy. The Federal Reserve has named AI one of the top systemic risks to financial stability, and banks including J.P. Morgan have warned that the economics of AI infrastructure are increasingly strained because revenue has not kept pace with spending.
The skeptics are blunt. Analysts at Bernstein have called a bubble the "likely outcome." Man Group has warned that the financial architecture of the boom is "oversized, overleveraged, and too dependent on a small set of interconnected actors." One strategist, Julien Garran of MacroStrategy Partnership, has called it the largest bubble in history. Even a Wells Fargo analyst who urged clients to buy in described the AI capex surge as "euphoric."
Three structural worries sit beneath the rhetoric. The first is circular financing: a tight web of deals in which the same dollars cycle between a handful of companies and inflate everyone's revenue. Nvidia has invested in OpenAI and committed enormous sums to it — money that flows back to Nvidia through chip purchases — while Microsoft owns a large stake in OpenAI and is also its primary cloud provider, booking Azure revenue that is partly spent on Nvidia hardware. Critics argue these arrangements manufacture the appearance of demand without creating independent economic value.
The second worry is the missing return on investment. Multiple enterprise surveys across 2025 and 2026 found that something like 95% of corporate AI pilots delivered no measurable financial return — a striking figure given how much has been spent. The third is depreciation: the graphics processors at the heart of the build-out lose value quickly and may need replacing far sooner than the decades-long timelines used to justify some data-center economics.
Then there are the valuations. SpaceX is pursuing an IPO at a price Morningstar pegs at nearly twice its estimate of fair value, opening at roughly 94 times sales. The database startup Supabase reached a $10.5 billion valuation at a price-to-sales multiple near 150. China's Moonshot AI is raising at a similar multiple. To bears, these are dot-com echoes — prices that only make sense if growth continues at rates almost no company has ever sustained.
The Case That It's a Revolution
Now look at the other ledger, and the picture changes. The revenue is not hypothetical, and it is not small.
Anthropic reached roughly $30 billion in annualized revenue by April 2026, up from about $1 billion at the start of 2025 — a 30-fold increase in 15 months, a growth curve almost without precedent in enterprise software. More than 1,000 of its business customers are each spending over $1 million a year, a number that doubled in under two months. In April 2026, Anthropic surpassed OpenAI in business adoption for the first time, according to data from the corporate-spending platform Ramp, and roughly 80% of its revenue comes from enterprises rather than consumers — the signature of a product companies rely on, not a novelty they are trying once.
The productivity gains show up in hard usage data, not just testimonials. Claude Code, Anthropic's agentic coding tool, became the fastest-growing product in the company's history, and by one estimate roughly 4% of all public commits on GitHub worldwide were being authored by it — double the share of a month earlier. Uber's chief technology officer said the company burned through its entire 2026 AI budget in four months as engineer adoption jumped from 32% to 84% and about 70% of committed code began coming from AI. Cursor's own data, drawn from 18 months of usage, shows the rate at which developers write code has roughly doubled year over year. These are not the metrics of a fad; they are the metrics of a tool that has embedded itself in how real work gets done.
The bulls also dispute the financial-fragility narrative on its own terms. For all the talk of reckless spending, the largest AI investors are, in aggregate, spending what they earn rather than what they borrow: the ratio of capital expenditure to free cash flow across the big spenders sits below one, which means the build-out is being funded largely from profits, not debt. That is not what the late stages of a classic credit-fueled bubble usually look like.
No one embodies the bull case more than Nvidia chief executive Jensen Huang, whose chips sit at the center of the build-out. At the World Economic Forum in Davos in January 2026, Huang dismissed bubble fears as a misreading of what he called "the largest infrastructure build-out in human history", arguing the spending is large because the industry must build the full stack beneath AI — energy, chips, cloud, models, and applications. As evidence of real demand rather than speculation, he pointed to how hard it has become to rent an Nvidia GPU at all, with spot rental prices rising even for two-generation-old chips, and to established firms such as the drugmaker Eli Lilly shifting research budgets from wet labs to AI supercomputing rather than burning venture capital. He has called Big Tech's roughly $700 billion in capex "just the start." The caveat is unavoidable: Huang is also the boom's single largest beneficiary, and the same Nvidia that anchors his optimism sits at the heart of the circular-financing arrangements the bears distrust — which is exactly why the argument is so hard to settle from the inside.
Why Both Sides Can Be Right
The instinct to force a single answer — revolution or bubble — may be the mistake. History suggests the two often arrive together. The railways of the 19th century were a genuine transformation of the economy and the subject of a speculative mania that ruined thousands of investors. The internet permanently changed commerce and communication and still produced the dot-com crash, in which the technology kept growing while the stock prices collapsed. Electricity, telephones, and the automobile each drew waves of capital that overshot near-term demand before the technology delivered on its promise.
The pattern is consistent enough to be a near-rule: when a technology is real and important, capital floods toward it faster than the technology can absorb, infrastructure gets overbuilt, valuations detach from cash flow, and a repricing follows that wipes out the weakest players without erasing the underlying progress. The fiber laid during the dot-com era was overbuilt and bankrupted its builders — and then quietly powered the broadband internet for the next two decades.
Applied to AI, that lens dissolves the false binary. The revolution can be real — the revenue, adoption, and productivity gains are measurable and large — while the financing can be overextended, the valuations stretched, and a painful correction still ahead. The question worth asking is not "is AI real?" but "is the capital structure built around it able to survive a repricing without breaking the real thing underneath?"
What Actually Decides It
A few specific variables will settle the argument over the next year or two, and they are worth watching more closely than any single headline.
The first is whether revenue catches capex. The bull case requires that the hundreds of billions flowing into chips and data centers convert into durable, profitable revenue at scale — and the 95%-of-pilots-failed figure is the bears' strongest evidence that it has not yet. If enterprise ROI broadens from a handful of use cases (coding, customer support, research) to the wider economy, the spending is justified; if it stalls, the overbuild becomes a reckoning.
The second is the concentration and circularity risk. Because so much of the boom runs through a small set of interlocked companies, a stumble at any one of them — a missed earnings number, a canceled commitment, a wave of GPU depreciation hitting the books — could cascade in a way that a more diversified market would absorb. The third is the IPO wave: SpaceX, and reportedly Anthropic and OpenAI, are moving toward public markets, and public investors, unlike private ones, reprice in real time. How those offerings trade will be the market's first broad verdict on whether the private valuations were visionary or delusional.
What It Means for You
For an ordinary reader, the practical takeaway is not to pick a team but to hold both truths at once. The technology is genuinely transforming knowledge work, and the evidence that it is doing so is stronger than at any prior point in the hype cycle. At the same time, the financial scaffolding around it shows the classic strain marks of a bubble, and the people warning about that strain include the Federal Reserve and some of the most respected analysts on Wall Street.
That combination has a specific implication: a correction in AI asset prices, if it comes, would not mean the revolution was fake — any more than the dot-com crash meant the internet was fake. It would mean the market got ahead of the timeline. For workers, the durable shift is in how jobs are done, and that is already underway regardless of where stock prices go. For investors, the lesson of every prior technology boom is that the winners are rarely obvious in advance and the prices paid at the peak rarely look wise in hindsight. The revolution and the bubble are not competing theories of the same event. They are, most likely, two names for two different things happening at the same time — and the skill the next two years will demand is telling them apart.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
