AI analyst topic
AI Hardware Research
Analyze the physical compute stack that enables AI training and inference, including its supply constraints and capital intensity.
Coverage
- Accelerators, CPUs, custom silicon, and semiconductor roadmaps.
- High-bandwidth memory, advanced packaging, and foundry capacity.
- Networking, optical connectivity, and system architecture.
- Semiconductor equipment and manufacturing bottlenecks.
Evidence to monitor
- Data-center revenue mix, backlog, lead times, and inventory.
- Gross margins, capital expenditure, and supply commitments.
- Performance-per-watt, software ecosystem support, and customer concentration.
Key risks
Hardware cycles can reverse quickly. Export controls, customer concentration, product delays, manufacturing yield, and shifts toward custom silicon can materially change an investment thesis.