China’s official internet statistics body has put a number on Chinese AI investment in H1 2026, and next to the US figures the gap is roughly tenfold. For search people, the bigger story is where the money went. In China it mostly went into robots and hardware. In the US, more than half went to two model labs.
China: RMB250 billion, and robots took the biggest slice
The China Internet Network Information Center (CNNIC) released its Generative AI Application Development Report (2026) at the China Internet Basic Resources Conference on 29 September (CLS and Sina Finance covered it the same day). The headline financing figures for H1 2026 are:
- 1,255 AI-related investment and financing events, already 78.0% of the 2025 full-year count.
- About RMB250.14 billion in total, already 182.0% of the whole of 2025.
The PDF breaks that money down by sector:
- Robotics / embodied AI: RMB93.78 billion, or 37.5% of capital. It also accounted for 529 deals, 42.2% of the count.
- AI hardware and technology: RMB88.85 billion, or 35.5% of capital, from just 21.5% of deals.
- AI applications themselves: RMB44.18 billion, or 17.7%.
- AI in traditional industries: 7.0%. Autonomous driving: 2.3%.
Put the two application buckets together and they took about a quarter of China’s AI money. Beijing led regionally with 34.9% of capital. A footnote says CNNIC built the dataset from publicly disclosed financing events, excluding private placements and IPOs, and that some of the figures are estimates.
At the 30 June central parity rate of 6.8109 yuan to the dollar, RMB250.14 billion is roughly $36.7 billion.
US: $412.7 billion of venture money, 86% of it AI
The Q2 2026 PitchBook-NVCA Venture Monitor puts US venture deal value at $412.7 billion in H1 2026. It says AI took 86% of all venture dollars, which works out to roughly $355 billion. Deals of $100 million or more took 87.5% of the money.
PitchBook’s separate Q2 2026 AI Report counts $407 billion of AI venture funding across 3,500 deals in H1, against $264.1 billion in all of 2025. PitchBook’s own write-up adds the concentration figures:
- OpenAI and Anthropic raised about $217 billion between them, more than half the total, from three rounds: OpenAI’s $122 billion in March, and Anthropic’s $30 billion Series G (February) and $65 billion Series H (May).
- Horizontal platforms, meaning the frontier labs plus companies such as ElevenLabs and World Labs, took 70.8% of capital ($288.1 billion).
- Vertical applications made up 62.9% of deals but got only 12.9% of the money.
The US is funding robots too: PitchBook says autonomous-machine funding already more than doubled 2025’s full-year total. It is still a side story next to the labs.
These datasets aren’t like for like: CNNIC counts disclosed “investment and financing events”, PitchBook counts venture deals. Read “about a tenth” as directional.
Why search people should care
The US money is a bet on the model layer, the labs whose assistants are becoming the answer box. CNNIC’s Chinese money is mostly a bet on robots and hardware. That isn’t because Chinese assistants lack users. The same report says China’s generative AI users passed 700 million, more than half the population, and 76.0% of users ask it to answer questions. That is search behaviour.
The usage table in the report, which uses QuestMobile data for July 2026, shows the shift happening:
- AI general assistants: about 43.5 billion monthly uses, up 162.8% year on year.
- AI search engines: about 38.5 billion, up just 9.9%.
In China, growth is going to the all-in-one assistants (Doubao, Qwen, DeepSeek and friends) rather than to the standalone AI search engines where Baidu’s AI search competes. Most of the money behind those assistants doesn’t show up as startup venture rounds. It comes from the platforms’ own budgets. CNNIC notes that Alibaba wired Qwen into Taobao, Alipay, Fliggy, Amap and Damai in February, and that Doubao launched paid tiers in June at RMB68, 200 and 500 a month. See our coverage of Qwen Go! and the 151 apps opting out of Doubao’s phone agent.
The exception is DeepSeek. CNBC reports that its first external round, about RMB50 billion, closed in June. If CNNIC counted it, that one deal alone would be a fifth of China’s half-year total. CNNIC doesn’t list deals, so we can’t confirm it. DeepSeek’s second round of up to RMB100 billion lands in H2. It is already spending on search: it just opened its in-house search engine to developers.
Our take
Don’t read a tenfold funding gap as a tenfold quality gap in Chinese AI search. CNNIC also says the top six models on global model-call leaderboards come from Chinese teams. Chinese startup capital is heading for the physical world, while the assistants are funded inside super-apps. US capital sits in two labs that now need ads, subscriptions and search revenue to justify $217 billion.
For SEOs, that means two very different worlds to optimise for. In the US, visibility depends on a handful of well-funded assistants. In China, it depends on whichever assistant owns the app your customer already has open, and those apps increasingly lean on Qwen, Baidu and DeepSeek, not US models.