AI Spending Is Exploding in 2026 — But Is the Boom Becoming a Bubble?
Artificial intelligence has gone from being a promising technology to one of the biggest investment stories in the world.
Companies are building enormous data centres, buying huge quantities of AI chips and signing multibillion-dollar computing contracts. Investors are pouring money into AI companies, while governments and technology firms are racing to secure the electricity and infrastructure needed to keep the industry growing.
The numbers are becoming difficult to ignore.
Global spending on data centres could eventually exceed $30 trillion by 2050, according to a projection cited by Reuters from PwC. At the same time, AI companies are making individual commitments worth hundreds of billions of dollars. Anthropic, for example, has disclosed plans involving at least $518 billion in cloud, computing and infrastructure obligations over the coming years.
That raises an uncomfortable question:
Is the world building the infrastructure for an AI revolution, or is too much money chasing a technology whose returns have not yet caught up with the hype?
Why Is So Much Money Going Into AI?
The basic argument behind the spending is straightforward.
The more powerful AI models become, the more computing power they require.
Training advanced models can involve enormous amounts of specialised hardware. Running those models for millions of users also requires data centres filled with processors, storage systems and networking equipment.
That has created a chain reaction.
AI companies need computing capacity.
Cloud companies need data centres.
Data centres need chips and networking equipment.
And all of them need electricity.
The result is an infrastructure race that extends far beyond companies making AI software.
The Numbers Are Getting Extraordinary
Anthropic provides one of the clearest examples of how large the spending has become.
According to its IPO filings reviewed by Reuters, the company expects to spend at least $518 billion over a decade on AI infrastructure through agreements with several partners. About 80% of those commitments are reportedly non-cancelable or require payment regardless of how much computing capacity is actually used.
The commitments include more than $111 billion with Google, $110 billion with Amazon and $31.4 billion with Microsoft.
Anthropic also has major equipment-related obligations involving Broadcom.
G7 Releases 100 Million Barrels of Oil:
Those figures are enormous even by technology-industry standards.
And Anthropic is only one company in a much larger race.
AI Spending Is Becoming an Energy Story
There is another part of the AI boom that is easy to overlook: electricity.
A data centre does not run on software alone.
Thousands of processors operating around the clock require huge amounts of electricity, while the equipment also generates heat that has to be removed.
That is why AI is increasingly influencing decisions about power plants, electricity grids and energy infrastructure.
In the United States, rising electricity demand from AI data centres, electric vehicles and cryptocurrency mining is already contributing to a push for additional power generation. Reuters reported that the U.S. government plans to lend about $4.2 billion to Vistra to increase output at several nuclear power stations.
So the AI investment story is no longer simply about technology companies.
It is becoming an infrastructure story.
But Where Are the Returns?
This is where the argument becomes more complicated.
Technology companies have already demonstrated that people are willing to pay for AI products.
Businesses are using AI for coding, customer service, research, writing, analysis and other tasks.
AI companies are generating billions of dollars in revenue.
But the amount being spent on infrastructure is growing at an extraordinary pace.
The question for investors is whether future AI revenue will grow quickly enough to justify the enormous cost of building all that computing capacity.
Reuters reported that Bain estimates U.S. hyperscalers and other companies involved in the AI race could need more than $4.2 trillion in additional revenue over five years to fund the planned buildout.
That is the heart of the bubble debate.
It isn’t necessarily an argument that AI is useless.
It is an argument about price, timing and returns.
Does That Mean AI Is a Bubble?
Not necessarily.
A technology can transform the economy while investors still lose money during a period of excessive speculation.
History has plenty of examples.
Railways transformed transportation, but railway companies still experienced spectacular financial failures.
The internet changed communication and commerce permanently, even though the dot-com bubble eventually collapsed.
The same thing could happen with AI.
The technology could prove enormously valuable while some of today’s companies, projects or investments turn out to have been overpriced.
That distinction is important.
The real question isn’t simply:
“Is AI real?”
It clearly is.
The better question is:
“Will AI generate enough economic value to justify the enormous amount of money being spent on it?”
Productivity Is the Missing Piece
One of the biggest arguments supporting the AI boom is productivity.
If AI allows workers to accomplish in an hour what previously took an entire day, businesses could produce more with fewer resources.
That could eventually raise economic growth and create new industries.
But the productivity gains economists are looking for have not yet appeared on the scale required to justify some of the most aggressive investment assumptions.
Reuters reported that JPMorgan wrote in August that broad-based U.S. productivity gains associated with AI remained difficult to see.
That doesn’t mean AI isn’t improving individual businesses.
It means the improvement hasn’t yet translated into a dramatic economy-wide productivity surge.
And that may take time.
AI Could Still Change the Economy Completely
There is a danger in judging a technological revolution too quickly.
Some technologies take years before their biggest economic effects become obvious.
Electricity, for example, did not immediately transform every factory simply because electric power became available.
Companies had to redesign their operations around it.
The internet followed a similar path.
The early web generated enormous excitement, followed by a financial crash. But the infrastructure that survived eventually became the foundation of modern commerce, communication and entertainment.
AI could follow a similar pattern.
Some investments may fail.
Some companies may disappear.
But the underlying technology could still become much more important than it is today.
