Washington’s Own Commission Just Confirmed Everything: Why the Chip War Was Aimed at the Wrong Target

A US government report explains, in cold institutional language, why export controls cannot stop China — because Beijing is winning a war Washington isn't even fighting. watchchina.ai has told you this all summer. Now America's own analysts have too

The Report That Says the Quiet Part Officially

For two months, watchchina.ai has advanced a single, consistent thesis: America is losing the AI race not because its technology is inferior, but because it is fighting the wrong war — obsessing over frontier model supremacy and chip controls while China wins through deployment, diffusion, and scale. It is one thing for an independent portal to make that argument. It is another thing entirely when the official analytical body of the United States Congress makes it in a formal report. That is what has now happened.

The report — "Two Loops: How China's Open AI Strategy Reinforces Its Industrial Dominance," authored by senior policy analyst Ngor Luong for the U.S.-China Economic and Security Review Commission — argues something remarkable for a US government document: that China's open-source AI strategy is not a second-best adaptation to semiconductor export controls, but a coherent doctrine for winning AI competition through systemic deployment rather than frontier model supremacy. In plain terms, Washington's own commission is telling Washington that its entire strategic assumption is wrong. China is not scrambling to catch up on the frontier. It is running a different race, on purpose, and winning it.

The Two Loops — And the One Washington Can't Touch

The report's central insight is an elegant piece of analysis that watchchina.ai's readers will recognize as the connective tissue running beneath dozens of our briefings. China's approach operates through two compounding feedback loops — one digital, one physical. The digital loop is open model diffusion driving community iteration: the free GLM, DeepSeek, Qwen, and Kimi releases that spread Chinese AI across the world and generate 100,000 derivative models. The physical loop is industrial deployment generating proprietary operational data: the factories, logistics networks, and robots — including the citizen-filmed chore data watchchina.ai documented — that produce real-world data no American lab can access.

Here is the devastating conclusion, in the report's own framing: it is the intersection of these two loops that gives China's AI strategy its compounding force, and controls calibrated to only one loop could leave the more durable source of advantage unaddressed. US export controls primarily target the digital loop — restricting access to advanced chips used for frontier model training — but are not well suited to addressing the physical loop of deployment-driven data creation across China's manufacturing base. This is the entire watchchina.ai thesis, rendered in the sober institutional language of a congressional commission: America built its whole containment strategy around chips, and chips address only half of China's advantage — and not the durable half.

Deployment Beats Benchmarks — Made Official

The report crystallizes into a single proposition that could serve as watchchina.ai's motto for the entire summer: China has operationalized a doctrine in which AI leadership is won not through the most capable model but through the widest deployment at the lowest cost. And it identifies the structural shift that makes this decisive — the emerging move toward small, specialized AI models for real-world deployment, rather than ever-larger frontier LLMs. As the competitive center of gravity shifts from large-scale pre-training toward small, specialized deployed models, China's dominance of the global open ecosystem and its unmatched manufacturing base may confer durable strategic advantages in manufacturing, logistics, robotics, and scientific research — independent of whether any Chinese frontier model outperforms its US counterpart.

watchchina.ai will, as always, hold the honest complexity. The report itself notes that the United States still leads in consumer-facing AI applications — ChatGPT's monthly traffic exceeds DeepSeek's by a factor of twelve — and that America retains genuine frontier advantages. This is not a story of Chinese total victory, and anyone selling that is selling propaganda. But the report's warning is precisely that this visible American lead at the consumer and frontier layers may mask a deeper structural risk, as Chinese AI companies shift their center of gravity from consumer products to industrial integration. And the most damning line of all, for US strategy: there is no coordinated U.S. response to China's open ecosystem dominance. Two months of watchchina.ai briefings — the "good enough" strategy, the 140 trillion tokens, the WAICO alliance, the Asian market surrender, the sanctions paradox written in silicon — all of it now finds its confirmation in a document produced not by Beijing, and not by this portal, but by the United States Congress's own China commission. When your adversary's own government analysts are telling their leaders that the strategy is aimed at the wrong target, the debate about who is winning is, for practical purposes, over. What remains is whether Washington will listen to its own experts in time. On the current evidence, watchchina.ai would not bet on it.

Sources: U.S.-China Economic and Security Review Commission ("Two Loops," Ngor Luong, March 23, 2026), TechTimes, Ruibao News, Epoch AI, Artificial Analysis