On November 5, the Nasdaq fell 5.52% and the Philadelphia Semiconductor Index dropped 12.25%; the next day, the Shanghai Composite fell 1.10% and the ChiNext Index fell 3.96%. We note that AI demand remains strong, but grid connection and permitting bottlenecks at data centers are transmitting the risks of expanding compute capacity to equipment suppliers and capital markets.
What This Is
Both model training and day-to-day inference depend on data centers, while large facilities require a continuous supply of electricity, water, and network connectivity. Projects in parts of Texas are facing grid connection queues, environmental permit reviews, and community opposition: a $1.5 billion data center zoning amendment was rejected by the San Marcos City Council in a 5-to-2 vote.
The problem is not simply a shortage of power. According to the U.S. Energy Information Administration, wind and solar met 36% of electricity demand in Texas’s ERCOT region during the first nine months of 2025. But power generation, energy storage, transmission lines, data center construction, and grid dispatch still need to work in coordination. If a project opens several months late, delivery and payment cycles for servers, optical communications equipment, and server-related companies may all be pushed back at once.
Inner Mongolia is also dealing with mismatched schedules for data center and grid construction. Envision’s Galaxy Base coordinates wind power, solar power, energy storage, and data center electricity consumption under a unified dispatch system, and says 80% of its electricity consumption is supplied by its own wind farms. CICC Data’s base, meanwhile, is building a 16-kilometer transmission line to work with the public grid.
Industry View
We believe this round of volatility is not a simple “recession in AI demand.” It reflects the market beginning to recalculate the time required to deliver infrastructure. Compute orders can run through 2027, and some companies are even guiding for demand in 2028, but if projects cannot connect to the grid on schedule, the conversion of those orders into revenue and profit will still be affected.
Local reviews warrant caution, but they are not purely an obstacle. Data centers consume substantial amounts of power and water and generate noise and land-use impacts. If residents bear these costs, project approvals will be difficult to sustain. Conversely, if approval standards are opaque and review timelines unpredictable, companies may reduce investment or move projects to regions with more abundant power and less public resistance.
China’s advantage lies in engineering coordination and renewable energy supply, but it still cannot overlook grid interconnection standards, transmission construction, and cross-agency coordination. Direct green-power connections do not mean electricity is automatically available. The real test is whether power generation, storage, transmission, and data center facilities can all come online according to the same schedule.
Impact on Regular People
For enterprise IT: When procuring AI services, companies should include data center power reliability, expansion timelines, and service availability in contracts rather than comparing only model pricing and demo performance.
For individual careers: Near-term volatility may depress valuations among related hardware companies, but strong orders indicate that compute demand remains intact. Energy dispatch, operations and maintenance, and compliance roles are more worth watching, as infrastructure expansion is likely to increase demand for them.
For consumer markets: The future spread of AI services will be constrained more by electricity prices and infrastructure costs. Regions with low-cost power and strong cooling efficiency may be the first to see a denser supply of applications.