Edge-AI Chip Startup SiMa.ai Raises $150 Million at $1.45 Billion to Build a 1,000-TOPS Robot Processor

SiMa.ai, a San Jose startup building custom processors for "physical AI" — robots, drones and vehicles that run AI on the device rather than in the cloud — announced on Monday a \$150 million Series C at a \$1.45 billion valuation, up from roughly \$960 million after its July 2025 Series B.

Fidelity Management & Research Company and Amplify co-led the round, with participation from Alter Venture Partners, Dell Technologies Capital and StepStone Group. AllianceBernstein, Baron Capital and J.P. Morgan joined as new investors. The raise pushes SiMa.ai's total funding above \$500 million.

Founder and CEO Krishna Rangasayee, who founded the company in 2018 after serving as COO at chipmaker Groq, framed the market in sweeping terms: "Physical AI in humanoids, automotive, and drones is the gateway to a \$50 trillion market that has remained largely untouched by modern innovation. While others are still figuring out the pieces or repurposing their cloud offerings, we've built the entire puzzle."

The new capital has two destinations. The first is Modalix, the company's next-generation system-on-chip for embedded robotics, which SiMa.ai says will target 1,000 TOPS (tera operations per second) when it arrives in the first half of 2028. That lands deliberately between Nvidia's edge lineup: the flagship Jetson AGX Thor modules already deliver 2,000 FP4 TOPS for full-scale humanoids, while the mainstream T2000 and T3000 modules cover roughly 400 and 865 TOPS. SiMa.ai's pitch is not raw throughput but power and price — its current production module fits an Nvidia form factor, needs no board redesign, and runs multiple large language models alongside vision and sensor models under 10 watts.

The second bet is software. Palette Neat, an open-source, agentic development environment launched in June, lets developers build applications with natural-language instructions and maps them onto SiMa.ai hardware; the company estimates it preserves about 90% of a developer's existing software investment — a crucial number, because Nvidia's CUDA ecosystem is exactly the switching cost SiMa.ai needs customers to escape.

For now, the customer list skews industrial rather than headline-grabbing: Robert Bosch, Emerson Electric, Micron Technology, Synopsys, ARK Electronics and AverMedia. Counterpoint Research projects cumulative shipments of physical AI devices reaching 145 million units by 2035, a forecast SiMa.ai cites as evidence of the runway ahead.

The honest read is that this is a bet on power envelope and developer experience, not peak performance — a 2028 chip targeting 1,000 TOPS will land into an Nvidia roadmap that already ships twice that number today. But drone makers and factory-floor robots do not buy TOPS counts; they buy watts, latency and time-to-deployment. If SiMa.ai can make "days, not months" the default for moving a model onto its silicon, the funding buys it a credible shot at the tiers Nvidia prices for margin. The next test is Modalix actually shipping on schedule in 2028 — chip years have a way of stretching.