Enterprises Can't Spend Their Way to AI Leadership
CIO, Thursday, September 10th, 2026
Cheaper distilled and open-weight models are undercutting massive AI compute spending; talent and culture now matter more.
The playbook of hoarding GPUs and hiring every AI expert is failing as the demand curve for frontier compute bends.
The top five tech giants are projected to spend over $750 billion on AI infrastructure in 2026, yet Chinese labs like DeepSeek and Moonshot use distillation and open-source foundations to close the gap at a fraction of the cost.
Open-weight models reaching about 90% of frontier performance on specialized tasks weaken the case for costly APIs, leaving some firms renting out excess capacity.
The author argues AI leadership depends on engineering culture, enterprise-wide adoption and retaining elite talent.