Arvind Krishna, the CEO of IBM, highlighted that, considering the existing costs associated with constructing and operating data centers, it is virtually unfeasible for the industry to recover its substantial capital outlays. AI firms are currently ramping up their computational capacities in a bid to achieve Artificial General Intelligence (AGI). However, the exorbitant infrastructure costs pose a significant challenge to the economic viability of large-scale investments. It is projected that erecting a 1-gigawatt data center would necessitate an investment of roughly $80 billion. The cumulative global pledges for AGI-related construction projects approach 100 gigawatts, translating to an investment of about $8 trillion, with the interest expenses alone demanding approximately $800 billion in profits for coverage. Moreover, the depreciation cycle of AI chips will also exert an impact on long-term returns.

Krishna is of the opinion that the likelihood of attaining AGI with the current technology stands at a mere 0% to 1%. Several tech experts also express reservations regarding the hastened development of AGI. Despite this, Krishna acknowledged the value of present AI tools in enhancing corporate productivity. He proposed that future advancements in AGI might hinge on integrating hard knowledge systems with large models, although he approaches this strategy with caution.