An aggressive AI investor warned that speculative data‑center builds could face a financing crisis in 2027–28, arguing that capital is flowing to "landlord" operators rather than owner‑users such as hyperscalers. The investor flagged extreme capex, weak enterprise traction, and risky valuation models for third‑party data‑center developers. Separately, Nvidia licensed Groq’s inference technology and hired the startup’s team in a landmark transaction, signaling that even dominant GPU vendors are hedging on inference economics. University research computing directors and CIOs should note both signals: procurement and partnerships for AI compute may face price volatility and shifting vendor consolidation as inference optimization becomes central to scale deployment.