AI infrastructure spending forecast to vastly exceed revenue

Summarised from 3 outlets · 4 reports · Updated 15 Oct, 14:20 · Archive

Analysis suggests AI infrastructure investment will far exceed expected returns unless the technology achieves unprecedented economic gains.

Since ChatGPT's release in late 2022, AI developers have invested heavily in data centre buildouts, chip development, and power grid expansion. Analysis by Marin Ivezic, Founder of Applied Quantum, finds the financial projections unsustainable. The $2.5 trillion invested comprises $1.5 trillion already spent on hardware since 2022 and $1 trillion in additional annual spending. Current AI revenues are estimated at between $150 billion and $250 billion per year.

To justify this investment, AI would need to generate $2.5 trillion in annual revenue. Ivezic calculates this would require the technology to deliver productivity gains equivalent to 250 million full-time jobs, assuming each costs $100,000. This exceeds the United States' total nonfarm workforce of 159 million.

Infrastructure providers face an additional challenge: high-performance AI chips become obsolete every five years, but the data centres housing them are designed to last twenty years, creating significant depreciation costs. Ivezic describes this mismatch as the '50-Cent Rule', requiring providers to generate 50 cents in annual revenue for every dollar invested. A Register analyst predicts that eighty per cent of companies providing AI infrastructure will disappear as standalone businesses by 2030.

How it is being reported

• Investors dump data centre bonds as AI borrowers face ‘Darwin moment’Debt issued by AI infrastructure providers and neocloud companies has sunk towards stressed territoryFinancial Times · 15 Oct, 14:26
• AI has a $2.5 trillion problem and replacing 250 million jobs is the only way to solve itInfrastructure providers could be left bankrupt unless AI delivers a return on investment far higher than current rates.TechRadar · 15 Oct, 14:05
• AI math isn't mathing: infrastructure spending forecast to outstrip expected revenueAnalyst predicts 80 percent of neoclouds will disappear as standalone businesses by 2030The Register · 14 Oct, 10:00
• The AI math isn't mathing as infrastructure spending forecast to outstrip expected revenueAnalyst predicts 80% of neoclouds will disappear as standalone businesses by 2030The Register · 14 Oct, 10:00

In this story: Marin Ivezic · Applied Quantum · ChatGPT · United States

This summary was written by AI from the headlines and standfirsts above, and states as fact only what at least two outlets report. How we use AI · Report a problem