‘$1.5 trillion bets’: IBM CEO sounds alarm on AI spending, says current tech has ‘0-1%’ chance of AGI

Arvind Krishna, the CEO of IBM, has raised significant concerns regarding the financial implications of the worldwide surge in investment towards artificial intelligence infrastructure. He cautioned that the push for artificial general intelligence (AGI) could lead even the most ambitious tech firms into perilously high spending. Speaking on The Decoder podcast, Krishna pointed out the enormous financial commitments being discussed, which bring into question the sustainability and potential returns of such investments. He emphasized the staggering costs associated with constructing state-of-the-art data centers, revealing that a single one-gigawatt data center could cost approximately $80 billion. For companies planning to build infrastructure of 20-30 gigawatts, the total capital expenditures could reach up to $1.5 trillion. Adding to the financial strain is the short lifespan of AI hardware. Krishna mentioned that advanced AI chips typically require replacement every five years, leading to a cycle of continual reinvestment that amplifies long-term costs. This, he warned, could severely impact company balance sheets as they strive to scale operations faster than their revenues can support. Krishna also expressed skepticism about the potential returns on such massive investments. He calculated that an $8 trillion outlay in AGI would necessitate $800 billion in profits merely to cover interest payments, casting doubt on the economic viability of current AGI goals. His comments come as OpenAI, known for developing ChatGPT, announced $1.4 trillion in infrastructure agreements with tech giants including Nvidia and Alphabet. While OpenAI's CEO Sam Altman is optimistic about strong returns on these investments, Krishna believes many current infrastructure plans are based more on hope than on solid evidence. Perhaps the most striking part of Krishna’s commentary was his estimation of AGI’s feasibility. He assigned existing technologies a mere “zero to 1%” chance of achieving true AGI, arguing that although large language models are impactful, they do not represent genuine intelligence. While acknowledging the productivity gains AI provides in business settings, he noted that these advancements do not necessarily guarantee a pathway to AGI. He suggested that any significant breakthroughs would likely require integrating language models with systems based on established knowledge, although he remains uncertain if this would suffice. Despite these cautions, major tech companies continue to ramp up their expenditures. Alphabet has revised its 2025 capital expenditure forecast to between $91 and $93 billion, while Amazon has upped its estimate to $125 billion. The industry anticipates that AI infrastructure investments will total around $380 billion this year. For the time being, Krishna urged companies to concentrate on deriving genuine productivity and business value from existing AI technologies, acknowledging that massive infrastructure investments may not yield the revolutionary advancements many are anticipating.