IBM CEO Challenges AI Bubble: Is Trillion-Dollar Race for AGI Economically Viable?

IBM CEO Challenges AI Bubble: Is Trillion-Dollar Race for AGI Economically Viable?​

IBM Chairman and Chief Executive Officer Arvind Krishna has injected a note of caution into the frenetic pace of artificial intelligence infrastructure expansion. He directly questioned whether the monumental scale of capital commitments being made by major tech players can realistically yield returns given current assumptions about Artificial General Intelligence (AGI).

The Economics of Gigawatt Spending​

Krishna laid out stark figures regarding the cost structure of advanced AI computing. He stated that it costs approximately $80 billion to fill a single gigawatt data centre at today's price points. Scaling this up, he noted that committing 20 to 30 gigawatts of capacity would equate to around $1.5 trillion in capital expenditure.

The scale of industry commitment is vast, Krishna noted. Announced commitments across the sector aimed at AGI appeared to amount to roughly 100 gigawatts. Based on current costs, he estimated this would translate into an astounding $8 trillion in capital expenditure for the technology.

Unanswered Questions Over Return on Investment​

A critical view was presented by IBM's leadership regarding the profitability of this colossal investment. "It’s my view that there’s no way you’re going to get a return on that," Krishna stated emphatically. He further elaborated that $8 trillion in capital expenditure would require about $800 billion in profit just to cover interest payments alone.

This commentary comes as the AI race intensifies, with companies like OpenAI continuing to secure massive infrastructure and chip agreements. These commitments underscore the high stakes attached to developing models capable of AGI, which OpenAI itself defines as highly autonomous systems surpassing human ability in economically valuable work.

Market Bets and Corporate Commitments​

The expenditure plans are backed by significant industry partnerships. In September 2025, OpenAI and Nvidia announced a strategic partnership involving at least 10 gigawatts of Nvidia systems for next-generation AI infrastructure. Nvidia intends to invest up to $100 billion in OpenAI as each gigawatt is deployed.

Further commitments include AMD; in October 2025, AMD and OpenAI signed an agreement where OpenAI would deploy six gigawatts of AMD GPUs, starting with a one-gigawatt deployment planned for the second half of 2026. Meanwhile, reports citing a Wall Street Journal piece placed OpenAI’s data centre commitments with Oracle at roughly $300 billion over five years.

Industry Concerns and Corporate Strategy​

While Krishna questioned the viability of the AGI path, he did not dismiss generative AI's immediate utility. He affirmed that the technology is "incredibly useful for enterprise" and capable of unlocking significant productivity gains. His critique, however, focused narrowly on the low probability of current technologies achieving true AGI.

IBM’s internal strategy remains geared towards enterprise adoption through products like watsonx and hybrid cloud services. An earlier CEO study conducted by IBM showed that while business leaders expected AI investment growth to more than double over two years, only 25 percent of these initiatives had achieved the expected return on investment.

The Wider Energy and Hiring Debate​

The heightened focus on computational power is reflected in global energy projections. The International Energy Agency projects that global electricity consumption for data centres will double by 2030 to around 945 terawatt-hours, driven largely by AI adoption and accelerated servers.

The debate over whether the industry constitutes a speculative bubble has permeated IBM’s internal hiring practices as well. One reporting detail noted that IBM leadership asked job candidates if they believed the industry was in an AI bubble, seeking responses that demonstrated comprehensive thought on the issue.