Is AI in a Bubble in 2026? Valuations, Spending & the Risks Investors Are Watching

Written by

Daniel Roy

Updated: September 17, 2026

Artificial intelligence has become one of the biggest investment themes in global markets.

Billions are being spent on chips, data centres, energy infrastructure, cloud computing and AI models. At the same time, some of the world’s largest technology companies have reached market valuations that would have seemed extraordinary only a few years ago.

That has revived an uncomfortable question:

Is AI in a bubble?

Receive your FREE exclusive PDF straight to your inbox and discover 50 of the best AI tools for 2026.

There is no simple yes-or-no answer.

There are genuine signs of speculative excess: elevated technology valuations, enormous capital spending, increasingly complex financing arrangements and private AI companies raising money at extremely high valuations.

But unlike the dot-com boom, much of today’s AI investment is also being led by companies with substantial revenue, profits and cash flow.

The more useful question is therefore not:

“Is all of AI a bubble?”

It is:

“Which parts of the AI boom depend on extremely optimistic assumptions, and what would happen if those assumptions prove wrong?”

This article looks at the evidence on both sides.

What Does “AI Bubble” Actually Mean?

A financial bubble occurs when asset prices rise substantially because investors expect future growth, while those prices become increasingly difficult to justify using the underlying cash flows, profits or economic value of the assets.

That does not mean the technology itself is fake.

History contains several examples where a transformative technology attracted too much capital too quickly.

Railways were real.

The internet was real.

Telecommunications networks were real.

Yet each experienced periods in which investment and valuations ran far ahead of the returns that businesses eventually earned.

The Bank for International Settlements has explicitly drawn parallels between today’s AI investment boom and earlier infrastructure booms, including railways, electrification and the dot-com era. Its concern is not that AI has no value, but that competition could encourage companies to invest more capital than future returns ultimately justify.

Why the AI Bubble Debate Has Intensified in 2026

The debate has become more serious because the scale of AI investment has increased dramatically.

The BIS estimates that the world’s five largest technology companies will invest more than $1 trillion in AI across 2025 and 2026, while global AI investment could reach roughly $4 trillion by 2030.

Reuters reported in September that industry spending is expected to exceed $795 billion in 2026 and could pass $1 trillion in 2027.

This investment is paying for an enormous physical buildout:

data centres, GPUs, custom AI chips, power generation, networking equipment and cooling systems.

That creates both opportunity and risk.

If demand for AI computing continues expanding rapidly, much of that infrastructure could be heavily utilised.

If revenue growth or AI adoption disappoints, the industry could be left with enormous amounts of expensive capacity producing lower returns than investors expected.

Evidence That AI Valuations May Be Stretched

The strongest argument for the bubble case is not that AI companies lack revenue.

It is that current valuations and spending assume very strong future growth for a long time.

The BIS said in September that equity valuations are elevated and increasingly concentrated among a small group of companies at the centre of AI development.

It also warned that AI investment is becoming increasingly connected through debt, private credit and what it calls circular financing — arrangements where chipmakers, hyperscalers and AI companies invest in one another while also entering large purchasing agreements.

That matters because an interconnected boom can amplify both gains and losses.

If one part of the ecosystem suddenly reduces spending, the impact can spread through chip suppliers, data-centre operators, energy providers, cloud companies and private lenders.

The Infrastructure Spending Numbers Are Enormous

One of the clearest signs of how aggressive the current AI buildout has become is the capital spending of the largest technology companies.

Alphabet increased its 2026 capital-expenditure forecast to approximately $195 billion to $205 billion after reporting rapid growth in Google Cloud.

Meta expects 2026 capital expenditure of roughly $130 billion to $145 billion, largely as it expands AI computing infrastructure.

Meta’s spending also shows the financial trade-off involved. Reuters reported that second-quarter free cash flow fell to only $784 million as investment in AI infrastructure consumed most of its operating cash flow.

Meanwhile, Microsoft reportedly plans to expand its global data-centre capacity to approximately 38 gigawatts by 2032 — more than three times its current footprint.

These are not small experimental investments.

The world’s largest companies are making infrastructure decisions that could shape their finances for years.

But There Is Also Strong Evidence Against the Simple “AI Is Dot-Com 2.0” Argument

The other side of the debate is important.

Many of the companies financing today’s AI boom are already highly profitable businesses.

That differs significantly from the late-1990s internet bubble, when investors often valued companies with minimal revenue and no sustainable earnings.

The IMF has specifically highlighted this difference.

In January 2026, IMF chief economist Pierre-Olivier Gourinchas noted that many of the companies driving AI investment have strong earnings and have been able to finance a substantial share of their spending from their existing businesses.

There is also real demand behind parts of the infrastructure buildout.

Nvidia Shows Why the Bubble Question Is Complicated

Nvidia provides perhaps the clearest example.

Its market value has climbed to around $5 trillion, making it one of the most valuable companies in history.

A valuation of that size obviously requires enormous expectations about future AI demand.

But Nvidia is not simply rising on a story.

