Big Tech’s AI Datacenter Investments Might Be In Big Trouble
NOBODY TALKING: Everyone is racing to build mega datacenters. Nobody is talking about what happens if Chinese open models make half of that capex worthless.
Why it matters
Chinese frontier models achieving parity with US closed models at significantly lower cost threatens the ROI calculus underpinning billions in hyperscaler datacenter capex. This forces a strategic reckoning: either the efficiency gains stick (reshaping AI infrastructure economics), or margin compression accelerates industry consolidation.
The key facts
5 to knowChinese open models (GLM-5.2, DeepSeek-V4) now rival frontier AI capability
Cost advantage claimed as 'fraction of the cost' vs US frontier models
Implies potential stranding of hyperscaler datacenter investments
Challenges the assumption that larger compute spend = sustainable competitive moat
Published Jun 2026 — positions this as emerging strategic threat, not hypothetical
Go to the source
Forbes Innovationforbes.com
Publisher excerpt: Chinese open models like GLM-5.2 and DeepSeek-V4 now rival frontier AI at a fraction of the cost, and that could strand the data center bet hyperscalers made.