A consequential development is emerging in the U.S.–China AI competition: China is shifting substantial investment from manufacturing increasingly sophisticated robots toward developing the intelligence that makes those robots economically productive.
Reuters reported on November 2 that Chinese embodied-AI companies attracted 93.5 billion yuan ($13 billion) across 322 financing transactions in the first half of 2026 — five times the previous year's level.
The significance extends beyond fundraising. China's industrial ecosystem is becoming a potential training environment for physical AI: factories, logistics facilities and commercial operations provide opportunities to collect real-world data and test autonomous systems. The United States retains important advantages in advanced robotics intelligence and computing infrastructure. China's opportunity lies in combining increasingly capable software with its extensive manufacturing base. This materially strengthens the industrial-diffusion dimension of our thesis, although it does not yet demonstrate a commercial breakthrough.
The New Competitive Frontier — From Tokens to Physical Tasks
three developments
Three developments stand out.
First, China's robotics investments are increasingly targeting autonomous decision-making, software and generalizable skills rather than hardware alone.
Second, industrial applications are moving beyond demonstrations. Battery manufacturer CATL reports that Spirit AI robots perform cable-connection tasks during battery-pack testing, achieving insertion success above 99%. Ant Group's Robbyant also reports deployments in two Shanghai pharmacies.
Third, Chinese developers are accumulating physical-task training data to help robots adapt across different machines and industrial environments.
These are early and company-reported results. They do not yet establish reliable, economically competitive deployment across factories. Nevertheless, the emerging direction is strategically important. China's AI challenge could extend from reducing the cost of generating intelligence to reducing the cost of applying intelligence to physical production. That would significantly enlarge the addressable economic opportunity.
South Korea — The Infrastructure Response
scale of the contest
A second development reinforces the scale of the emerging industrial competition.
On November 2, South Korean President Lee Jae Myung called for 21.3 trillion won, approximately $16 billion, in 2027 public investment across AI and three major strategic initiatives: semiconductors, physical AI and data centers. That represents nearly twice the comparable spending planned for 2026. The proposal still requires parliamentary approval.
South Korea's initiative matters because the country occupies a critical position in the advanced-memory supply chain. Its investment strategy could reinforce the computing infrastructure supporting American frontier-model development while expanding South Korea's participation in industrial AI. The competitive landscape is therefore becoming more complex than a bilateral technology contest.
The Financial Inflection — The Next Value Migration
the transmission mechanism
For The Velocity Edge, the emerging transmission mechanism is: Frontier intelligence → More efficient models → Embodied AI → Industrial deployment → Productivity gains → Value migration toward economically productive applications.
The investment implications are substantial. Advanced chips and memory remain essential, but the potential beneficiaries extend to robotics software, industrial automation, machine vision, sensors and manufacturers capable of integrating AI into their production systems.
The risk is equally significant. Large investments in robotics intelligence do not guarantee commercially viable automation. Reliability, adaptability, energy consumption, deployment costs and maintenance requirements remain important obstacles. The next test is whether physical AI can generate measurable operating-margin improvements — not simply demonstrate increasingly sophisticated capabilities.
What Comes Next — Three Decisive Tests
technology, economics, infrastructure
The technological test: Can Chinese embodied-AI models approach American frontier capabilities in generalizing across unfamiliar industrial tasks?
The economic test: Can Chinese manufacturers demonstrate lower costs per successfully completed task, including hardware depreciation, supervision and maintenance?
The infrastructure test: Can China's increasingly capable domestic semiconductor ecosystem support industrial AI deployment despite restrictions on advanced computing hardware?
The November 2 review identified no newly verified DeepSeek frontier-model release or implemented U.S.–China export-control change that would establish a reversal of American technological leadership. The most consequential new evidence concerns China's growing commitment to embodied intelligence.
The Investment Verdict
thesis intact, scope widening
Our thesis remains intact, but its scope is widening. America continues to lead important areas of frontier AI research and advanced computing. China is challenging the economics of intelligence and increasingly seeking to translate AI capabilities into industrial applications.
The emerging opportunity is not limited to software platforms or inference pricing. It encompasses the potential transformation of manufacturing productivity.
For investors and boards, the critical distinction is between financing increasingly sophisticated automation and capturing economically measurable productivity gains. The former is accelerating rapidly. The latter remains to be demonstrated at scale.
The next phase of the AI Supercycle will be determined by which system converts those capabilities into durable industrial productivity and financial returns.