For its fiscal second quarter reported in August 2026, Nvidia recorded:

$96.2 billion in quarterly revenue, up 106% from a year earlier.

Its Data Center business generated $89 billion, up 117% year over year.

Gross margins were approximately 75%.

Nvidia subsequently forecast roughly 70% revenue growth for the following fiscal year, reinforcing expectations that demand for AI infrastructure remains very strong.

That does not tell us whether Nvidia’s share price is cheap or expensive.

It demonstrates why today’s situation cannot simply be described as a market built entirely on imaginary revenue.

The earnings are real.

The difficult question is whether future earnings can grow quickly enough to justify today’s expectations.

Cloud Revenue Is Also Growing Rapidly

The same tension exists among the hyperscalers.

Alphabet’s Google Cloud revenue grew 82% year over year to $24.8 billion in its second quarter of 2026.

Microsoft also reported cloud performance strong enough to ease some investor concerns that its massive AI spending was running ahead of monetisation.

Amazon reached a $3 trillion market valuation in August after strong earnings and accelerating AWS growth helped reinforce confidence in AI-related cloud demand.

Again, that doesn’t prove current valuations are justified.

But it means there is a measurable commercial engine underneath the investment boom.

Where Bubble Risk May Be More Concentrated

Rather than labelling entire sectors “safe” or “dangerous,” it is more useful to look at the characteristics that make an investment more dependent on optimistic assumptions.

Private AI companies

Private markets can involve extremely high valuations despite limited public financial disclosure.

Some frontier AI companies require enormous amounts of additional capital to train models, build infrastructure and compete globally.

That creates a difficult valuation problem: investors are trying to price companies based partly on revenue that may not exist for many years.

Highly leveraged infrastructure businesses

Independent AI data-centre and GPU-cloud companies can be vulnerable because they require huge upfront investment while depending heavily on continued demand for expensive computing capacity.

Reuters Breakingviews has compared parts of the current data-centre boom with the telecom-network overbuilding of the late 1990s.

Companies whose valuation depends mainly on an “AI story”

Another warning sign is when a company’s market narrative changes much faster than its financial performance.

A business adding “AI” to its investor presentation does not automatically become an AI winner.

Revenue growth, margins, customer adoption and cash generation still matter.

The Hyperscalers Have a Different Risk Profile

Microsoft, Alphabet, Amazon and Meta are also spending enormous amounts on AI.

The difference is that their businesses are diversified.

They generate revenue from areas such as:

That gives them larger existing cash-generating businesses from which to fund AI investment.

It does not make their stocks inherently low-risk.

A profitable company can still be overvalued.

The relevant question is whether future earnings justify the price investors are currently willing to pay.

A New Risk: Debt and Private Credit

One of the more important developments in 2026 is that AI expansion is increasingly being financed with debt.

Earlier in the boom, much of the infrastructure investment could be financed from the enormous cash flows of technology companies.

As spending has accelerated, borrowing and private credit have become more important.

The BIS has warned that capital expenditure at some of the largest AI-linked companies is beginning to outpace cash flows, increasing dependence on debt and private credit.

This creates a new channel of risk.

If the expected returns on AI infrastructure fall, losses may not remain confined to technology shares.

They could affect lenders and private-credit investors as well.

AI Market Concentration Is Another Risk

A relatively small group of AI-related companies now accounts for a large amount of market value, capital spending and investor expectations.

That concentration can make broader stock indexes increasingly sensitive to developments in AI.

The IMF has warned that a downward revision to expected AI profitability or productivity could cause a sharp retrenchment in technology investment and a correction in highly valued, technology-heavy equity markets.

The important point is not that such a correction will happen.

It is that a disappointment affecting a relatively small number of very large companies could have an unusually large effect on broader markets.

What Would Strengthen the “AI Bubble” Case?

The argument that AI is in an unsustainable investment bubble would become stronger if several things started happening at once.

The clearest warning signs would include slowing cloud and AI revenue growth while capital expenditure stayed extremely high; falling utilisation of data centres and AI chips; worsening free cash flow across major spenders; increasing reliance on debt; weaker corporate adoption; falling model prices without equivalent improvements in operating efficiency; and major AI companies repeatedly missing revenue expectations.

The relationship between spending and financial returns matters more than spending alone.

A company investing $100 billion isn’t necessarily overinvesting if that infrastructure generates substantially greater future cash flows.

The problem appears when investment continues rising while returns deteriorate.

What Would Weaken the Bubble Argument?

The bubble thesis becomes less convincing if AI continues producing large measurable commercial returns.

That could include continued rapid cloud revenue growth, strong demand for inference, rising enterprise adoption, productivity gains becoming visible across more industries, improving margins for AI services and hyperscalers maintaining strong cash generation despite infrastructure spending.

The latest Nvidia and cloud-company results provide some evidence for this side of the argument.

The Productivity Question May Decide the Long-Term Outcome

Ultimately, trillions of dollars are being invested because companies expect AI to increase economic productivity.

