The scale of global investment in artificial intelligence infrastructure is drawing increasing scrutiny after IBM CEO Arvind Krishna warned that the multi-trillion-dollar AI data center buildout could be a bubble. The remarks have since been amplified by Zoho Chief Scientist Sridhar Vembu, who said the company would focus on areas such as data curation, reinforcement learning, and AI verification technologies rather than joining what he described as an investment bubble.

Sharing Krishna's views on X, Vembu said, "We are investing in creating capabilities like data curation, reinforcement learning, and most crucially the compiler infrastructure to ensure AI output can be verified but we will not chase the investment bubble." He added that while some may view the strategy as conservative, the company would revisit the debate in five years.

The discussion gained momentum after cloud computing expert and author David Linthicum published a post titled "The Emperor Has No Clothes: Why the AI Infrastructure Buildout Math Doesn't Work." In the post, Linthicum argued that the current pace of AI infrastructure spending is based on demand projections that may not materialize.

According to Linthicum, companies across the industry are committing hundreds of gigawatts of data center capacity over the next decade, with total capital expenditure running into trillions of dollars. However, he believes enterprise adoption remains slower than expected, while consumer AI products are still searching for sustainable business models.

"The math just doesn't add up," Linthicum wrote, comparing the current enthusiasm around AI infrastructure to previous technology bubbles such as the dot-com boom, blockchain, and the metaverse.

As cited by Futurism, Krishna reportedly estimated that a modern 1-gigawatt AI data center can cost around $80 billion to build. Under that framework, a company committing to 20–30 gigawatts of capacity could spend approximately $1.5 trillion. Across the industry, planned AI infrastructure investments linked to AGI ambitions could approach 100 gigawatts, translating to roughly $8 trillion in capital expenditure.

Krishna argued that such investments would require around $800 billion in profits simply to cover financing costs, raising questions about whether future AI demand can justify the spending.

Not everyone agrees with the cautious outlook. Some industry observers argue that AI adoption is accelerating rapidly and that current power users already spend several hours each day using AI tools. Others believe compute infrastructure, energy availability, and access to advanced foundation models will become strategic assets in the coming decade.

The debate highlights a growing divide within the technology industry. While some companies are racing to build AI infrastructure at unprecedented scale, others are prioritising software, data quality, model reliability, and verification systems, betting that long-term value may come from practical deployment rather than raw computing capacity alone.