The United States added 8.3 gigawatts (GW) of utility-scale battery storage in the first half of 2026, reaching nearly 52 GW. In August, the U.S. Department of Energy (DOE) also announced a $100 billion Paducah AI energy hub with plans for two GW of new natural-gas generation, up to 2.6 GW of battery storage, and transmission upgrades. The project is expected to help lower regional electricity costs, create thousands of high-paying American jobs, and strengthen national security.
U.S.–China AI Leadership Scorecard
A monthly read on who is winning the race for artificial intelligence — scored across the four long-term drivers that decide it: energy, compute, adoption and talent.
AI leadership rests on four long-term drivers.
Each month this scorecard highlights the most important developments affecting each country’s AI competitiveness — then scores the month, driver by driver. Advances and setbacks are graded per country: green moves that nation forward, red holds it back.
Training and running frontier models consumes enormous power. The nation that builds abundant, affordable and reliable electricity fastest can field the largest AI systems.
Advanced semiconductors, frontier models and data centers are the raw horsepower of AI. Leadership across the compute stack sets the pace for everyone else.
Capability only counts once it is deployed — across businesses, government and allied nations. Widespread use, plus exporting the AI stack abroad, converts research into real advantage.
AI is built and sustained by people: researchers, engineers and the skilled tradespeople who construct and power data centers. A deep talent pipeline protects leadership over the long run.
China made more net progress toward AI leadership in August — its massive electricity buildout continued, while 546 million generative-AI users and the global spread of Alibaba’s Qwen strengthened its position in adoption. The United States kept its edge in frontier chips and compute investment and invested heavily in the skilled workforce behind AI infrastructure, but 299 active data-center restrictions across 44 states threaten its ability to turn that advantage into operating infrastructure.
In August, DOE issued three emergency orders to keep power generation available during periods of high electricity demand. The orders covered Pennsylvania’s Eddystone and Maryland’s Wagner plants in the Pennsylvania-New Jersey-Maryland Interconnection (PJM) grid, which serves 65 million Americans, and Michigan’s J.H. Campbell plant in the Midcontinent Independent System Operator (MISO) region.
China’s installed generating capacity reached 4.08 TW at the end of July, up 11 percent year over year. Solar capacity reached 1.29 TW, up 16.1 percent, while wind reached 690 GW, up 19.5 percent.
In August, China announced a new national power plan that calls for integrating more than 2.8 TW of new-energy capacity into its electric grid by 2030 and generating 50 percent of its electricity from non-fossil sources. The plan also calls for a larger, more resilient grid capable of supporting China’s rapidly growing electricity demand.
A new report found North American data center demand reached a record 25 GW in the first half of 2026, double the previous year and five times the level two years earlier. That demand is translating into massive new projects: in August, OpenAI announced plans for eight GW of computing capacity at a new Ohio AI data center campus. Data center vacancy rates will remain near zero through 2028.
Interconnected Capital’s tracker showed 299 currently active data center moratoriums and other restrictions across 44 states, creating a growing obstacle to converting America’s chip advantage into operating compute.
A new report found that 79 percent of Americans support U.S. leadership in AI, but only 14 percent support data center development in their own community – a 65-point gap that threatens America’s ability to build the infrastructure needed to maintain AI leadership.
In August, China’s Envision began operating the first phase of a planned two GW AI data center campus, larger than many of the biggest individual AI campuses being built in the United States. DeepSeek is also planning a one GW campus nearby, while RedNote is considering another 600 MW facility.
Alibaba reported 45 percent growth in its AI cloud and computing business and said its Chinese-designed AI chips are now being used by more than 650 customers across 20-plus industries.
New data published in August found 39.2 percent of employed Americans use generative AI for work, up from 28.2 percent in 2024. AI now assists with 6.3 percent of work hours, showing how quickly the technology is becoming part of everyday work.
Companies making the largest investments in AI increased employment 10.2 percent in the two years following adoption, with entry-level employment growing 12 percent, evidence that greater AI adoption is creating job growth, not simply replacing workers.
Qwen recorded roughly 2.05 billion downloads on Hugging Face, with developers creating roughly 200 new AI models every day based on Qwen, helping spread Chinese AI technology into more applications and markets. Alibaba reports even broader adoption across platforms, with Qwen surpassing three billion downloads and 300,000 Qwen-based models worldwide.
An August report estimates 545.6 million people in China will use generative AI each month in 2026, making China the world’s largest national market for generative AI and accounting for nearly one-quarter of the world’s 2.39 billion GenAI users.
Meta partnered with North America’s Building Trades Unions (NABTU), tapping a network of 3.2 million skilled workers and 1,900 training facilities to help build AI infrastructure. Meta also invested $115 million in its first year of America’s Workforce Academy, which provides industry-recognized credentials and guaranteed jobs for graduates, with its first class now heading to Meta
The new Boise facility provides hands-on semiconductor training while Micron expands its registered apprenticeship program and advanced-manufacturing workforce pipeline.
China is moving to integrate AI throughout its education system, including expanding AI education in primary and secondary schools and encouraging interdisciplinary teaching, as it builds a larger pipeline of students with AI skills.
China made more net progress toward AI leadership in July — launching a 29-nation AI governance bloc, closing the compute gap with competitive open-weight models and a new domestic supercomputer. The United States advanced on energy policy, federal siting and skilled-trades funding, but state and local restrictions kept slowing data-center development.
New executive action requires data-center operators — not ratepayers — to fund the power and infrastructure their AI projects need.
Consumption is projected to climb from roughly 183 to 426 terawatt-hours, with new generation not on pace. PJM, the nation’s largest grid, keeps missing supply targets.
Installed generation capacity hit four billion kilowatts (+10.8% year over year) — nearly double its own demand growth.
The IEA finds China’s 2026 electricity demand is rising three times faster than that of the United States, and supply is meeting it.
Local opposition continues to slow AI-infrastructure buildout nationwide, with some states moving to freeze — or considering freezing — new projects.
Faster siting on federal land — former industrial areas, power-generation facilities and dormant sites — aims to unstick permitting bottlenecks.
CHIPS Act funding for next-generation chip packaging, AI memory and more, aimed at cutting U.S. reliance on foreign suppliers for critical compute.
A coordinated national effort treats domestic AI-chip independence as a mandate, not a business decision.
A coordinated national effort treats domestic AI-chip independence as a mandate, not a business decision.
Multiple Chinese chipmakers showcased large-scale AI systems built entirely on domestic chips, underscoring Beijing’s push to replace U.S. technology across the stack.
Washington countered Beijing’s bloc by rallying 35 partner nations behind its own vision for open, democratic AI adoption.
steep jump from 55 percent a year ago as AI spreads across the broader U.S. economy.
A new Shanghai-based body works to expand Beijing’s influence over global AI standards and adoption, especially across the Global South.
A cross-agency initiative blends federal compute, data and research infrastructure with AI to tackle grid planning, microelectronics and manufacturing.
Meta, Google and BlackRock fund electrician and carpenter pipelines for the data-center boom.
A new policy push channels a record graduating class toward AI-related occupations and continuous learning.