There is evidence that AI can improve productivity substantially for particular tasks.

But translating those task-level gains into economy-wide productivity growth takes time.

The IMF said in March that AI investment is already contributing materially to US growth, while cautioning that lasting benefits depend on how quickly the technology spreads through the broader economy.

Reuters Breakingviews similarly noted in September that AI adoption has expanded rapidly while the productivity gains visible in economy-wide statistics remain relatively modest so far.

That gap is important.

If productivity eventually accelerates sharply, current investment may look more rational in hindsight.

If the productivity dividend remains modest, today’s spending could prove much harder to justify.

Is This Another Dot-Com Bubble?

There are similarities.

Both periods featured:

But there are major differences.

Many leading AI companies are already among the world’s most profitable businesses.

Demand for AI chips and cloud infrastructure is producing substantial revenue today.

And AI products are already being deployed across software development, advertising, customer service, research and enterprise workflows.

The dot-com comparison is therefore useful as a warning about valuation and overinvestment.

It is less useful if interpreted as evidence that AI itself is mostly hype.

Could AI Stocks Still Fall Sharply Even If AI Succeeds?

Yes.

This may be the most important point in the entire debate.

A technology can transform the economy while investors still lose money buying companies at the wrong price.

A company can:

and still experience a falling share price if investors previously expected even faster growth.

This is known as valuation compression.

The distinction between:

“AI will succeed”

and:

“AI stocks are correctly priced”

is crucial.

They are not the same question.

What Investors Can Monitor Instead of Trying to Predict a Crash

Predicting exactly when a market bubble will burst is notoriously difficult.

A more useful framework is to watch whether financial performance is keeping pace with expectations.

Relevant indicators include capital expenditure relative to operating cash flow, AI and cloud revenue growth, data-centre utilisation, gross margins, debt growth, free cash flow, enterprise AI spending, chip demand and whether productivity improvements become visible outside the technology sector.

The BIS itself has argued that the scale and speed of the AI investment boom warrant caution without saying that a bust is inevitable.

So, Is AI in a Bubble in 2026?

The evidence supports a more nuanced conclusion than either extreme.

There are legitimate bubble concerns.

AI-related valuations are elevated, infrastructure spending is historically large, financing is becoming more complex and some businesses depend heavily on extremely optimistic assumptions about future AI demand.

At the same time, the boom is producing very real revenue.

Nvidia’s data-centre business is growing at triple-digit rates. Cloud companies are reporting strong AI-related growth. Large technology companies funding the boom remain highly profitable.

The most defensible conclusion is therefore:

AI may be experiencing an investment boom with pockets of speculative excess, rather than one uniform bubble across the entire sector.

Whether today’s spending eventually looks visionary or excessive will depend largely on how quickly AI generates sustainable revenue and productivity gains relative to the enormous amount of capital being invested.

Frequently Asked Questions

Is AI in a bubble right now?

There are signs of elevated valuations and unusually aggressive investment, and institutions including the BIS and IMF have warned that disappointment around AI profitability could cause a significant market repricing.

However, many leading AI-related businesses are also reporting strong revenue and earnings growth, which makes the current environment different from a purely speculative bubble.

Is Nvidia in an AI bubble?

There is no objective way to determine from today’s data whether an individual stock is definitively “in a bubble.”

Nvidia combines an extremely large market valuation with exceptionally strong current growth. Its latest reported quarter produced $96.2 billion in revenue, up 106% year over year.

The investment debate therefore centres on whether that growth can remain strong enough to justify expectations embedded in the company’s valuation.

Why is so much money being spent on AI?

Modern AI systems require enormous quantities of computing infrastructure, including GPUs, data centres, networking equipment, electricity and cooling.

Major technology companies are also competing to secure enough capacity to train new models and serve rapidly growing inference demand.

What could cause AI stocks to fall?

Possible triggers include weaker-than-expected earnings, slower AI adoption, falling infrastructure demand, rising financing costs, lower margins, regulation, technological disruption or investors simply reducing the valuations they’re willing to pay for future growth.

Could AI succeed while AI stocks fall?

Yes.

A technology can become commercially successful while share prices decline if investors previously priced companies for even stronger growth.

Business success and investment returns are related, but they are not identical.

Is today’s AI boom the same as the dot-com bubble?

There are similarities in valuations, enthusiasm and infrastructure investment.

A major difference is that many of today’s largest AI investors are already highly profitable companies generating substantial revenue.

The comparison is therefore useful, but imperfect.

Final Takeaway

The AI bubble debate is ultimately a debate about expectations.

Very few serious observers dispute that AI is an important technology.

The uncertainty is whether companies will generate enough future cash flow and productivity growth to justify the extraordinary amount of money currently being invested.

Both things can be true at once:

AI can transform the economy.

Investors can still overpay for that transformation.

That distinction is more useful than trying to reduce the entire AI market to a simple “bubble” or “not a bubble” label.

Receive your FREE exclusive PDF straight to your inbox and discover 50 of the best AI tools for 2026.