AI investment boom raises fears of global spending correction as trillion-dollar bets outpace returns

The global race to dominate artificial intelligence is fuelling one of the largest technology investment cycles in history, with the world’s biggest technology companies committing more than $1 trillion to AI infrastructure over 2025 and 2026.

However, industry experts warn that the spending surge could end in a painful correction if revenue growth fails to keep pace.

The Bank for International Settlements warns that the $1 trillion AI capital expenditure by major tech firms risks creating a bubble resembling the dot-com era. If expected returns fail to materialise, it could trigger a sharp spending pullback, a stock market correction, and a credit crisis across private markets.

In its Annual Economic Report 2026, the BIS cautioned that massive AI-related spending by the world’s largest tech companies, exceeding $1 trillion across 2025 and 2026, is outpacing current earnings and free cash flow.

The warning echoes concerns raised by the Bank for International Settlements, which has cautioned that rapid AI-related capital expenditure could expose companies and investors to risks similar to previous technology investment booms, including the dot-com era, if expected returns fail to materialise.

The AI frenzy has caused spending on advanced chips, hyperscale data centres, cloud infrastructure, electricity generation and AI startups as companies race to secure computing capacity for the next generation of intelligent systems.

Among the biggest spenders is Microsoft, which plans to spend approximately $80 billion during its 2025 fiscal year to build out AI-enabled data centers. The company plans to use this massive capital investment to train complex AI models and deploy cloud-based applications globally, with over half of the investment allocated to the United States.

Meta’s 2025 capital expenditure range of $64 billion to $72 billion has already been surpassed by its subsequent investments. The company is spending $125 billion to $145 billion on capital expenditures in 2026, roughly double its 2025 budget, as it pivots further into AI infrastructure and compute capacity.

Amazon expects capital expenditure of more than $100 billion this year, with most of the investment directed at AI and cloud infrastructure, while Alphabet increased its 2025 capital expenditure forecast to about $85 billion, citing growing demand for AI computing capacity.

The investment wave extends beyond cloud providers as OpenAI, Oracle and SoftBank announced the Stargate Project, a venture expected to invest up to $500 billion over several years in AI infrastructure across the United States, underscoring the scale of global ambitions in artificial intelligence.

The boom has also propelled NVIDIA into the world’s most valuable companies as demand for its graphics processing units (GPUs), the chips that power large AI models, continues to soar.

Read also: Meta unveils Muse Image as Zuckerberg escalates AI race against OpenAI, Google

Spending grows faster than returns

While investment continues to increase, there are further questions over returns, if AI revenues can justify the enormous capital put into it.

Oluwajuwon Omotayo, founder of Comply54, said the financial mechanics behind the AI boom resemble previous investment cycles where infrastructure spending raced ahead of proven commercial demand.

“The five biggest tech companies are set to spend over a trillion dollars combined on AI infrastructure across 2025 and 2026, and that spending is now growing faster than the actual cash they’re bringing in,” he said.

According to him, part of the apparent demand is being driven by circular investment arrangements in which major technology firms fund AI startups that subsequently spend much of the capital purchasing computing resources from the same companies.

“On paper it looks like strong demand, but some of that money is really just moving in a loop between the same few companies,” Omotayo said, comparing the trend to the railway investment boom of the 19th century and the dot-com bubble of the late 1990s.

The concern is not only financial

Omotayo said that while companies can cut AI spending if markets weaken, the risks associated with deploying AI systems into sensitive business operations could have longer-lasting consequences.

“Companies have been plugging AI agents into moving money, handling customer data and verifying identities much faster than they’ve built real checks to ensure those AI agents comply with the law,” he said.

“If an AI agent wrongly processes someone’s private data or authorises a payment it shouldn’t have, the damage has already been done.”

Rising AI bills add pressure

Akande Adedayo, specialist solutions architect at 54Pay Technologies, said the economics of enterprise AI differ significantly from what consumers experience.

“Most people only see ChatGPT subscriptions costing around $20 per month,” he said. “For businesses building products on OpenAI, Claude or Gemini, costs are usage-based. Every prompt, API call, document processed, image generated or AI agent running continuously increases the bill.”

He cited OpenClaw, whose creator reportedly disclosed using about $1.3 million worth of OpenAI tokens in 30 days, illustrating how quickly operating costs can escalate for AI-native applications.

“The bigger risk is not that AI is fake,” Adedayo said. “The danger is overinvestment—too much money chasing returns that may take longer to arrive than markets expect.”

He noted that if productivity gains and customer adoption fail to justify today’s spending, companies may begin reducing AI capital expenditure, which will affect lower technology valuations, reduced lending, slower hiring, and broader pressure across global markets.