Bloomberg Intelligence: US-China AI Gap Narrows to 6%, Zhipu Is Not an Isolated Case
0xBroomberg
Bloomberg Intelligence reports that China's AI models closed the performance gap with US peers to 6% in June, a record low; analysts say Zhipu AI's global top ranking is no fluke, and the verdict on which development path wins remains open.
A 6% gap — what does that actually mean?
The China-US AI performance gap narrowed to 6% in June, down from 9% in May and an average of 10–15% over the past year.
This means → Chinese AI is no longer trailing by a comfortable margin — it has entered a neck-and-neck zone.
In plain terms = a year ago the gap was like scoring ten-plus points behind on an exam; now it is three or four — still behind, but close enough to read the front-runner's paper.
Zhipu topped the charts — why call it "no outlier"?
Zhipu AI's GLM-5.2 ranked first globally in agentic coding — a benchmark testing an AI's ability to write and debug code autonomously.
Moonshot AI's Kimi 3 also drew wide attention; Chinese models claimed two seats on the LiveBench global leaderboard last month.
This reflects a broad-based advance, not a single-company breakout. Analyst Robert Lea called it "no outlier" — multiple Chinese firms are converging on the frontier simultaneously.
Two paths — what do they actually look like?
The US path: leverage Nvidia chips, push AI performance ceilings higher at greater cost — spend more, reach further.
The Chinese path: build smaller, more compute-efficient models, but potentially trade off peak performance — spend less, squeeze more.
In plain terms = one side drives a high-displacement sports car chasing top speed; the other drives a smaller engine tuned for fuel economy. Which reaches the finish line first is an open question.
For markets, what is the real question to track?
The report states plainly: "Which path ultimately produces the greater outcome remains to be seen."
This means → if the gap keeps narrowing, markets will have to reassess whether US AI companies' valuation premiums are still justified.
Conversely, if China's efficiency-first path proves viable, the logic that "cheaper can still deliver" would reshape capital flows across the global AI supply chain.
Content is for reference only, not financial advice.