ai PM market analysis — 2026-09-12

The AI market is shifting from a race over model releases toward a harder question: whether the underlying economics can hold. Research on open foundation models flags that long-run commercial payoffs remain poorly understood as providers keep competing dynamically, which raises doubt about how much genuine openness the market can sustain.

Stanford’s work on responsible open models frames this as a balancing act between broader access and the governance, safety and accountability demands that come with it. Bruegel’s analysis suggests competition is still active across both open and closed approaches, with market structure and access conditions shaping the policy debate rather than settling it.

Enterprise inference is emerging as a steadier driver of infrastructure demand, as deployed systems move from pilot projects into routine business workloads. That points to spending gradually reorienting from headline model training toward serving capacity and operational efficiency.

Taken together, the evidence points to a market in transition rather than one moving cleanly in a single direction. Durable monetisation for open models is unresolved, governance frameworks are still forming, and infrastructure investment is only beginning to reflect inference-led demand.