The hype cycle loved the chatbots. The money is now quietly flowing to the chips that make robots move. On 17 September, a company most Western readers have never heard of raised US$400 million — and explained where China's AI capital really thinks the value is.
D-Robotics (地瓜机器人) is the robotics-computing arm spun out of autonomous-driving chipmaker Horizon Robotics (地平线) in early 2024. It does not build robot bodies. It sells the "picks and shovels" — the Sunrise (旭日) series chips, developer kits and software that sit inside humanoid (人形机器人), quadruped and logistics robots. In the words of one Chinese report, it wants to be the "Nvidia of the embodied-AI era." This month, investors agreed that the shovels are worth backing.
The round
D-Robotics closed a US$400 million Series C (about RMB 2.9 billion / HK$3.1 billion) on 17 September. The round was led by Mirae Asset, with a strategic investment from Meituan and participation from a long list of funds: Hefei State-owned Investment, Nanshan Venture, Cathay Capital (凯辉基金), GF Xinde, Hillhouse (高瓴创投), 5Y Capital (五源资本), Linear Capital (线性资本), Temasek's Vertex Growth, Yunfeng (云锋基金) and Meituan Longzhu, among others.
The company said the proceeds will go into two things: widening the Sunrise chip family across compute tiers, and building the software platform that links data collection, model training, simulation and inference deployment — the full pipeline a robot maker needs before a product can ship.
The numbers that matter
The raise was backed by operating data, not just a deck:
- Cumulative shipments of Sunrise-series chips have passed 8 million units.
- First-half 2026 revenue grew several-fold year on year.
- The customer list runs to more than 20 names, including UBTech (优必选), Zhiyuan (它石智航), Qianxun and the Beijing Humanoid Robot Innovation Centre, several already at batch-delivery stage.
This was also its second nine-figure raise of 2026. D-Robotics took a US$150 million B2 round in April (roughly RMB 1.1 billion); together, the two rounds put about US$550 million behind the company this year. For a firm that only split from its parent two years ago, that is a fast escalation in both cheque size and expectations.
Why capital is moving down the stack
For two years, Chinese AI fundraising chased the large model (大模型) developers — the chatbot and video-generator builders. That market is now crowded, expensive and hard to monetise. D-Robotics represents a different bet: that the durable profits sit one layer down, in the computing substrate that every physical-AI product must buy.
The logic is the same one that made Nvidia the centrepiece of the Western AI boom. If humanoid (人形机器人) and autonomous machines ship in volume, someone has to sell them the brain. D-Robotics is positioning as that someone — and Meituan's strategic ticket hints at where the first large buyers may come from: logistics, last-mile delivery and warehouse robotics, exactly the segments where a fleet of vehicles is a direct customer.
It is not alone in attracting that conviction. The same September window saw ByteDance's AI drug-discovery spin-off Anew Labs raise US$290 million at a US$1.5 billion valuation (Reuters, 16 September), and embodied-AI names such as Paxini begin mainland IPO preparation — evidence that capital is concentrating below the model layer rather than above it.
Honest limitations
- The US$400 million headline and the 8 million chips are company and press-release claims, not audited figures. "Cumulative shipments since 2024" is not an annual run-rate, and chip "shipments" to customers are not the same as robots actually operating in the field.
- Valuation was not disclosed. Without a price, readers cannot judge whether the round was cheap or rich, only that the cheque was large.
- Embodied-AI commercialisation is still early. Humanoid (人形机器人) demand at scale is unproven; much of the 8-million-unit base is earlier AMR, delivery and education kits, not the high-value humanoids the investment thesis ultimately depends on.
- The investor roster is heavy with corporate-strategic and local-government funds (Meituan, Hefei, Nanshan). That may reflect policy alignment and industrial strategy as much as pure market conviction about near-term returns.
- Geopolitics is a ceiling risk. Advanced-compute export controls and tighter supply chains could cap both the addressable market and the component supply that a chipmaker like this relies on.
What readers can do now
- Track D-Robotics' disclosed shipment and run-rate as a proxy for how fast China's embodied-AI stack is actually building, not just announcing. A rising annualised run-rate matters more than a cumulative milestone.
- Follow where strategic capital flows. Meituan, BYD and similar operators putting money into infrastructure — rather than apps — is the sharper signal of where real demand is expected, because those firms are also the future customers.
- Wait for hard disclosure. A single round is not a market verdict. Treat IPO filings, audited financials and per-customer deployment numbers as the moment to reassess, not the press release.