Physical AI's Constraint Stack Is Going Global

00Opening read

The physical-AI race is starting to split into several industrial strategies.

Huawei says demand for its AI computing hardware is outrunning supply and is responding with new chips plus architectures designed to connect enormous numbers of processors. India opened SEMICON India with a $5 billion commitment from Applied Materials as it tries to build a semiconductor ecosystem beyond packaging and incentives. D-Robotics raised $400 million to expand chips and software for embodied AI. Tower Semiconductor and NewPhotonics moved laser-integrated AI interconnect into high-volume production.

Meanwhile, the U.S. House voted 417-3 for legislation aimed at preventing data centers from shifting their incremental grid costs onto ordinary ratepayers, while Silicon Valley communities are pushing back against new facilities over power, water and pollution.

The pattern is bigger than any one announcement.

Physical AI is becoming a contest over the constraint stack: silicon, interconnect, power, factories, materials, capital and permission to build.

Solve one bottleneck and pressure moves to the next.

01What moved

Huawei scales around silicon constraints. New 2027 Ascend chips plus increasingly large compute architectures put networking and system design at the center of China's NVIDIA challenge.
Applied Materials commits $5 billion to India. India's semiconductor strategy is moving deeper into equipment, research and supply-chain localization.
D-Robotics raises $400 million. Robotics infrastructure — chips plus development software — is attracting growth capital alongside robot manufacturers themselves.
AI optics enters volume production. Tower Semiconductor and NewPhotonics are shipping 800G-1.6T laser-integrated photonic engines.
The grid bill becomes part of the compute bill. The U.S. House voted 417-3 to address who pays for new infrastructure serving large data centers.
Lucid and Bolt target 25,000 European robotaxis. Autonomous mobility is moving toward fleet-scale commitments.
China doubles down on intelligent manufacturing. Beijing is explicitly connecting AI, advanced manufacturing and supply-chain autonomy.

02The throughline

The important question in physical AI is becoming less "Who has the best technology?" and more "Who controls enough of the system to deploy it?"

03What to watch