IBM CEO Arvind Krishna warned on The Decoder podcast that the technology industry’s race to build artificial intelligence infrastructure may lack financial sustainability due to enormous costs estimated at trillions of dollars.
Krishna calculated that equipping a single one-gigawatt data center requires approximately $80 billion at current prices. Companies targeting 20-30 gigawatts of infrastructure would incur capital expenditures around $1.5 trillion. He specified that these figures constitute “today’s number,” noting costs could shift as the industry expands.
Krishna pointed out the ongoing financial strain from hardware depreciation. Advanced AI chips demand replacement every five years, establishing a perpetual reinvestment cycle. He stated, “If I look at the total commits in the world in this space, in chasing AGI, it seems to be like 100 gigawatts.” This global pursuit of artificial general intelligence equates to roughly $8 trillion in capital expenditures at $80 billion per gigawatt. Servicing the interest on this investment would necessitate about $800 billion in annual profit.
Krishna rated the prospects of current large language model technologies achieving artificial general intelligence at a “zero to 1 %” chance absent additional breakthroughs. He recognized that artificial intelligence holds capacity to unlock trillions of dollars of productivity in the enterprise. True artificial general intelligence, however, would demand integration of language models with “hard knowledge” systems. Krishna expressed doubt on the success of this combination.
