ai AM market analysis — 2026-08-09

This session’s coverage separates into two threads: academic scrutiny of foundation-model competition, and continued capital flowing into AI infrastructure. Researchers writing in the Oxford ICC journal, the University of Florida PURC working paper series, Stanford HAI and an arXiv-hosted economics paper converge on a shared framing: openness in foundation models is not a fixed stance but a strategic choice that developers revisit as competitive and policy pressures shift. One arXiv-hosted paper adds a governance caveat, arguing that subsidies designed to widen AI access can instead be captured mainly by already-dominant developers, who adjust pricing and openness to preserve their position against more efficient rivals.

On the infrastructure side, Intellectia.ai, VentureBurn and Archynewsy describe steady momentum in AI-specific data centre spending. Firmus’s newly reported Asia-Pacific funding round points to investor appetite for regional compute buildout outside the traditional US and China axis. Archynewsy frames chip efficiency, liquid cooling and alternative power sourcing, including nuclear and geothermal, as the real constraints on future hosting growth rather than headline spending totals. A Vultr social post echoes this, tying larger models and rising inference demand to changing enterprise infrastructure needs.

Taken together, the model-side research counsels caution about assuming openness or subsidy programmes will broaden competition on their own, while the infrastructure narrative shows physical buildout continuing to track model-side ambitions. This is information only, not financial advice.