ai PM market analysis — 2026-08-25

The AI market is broadening beyond a straight contest between frontier model developers. Funding rounds, chip-software tie-ups and acquisition talk across the stack point to a structure that is consolidating around a smaller set of strategically important layers, not just the models themselves.

Model development continues to prioritise reasoning, multimodality and cheaper inference, which raises the bar for both providers and the infrastructure operators supporting them. Managed cloud is being floated as a further picks-and-shovels play on AI demand, though slowing cloud growth, heavy capital requirements and unproven margin expansion temper that case.

Enterprise behaviour is a useful counterweight to the infrastructure story. Thomson Reuters continuing to run Anthropic’s models alongside its own shows large buyers hedging rather than committing to a single proprietary stack, and the growing accessibility of open multimodal models through routing marketplaces adds further pressure on closed providers’ pricing power.

Taken together, the picture is one of intensifying competition and capital intensity rather than a settled hierarchy. The infrastructure spending case and the enterprise adoption case are pulling in different directions, which argues for a measured read on where value in the stack actually accrues.