While focus is on massive supercomputers for training next-gen models, the real supply chain constraint will be 'inference' chips—the GPUs needed to run models for billions of users. As adoption goes mainstream, demand for everyday AI use will far outstrip the supply of available hardware.

The shift from simple chatbots (one user request, one API call) to agentic AI systems will decouple inference requests from direct user actions. A single user request could trigger hundreds or thousands of automated model calls, leading to an exponential increase in compute demand and cost.

The focus in AI has evolved from rapid software capability gains to the physical constraints of its adoption. The demand for compute power is expected to significantly outstrip supply, making infrastructure—not algorithms—the defining bottleneck for future growth.

A critical, under-discussed constraint on Chinese AI progress is the compute bottleneck caused by inference. Their massive user base consumes available GPU capacity serving requests, leaving little compute for the R&D and training needed to innovate and improve their models.

The current focus on building massive, centralized AI training clusters represents the 'mainframe' era of AI. The next three years will see a shift toward a distributed model, similar to computing's move from mainframes to PCs. This involves pushing smaller, efficient inference models out to a wide array of devices.

While the growth of new consumer AI users is slowing into an S-curve, the compute consumption per user is still growing exponentially. This is driven by the shift from simple queries to complex, token-intensive tasks like reasoning and agents, sustaining massive demand for GPU infrastructure.

While AI training requires massive, centralized data centers, the growth of inference workloads is creating a need for a new architecture. This involves smaller (e.g., 5 megawatt), decentralized clusters located closer to users to reduce latency. This shift impacts everything from data center design to the software required to manage these distributed fleets.

While training has been the focus, user experience and revenue happen at inference. OpenAI's massive deal with chip startup Cerebrus is for faster inference, showing that response time is a critical competitive vector that determines if AI becomes utility infrastructure or remains a novelty.

AI's computational needs are not just from initial training. They compound exponentially due to post-training (reinforcement learning) and inference (multi-step reasoning), creating a much larger demand profile than previously understood and driving a billion-X increase in compute.