SiMa.ai raises $150M Series C at $1.45B valuation to compete with Nvidia in humanoid and drone chips

SiMa.ai closed a $150 million Series C on Sept. 28, co-led by Fidelity Management & Research and Amplify, valuing the eight-year-old chipmaker at $1.45 billion and pushing its total capital raised to $500 million. The San Jose company, founded in 2018, builds chips and software meant to let robots, drones, and cameras run AI directly on the device rather than round-tripping to the cloud -- a category it and its backers are betting will absorb a meaningful share of the money currently flowing to Nvidia's edge hardware.

The distinction matters because of where the compute physically happens. A cloud-hosted model answers a query by sending data to a data center and waiting for a response; a robot deciding whether the object in front of it is a person cannot afford that round trip, so the inference has to run on the device itself, within a power and heat budget a data-center GPU was never designed to meet. SiMa.ai's software layer, Palette Neat, is the company's attempt to make that on-device deployment as routine as calling a cloud API -- what it calls an "agentic software environment for Physical AI," letting a developer target the company's chips without hand-tuning every model for the hardware's specific constraints.

Founder and chief executive Krishna Rangasayee, previously chief operating officer at Groq and an 18-year veteran of Xilinx, said the company's revenue quadrupled year-over-year between 2024 and 2025, with growth continuing into 2026 -- a company-reported figure, not an independently audited one. The round drew a mix of return and new backers: existing investors from a $85 million Series B closed in July 2025 at a $960 million valuation were joined this time by four new names -- AllianceBernstein, Baron Capital, J.P. Morgan, and the State of Michigan, the kind of public-pension participation state retirement systems typically make through committed venture-fund allocations rather than a named-company check; SiMa.ai's press release does not specify which Michigan fund or vehicle wrote it. Series B and C together account for $235 million of the $500 million total raised to date -- meaning roughly $265 million came from Series A and earlier rounds since 2018, a period SiMa.ai's current disclosures do not break out round by round. The release also names a dozen "Fortune 500 leaders and technology innovators," including Bosch, Emerson, Micron, Synopsys, and TRUMPF, as customers and technology partners, though it does not disclose revenue or unit-volume figures for any of them individually.

SiMa.ai is not the only company chasing this exact gap. Positron AI claims five times more tokens per watt than Nvidia's upcoming Rubin GPU; Lyte raised $165 million at a $1.6 billion valuation for robot-perception hardware; Encord closed a $60 million round for the data infrastructure that trains physical-AI systems. The pattern across all of them is the same wager: that Nvidia's dominance in data-center AI chips does not automatically transfer to the edge devices -- robots, drones, cameras, cars -- where power, heat, and cost constraints are different enough that a specialist could out-compete a generalist. Nvidia is not conceding the point; Jetson Thor is the company's own answer to exactly this bet, shipping now rather than promised for later.

SiMa.ai's pitch against Nvidia has two parts, on two different timelines. For chips shipping today, the company claims its MLSoC line delivers roughly twice the performance of Nvidia's Jetson Orin at about a fifth of the power draw, in a pin-compatible module a customer can swap in without a board redesign. For 2028, the money from this round funds a wholly new chip family -- spanning machine-learning IP, chiplets, and full systems-on-chip -- targeted at 1,000 dense TOPS of compute for humanoid robotics, automotive driver-assistance, and drones.

"Physical AI in humanoids, automotive, and drones is the gateway to a $50 trillion market," Rangasayee said in the round's announcement -- a company's own framing of its addressable market, not an independent estimate, and one this piece does not adopt as fact. The more grounded comparison is the one directly in front of investors: SiMa.ai is asking them to fund a chip targeting 1,000 dense TOPS that won't exist for roughly 18 more months, against Nvidia's Jetson Thor, which already ships today at up to 2,070 TOPS at 4-bit precision within a 130-watt power envelope and has already been independently benchmarked by third parties since launch.

None of that makes the round irrational -- Fidelity and Amplify are underwriting a bet on the category, not a guarantee of the 2028 chip's success, and the revenue-quadrupling claim, if it holds up, suggests SiMa.ai's current-generation hardware is already finding paying customers regardless of what happens to the roadmap. But it does mean the $1.45 billion valuation prices in a product that is still a target on a slide, not silicon in a lab.

Against this newsroom's own register of closed AI funding raises, $150 million is a small number -- a reminder that even a name-brand, oversubscribed round in the "physical AI" category is still modest money next to the tens of billions moving through frontier-model labs and hyperscaler data-center deals the same month. The chip market SiMa.ai is chasing may be real; the capital behind it, so far, is not yet close to hyperscaler scale.