Zoho founder Sridhar Vembu pushes a pragmatic path for AI, focusing on core AI capabilities rather than chasing an investment frenzy amid the big data-center buildout.

The debate centers on AI infrastructure spend versus real demand, with Arvind Krishna arguing for profitability benchmarks and Linthicum warning of asset write-downs if demand never materializes.

Krishna estimates that building 1 gigawatt of data centers could cost about $80 billion, implying that 20–30 gigawatts might require around $1.5 trillion in capital expenditure.

Industry-wide questions linger about AI economics, productivity gains, and the risks of job cuts amid rapid AI-driven transformation.

The narrative frames a tension between ambitious AI infrastructure expansion and uncertain, potentially mismatched demand and profitability.

IBM emphasizes breakthroughs beyond current LLMs, such as fusing knowledge with models, while remaining unsure of whether these paths lead to AGI.

Industry opinions are split: some see rapid AI adoption and compute infrastructure as strategically valuable, while others prioritize software, data quality, and verification over massive compute.

Some argue the AI investment boom may be temporary and that AI alone cannot fix broader economic imbalances or guarantee productivity, reflecting rising industry skepticism.

The sector backdrop includes OpenAI–Nvidia–AMD collaborations and Oracle cloud commitments driving multi-gigawatt plans, drawing investor scrutiny on business models.

Linthicum cautions about mismatches between enterprise adoption pace, consumer AI maturity, and the scale of investment, warning of risk if demand underwhelms.

He notes slow enterprise adoption and evolving consumer AI, raising the prospect of write-downs or pivots if demand fails to materialize.

Krishna concedes generative AI can boost enterprise productivity even if returns on AGI-scale bets remain uncertain.