IBM CEO: The AI industry is a "hard-to-break-even" gamble, with only a 1% chance that LLMs will successfully build AGI

The AI industry is experiencing an unprecedented capital frenzy, but IBM CEO Arvind Krishna has poured cold water on it with elementary-level calculations. He pointed out that global capital expenditures in AI data centers are approaching $8 trillion, but with the current business models, it’s simply impossible to break even. Especially since he believes the success rate for current large language models (LLMs) to build AGI is only 1%, making monetization difficult.

The “$8 Trillion Bet” on AI Data Centers: Why It Will Never Break Even

On the Decoder podcast, Krishna made a simple estimate based on current costs.

Building a 1 GW AI data center requires about $8 billion. If large tech companies are targeting a scale of 20 to 30 GW, each company would need to invest at least $1.5 trillion. Looking at the entire industry, the AI race is driving about 100 GW of global computing capacity, which is equivalent to $8 trillion in capital expenditures:

Krishna stated bluntly: “I don’t think you can get a return on that.”

To break even on an $8 trillion investment, you would need about $800 billion in annual profits, and currently, no AI company’s business model can support investments of this scale.

He added that, since GPUs need to be replaced every five years, depreciation costs make AI infrastructure even heavier.

(The Big Short’s Michael Burry Criticizes AI Giants Again: Underestimating Depreciation and Inflating Earnings Is Modern Fraud)

Microsoft and OpenAI’s High-Risk Bet on AGI: Only 1% Chance of Success

When host Nilay Patel mentioned that OpenAI seems convinced its massive investments will pay off, Krishna responded: “That’s a belief (belief), I understand it but don’t agree.”

I think that’s fine—some will fail but leave behind useful infrastructure; but once some succeed, they’ll be able to dominate the market.

Krishna pointed out that OpenAI and its backer Microsoft are actually betting on the premise that “AGI will inevitably arrive, and it will be built by them.” However, he does not think this assumption holds up:

I believe the probability that current LLM technology will achieve AGI is only 1%. AGI must rely on brand-new, groundbreaking technological breakthroughs, not just scaling up models. The key is how to combine it with thousands of years of human knowledge.

(Balaji’s Five-Point Analysis on the Future of AI: It Won’t Replace Human Jobs, and There Won’t Be a Single Omnipotent AGI)

The AI Bubble Debate Returns, Krishna Calls the Hype Reasonable

As tech giants like OpenAI, Meta, and Google double down on AI, concerns about a bubble are emerging in the market.

Krishna directly said: “I don’t think we’re facing an AI bubble, but it’s definitely like the internet race of 2000—some capital investments will end up losing money.”

This AI race is pursuing internet-scale success, and the last generation of social media is the best proof, so I’d say the current hype is reasonable.

(AI Crypto Czar: OpenAI Mistakenly Seen as Start of AI Bubble, Tech Stock Decline Is a Natural Correction)

How Many Jobs Will AI Replace? Krishna: Limited Impact, But Will Reshape the Talent Structure

While many fear that AI’s advance will trigger massive layoffs, Krishna takes a milder view:

I believe this is very likely to happen in the coming years, but it won’t be 30% or 40%. At most, it will affect 10% of the workforce, and it will be highly concentrated in certain fields.

He believes companies should invest in helping new workers use AI to boost productivity and efficiency, rather than laying them off: “Sometimes amplifying your workforce is more cost-effective than reducing it.”

The Next Chapter for AI: Huge Gamble or Rational Investment?

Amid the capital-driven AI frenzy, the IBM CEO offers a clear perspective: “AI will change corporate structures, and today’s AI infrastructure bets are fraught with both financial and technological risks—especially as the road to AGI remains unclear.”

This article, “IBM CEO: The AI Industry Is a ‘Hard-to-Break-Even’ Bet, Only 1% Chance for LLMs to Build AGI,” first appeared on Chainnews ABMedia.