ai PM market analysis

The PM session presents a structural tension at the centre of the AI investment thesis. Gartner projects global AI spending approaching one and a half trillion dollars in 2025, growing close to fifty percent year over year per CIO Dive. Yet Epoch AI research, cited by The AI Insider, finds that inference prices required to reach a given model performance milestone have fallen between nine and nine hundred times per year. That deflation rate fundamentally challenges the revenue models being built on top of the infrastructure that the spending surge is funding.

Platform competition is intensifying in parallel. AWS is positioning Amazon Bedrock as the enterprise production layer for generative AI agents, while CoreWeave is carving out a specialist position with purpose-built rack-scale compute for frontier AI workloads. Both are targeting the same enterprise AI buying decision from different angles — scale versus specialisation.

The most strategically significant development today is NVIDIA and Hugging Face’s joint announcement of an open-source AI foundation model for robotics, reported by BigGo Finance. An open-source foundation model co-developed by the dominant AI chip maker and the leading model-sharing platform could meaningfully lower barriers to physical AI adoption — and in doing so, extend the demand frontier for the compute infrastructure both companies supply.

Worth Tracking

Inference cost deflation vs. usage growthEpoch AI data shows inference prices falling at an extraordinary rate; whether usage growth can outpace price compression fast enough to sustain current valuation assumptions is the central AI monetisation question.

NVIDIA-Hugging Face open-source robotics modelA jointly developed open-source foundation model for robotics could significantly accelerate physical AI adoption by lowering access barriers, while positioning both companies as infrastructure gatekeepers to the next application layer.

Amazon Bedrock enterprise adoption vs. Microsoft and GoogleAWS is pushing Bedrock as the production platform for enterprise AI agents; how quickly it captures that layer relative to Microsoft and Google will determine whether AWS translates its cloud position into durable AI application revenue.