- SiMa.ai raised $150M at a $1.45B valuation, co-led by Fidelity and Amplify.
- Revenue quadrupled from 2024 to 2025, SiMa.ai says, but it gave no absolute figures.
- Its next chip arrives in the first half of 2028, while Nvidia’s Thor is already shipping.
SiMa.ai has raised a $150 million Series C at a $1.45 billion valuation, taking its total funding to $500 million, according to the company’s announcement.
Fidelity Management & Research Company and Amplify co-led the oversubscribed round. Alter Venture Partners, Dell Technologies Capital, Maverick Capital, +ND Capital, Point72, and StepStone Group also took part. AllianceBernstein, Baron Capital, J.P. Morgan, and the State of Michigan joined as new investors.
The price has moved fast. TechCrunch, citing PitchBook, reports that SiMa.ai was valued at $960 million after an $85 million Series B in July 2025, which makes the new valuation about 51% higher. When Tech Funding News covered its $70 million round in April 2024, total funding stood at $270 million, and Maverick, Point72, Amplify, and Dell Technologies Capital were all in that round.
Selling a way out of Nvidia
SiMa.ai builds chips and software for machines that must run AI on the device itself, inside a drone, a robot, or a car, within tight limits on power and heat. That is the opposite of the data-centre approach. Krishna Rangasayee founded the company in 2018. He was previously chief operating officer ofGroq and spent 18 years at Xilinx, the chipmaker.
Its second-generation chip, Modalix, is in production on TSMC’s 6-nanometre process. It is paired with Palette Neat, which SiMa.ai launched on June 16 and describes as an agentic software environment for physical AI. Developers describe their dataset and their limits on speed, delay, power, and accuracy, and the software builds the application. The software is open source and it cuts deployment from months to days or hours.
The pitch to Nvidia’s customers is about switching costs. According to Rangasayee, SiMa.ai’s module is pin-compatible with Nvidia’s, so a robot maker can swap one for the other, and its software can port code written for Nvidia’s own tools. The company is offering drone makers a live port during the sales meeting. SiMa.ai’s main target is Nvidia’s Jetson Orin family, Rangasayee claims about twice the performance at one-fifth the power.
“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. This investment will further catapult us into a dominant position, giving us the resources to scale our purpose-built platform and secure an undeniable lead in the Physical AI race,” said Rangasayee, founder and CEO of SiMa.ai.
The 2028 gap
The new money goes mainly to scaling Palette Neat and funding the next generation of hardware. SiMa.ai says that hardware will deliver 1,000 dense TOPS (trillions of operations per second) and arrive in the first half of 2028 as licensable machine-learning designs, chiplets (small modular chips), and full systems on a chip. It targets medium to high-end drones, humanoid robots, automotive driver-assistance systems, and AI-powered cockpits.
Nvidia is not standing still. It lists its Jetson Thor module, already on sale, at up to 2,070 trillion operations per second at 4-bit precision and up to 130 watts. The two figures are not like for like. Third-party spec listings describe Nvidia’s 2,070 as a sparse figure, meaning it counts shortcuts that skip zeros in a model’s data, with a dense figure roughly half that.
Efficiency is the standard pitch against Nvidia: Positron AI claims its next chip will deliver five times more tokens per watt than Nvidia’s upcoming Rubin GPU.
The announcement leans on Counterpoint Research, which projects 145 million cumulative physical AI device shipments across vehicles, robots, and drones between 2025 and 2035. Computer Weekly’s summary of the report says autonomous vehicles should be the largest market and commercial drones a major driver of shipment growth. Humanoid robots are the fastest-growing category, with cumulative installations projected to pass 100,000 by 2028.
SiMa.ai is raising into a crowded field. TFN reported Encord’s $60 million Series C in February, for the data behind robots, automotive, and drones. Humanoid robot startups raised $8.6 billion in the first half of 2026, 1.8 times all of 2025, according to Dealroom data. Embedd, a London startup, raised $2.7 million to cut the software integration work behind robots and drones, the same layer Palette Neat targets. Lyte, which builds robot perception, raised $165 million at a $1.6 billion valuation, with Fidelity returning.
What the investors say
Amplify, one of the co-leads, has been with SiMa.ai from the start.
“SiMa.ai has consistently proven its ability to out-innovate and out-execute the rest of the market, dramatically reducing the complexity of deploying advanced AI models on physical edge devices, and accelerating time to market for the world’s most sophisticated autonomous systems,” said Mike Dauber, general partner, Amplify Partners.
“As enterprise intelligence migrates into the physical world, it’s been challenged by the tradeoff between performance and power. SiMa.ai is shattering that barrier. Its platform delivers the software agility and Physical AI-optimised compute that modern enterprises require to scale computer vision and autonomous capabilities,” added Scott Darling, president and managing director, Dell Technologies Capital.
“SiMa.ai has created a differentiated, purpose-built Physical AI platform that answers the rigorous demands of real-world applications,” notes John Avirett, partner, StepStone Group.
Fidelity and Amplify are backing a company that has a second-generation chip in production and a software pitch aimed squarely at Nvidia’s customers. It has no 2028 chip yet, no named humanoid or carmaker customer, and no independent benchmarks.
The investors’ thesis is that the next fight in AI infrastructure happens inside the machine, not in the cloud. The open question is whether robot makers will switch silicon now, on SiMa.ai’s numbers, or wait to see the chip that is supposed to justify the valuation.
