D-Robotics just raised $400M to build the brain inside your next robot
Every prospector needs tools. And every gold rush needs a picks-and-shovels play. In the humanoid robot gold rush, D-Robotics is making a serious bid to be the company selling the shovels.
D-Robotics has just closed a massive $400 million Series C led by Mirae Asset with Meituan and a consortium of strategic and financial investors. The round follows a $270 million Series B from April of this year, meaning that D-Robotics has now raised well over half a billion dollars in the past six months.
That's impressive.
At least, apparently, to investors.
D-Robotics is building the Sunrise™ chip line, physical AI software, and development kits for robots. It's essentially edge compute infrastructure plus software for embodied AI robots: components for robot brains.
Eight million chips shipped before the Series C closed
D-Robotics reports that cumulative Sunrise chip shipments exceeded 8 million units by mid-2026.
Most of those went into a category the company doesn't always lead with: consumer smart hardware, which was the Sunrise line's original market. But the Sunrise S600, launched in November 2025 as the company's dedicated embodied AI chip, is for humanoid robots.
Within six months of launch, the S600 was adopted by more than 20 humanoid and embodied AI customers, including companies like UBTECH, PaXini Tech, Astribot, Fourier Intelligence, Booster Robotics, TARS, Spirit AI, and X Square Robot.
Many are already in mass production.
If you can hear the siren, that's the sound of the S600 becoming a standard component in Chinese humanoid robots at pretty much the exact moment the industry is beginning to scale from hundreds to thousands of units annually.
What the Sunrise S600 actually does
The S600 is designed for the specific compute requirements of embodied AI, which are substantially different from the inference demands of a language model running in a data center.
A robot needs real-time perception, sensor fusion, motor control, and environmental reasoning happening simultaneously, with strict latency requirements and within tight power budgets.
A chip optimized for cloud inference might not be perfect here.
D-Robotics built the S600 around those constraints: low power, real-time performance, support for multi-modal perception, and compatibility with the robotics software stack that physical AI developers are actually using. The company's developer ecosystem — 100,000 developers across 20 countries, 500+ universities, 500+ startups in its Gravity Program — suggests the platform has real traction beyond the chip itself.
D-Robotics doesn't just sell chips.
It supports developers through product development, supply chain selection, go-to-market strategy, and even fundraising introductions. That's stickier
The infrastructure bet
The $400M round announcement frames D-Robotics as infrastructure for the next generation of robotics categories. That framing matters. The company isn't betting on any single robot design winning the market — it's betting that whatever wins, it will need compute, and the compute might as well run on Sunrise silicon.
This is the Qualcomm play for robots. Qualcomm didn't make the winning smartphone. It made the chips that went into almost every winning smartphone. Arm didn't build a device — it licensed the architecture that powers most of the world's mobile computing. D-Robotics is positioning for the equivalent outcome in physical AI.
Whether it gets there depends on several things: whether the S600 and its successors stay competitive as the performance requirements of humanoid robots escalate, whether US-headquartered companies buy Chinese AI chips given the current geopolitical environment, and whether NVIDIA — which is pursuing embodied AI with IGX Thor and Halos for Robotics — decides to compete directly for the robot chip market.
NVIDIA already powers Digit 5. The IGX Thor platform is in multiple high-profile Western humanoid programs. D-Robotics dominates the Chinese market; NVIDIA is the default for US and European programs. That geographic division may hold. Or it may not, depending on how export controls evolve.
Here's what I'm seeing
Revenue growing "several times year-over-year" in H1 2026, 8 million chips shipped, 20+ humanoid robot customers already in production, $400M just closed ... D-Robotics is executing.
The company is building the infrastructure layer before the application layer fully exists, which is exactly the right timing if you believe the humanoid boom is real.
And, likely, it's being promoted by virtual of the fact that it's a Chinese competitor to NVIDIA, which is making significant revenue in robotics ... and China is its best customer.
It's not quite there yet ...
Importantly, these are very different chips.
D-Robotics says the S600 delivers 560 TOPS at INT8. That means roughly 560 trillion 8-bit integer operations per second, which is a common metric for highly quantized edge inference. On the other hand, NVIDIA says Jetson Thor delivers 2,070 TFLOPS at FP4. FP4 uses 4-bit floating-point math, aimed especially at transformer and generative-AI workloads.
| S600 | Jetson Thor | |
|---|---|---|
| Quoted metric | 560 TOPS INT8 | 2,070 TFLOPS FP4 |
| Optimized for | Efficient edge inference + robotics control | Large transformers, VLMs, VLAs, generative AI |
| Compute design | Dedicated BPU + CPUs + real-time MCUs | Huge Blackwell GPU |
| Robot implication | Strong integrated perception + control platform | Stronger emphasis on big AI models/reasoning |
So you cannot say Thor has 3.7× the AI compute just by dividing 2,070 by 560. They aren't apples-to-apples measurements.
Overall, this is an ecosystem bet
The bet is simple: humanoid robots need chips. Lots of chips.
And the company that earns the right to be the default compute platform for physical AI (the way Qualcomm became default for mobile) captures value across every robot that ships, regardless of which robot brand wins.
That's a compelling position.
There's competition with NVIDIA, of course, but there's also national pride in China at using a Chinese solution for Chinese robots.
Sources
Written by
I tell stories with data and build startups. I've raised angel and VC funds, led software companies and startups, and chronicled the rise of the mobile economy. Currently I write for Forbes, host several podcasts, and am starting up HumanoidMegaHub. I was TUNE's mobile economist and Singular's VP of Insights, built the research team at VentureBeat, and managed teams creating software for partners like Intel and Disney. In past lives I built websites, e-learning platforms, and mobile apps, and, a long time ago, was named to Folio's top 100 of the media industry's "most innovative entrepreneurs and market shaker-uppers.”