What Happens If AI Spending Slows Down?
A slowdown would not necessarily mean the end of AI.
It could simply mean that companies become more selective.
Instead of building as much infrastructure as possible, they could focus on projects that generate measurable returns.
Investors could also start paying more attention to revenue, profits and cash flow instead of customer growth and ambitious promises.
That transition could be uncomfortable for some AI companies.
Companies that depend on continuously rising valuations or enormous amounts of outside financing could face the most pressure.
What If AI Demand Grows Faster Than Expected?
There is another possibility.
AI could become so useful that today’s infrastructure eventually turns out to be insufficient.
If AI agents, robotics, scientific research and automated business systems become widely adopted, computing demand could rise dramatically.
That would make today’s infrastructure spending look less excessive in hindsight.
Anthropic’s own projections include scenarios in which AI produces significant increases in economic growth, although the company also acknowledges substantial uncertainty around those outcomes.
The problem is that nobody knows which scenario will happen.
AI Is Already Affecting Jobs
The economic debate isn’t only about corporate profits.
AI is also beginning to change the labour market.
Researchers have reported weaker early-career hiring in some occupations that are particularly exposed to AI, including certain white-collar roles.
Reuters cited Stanford research finding that employment among U.S. workers aged 22 to 25 in AI-exposed industries was 19% lower than employment in occupations less exposed to AI.
That does not prove that AI is responsible for every job loss.
Hiring is affected by interest rates, economic growth, company budgets and many other factors.
But it is becoming increasingly difficult to discuss AI’s economic impact without discussing work.
The AI Boom Could Have a Long Way to Run
Despite the concerns, it would be premature to assume the AI boom is about to collapse.
There is genuine demand for AI.
Companies are paying for AI services.
Consumers are using AI tools.
Businesses are experimenting with automation.
And governments are increasingly treating AI infrastructure as strategically important.
The bigger uncertainty is whether the industry can turn that demand into enough long-term revenue to support the enormous infrastructure investments now being made.
What Should Investors Be Watching?
Rather than focusing only on the latest AI valuation, investors may want to watch several underlying indicators.
Revenue growth is one.
Are AI companies generating enough actual revenue to support their valuations?
Profitability is another.
Rapid revenue growth means less if costs are growing even faster.
Data-centre utilisation also matters.
Building a data centre is expensive. If computing capacity sits unused, the investment becomes much harder to justify.
Electricity availability could become another major constraint.
AI companies may have enough money to buy chips but still struggle to obtain the power needed to operate them.
And finally, investors will want to see whether AI produces measurable productivity gains across the wider economy.
So, Is AI in a Bubble?
The honest answer is that nobody knows yet.
There are clearly signs of exuberance.
The amounts being committed to AI infrastructure are enormous, valuations are high and some of the industry’s most ambitious predictions depend on productivity gains that have not yet fully appeared.
At the same time, dismissing AI as just another speculative craze would be equally simplistic.
The technology is already being used by millions of people and thousands of businesses.
It is improving rapidly, and the infrastructure being built today could support technologies that have not even been fully developed yet.
The more likely outcome may be somewhere between the two extremes.
AI could be genuinely transformative and some of today’s investments could eventually prove to have been excessive.
The Bottom Line
The AI boom has reached a point where the biggest question is no longer whether companies will spend money on artificial intelligence.
They already are.
The question is whether the economic value created by AI will eventually catch up with the enormous cost of building the infrastructure behind it.
With global data-centre investment potentially reaching tens of trillions of dollars over the coming decades, and individual AI companies making infrastructure commitments measured in hundreds of billions, the stakes are unusually high.
If AI delivers the productivity gains its supporters expect, today’s spending could eventually look like the foundation of a new industrial era.
If those gains arrive too slowly, investors could discover that even a revolutionary technology can be an expensive investment.
For now, the AI race continues.
And the biggest test may not be how powerful the next AI model becomes.
It may be whether the world can make enough money from it to pay for everything being built around it.
Frequently Asked Questions
How much is being spent on AI in 2026?
AI spending is spread across chips, data centres, cloud computing, electricity and software, making one precise global figure difficult to establish. However, the scale is enormous, with major technology companies committing hundreds of billions of dollars to infrastructure. Reuters reported that five major AI hyperscalers were expected to spend more than $800 billion in 2026.
Is AI a bubble?
There are concerns that some AI investments and valuations may be excessive, but that does not mean the underlying technology is a bubble. AI is already being used commercially, while the debate is mainly about whether future economic returns will justify today’s spending.
Why do AI companies need so much infrastructure?
Advanced AI systems require large amounts of computing power for training and operation. Data centres provide the processors, networking, storage and electricity needed to run those systems.
Why does AI require so much electricity?
AI processors consume substantial amounts of electricity, particularly when thousands of chips operate simultaneously in large data centres. Cooling the equipment also requires additional energy.
Could AI spending eventually slow down?
Yes. Companies could reduce infrastructure spending if AI demand, revenue growth or productivity gains fail to meet expectations. A slowdown would not necessarily mean that AI itself is failing.
Could AI investment still be worth it?
Yes. If AI produces major productivity improvements, creates new markets and becomes deeply integrated into business and scientific research, today’s infrastructure could eventually generate substantial economic value.
About the Author
