SiMa.ai raises $150M at $1.45 billion valuation, targets Nvidia's Orin with 2028 chip

TechCrunch, citing PitchBook, put SiMa.ai at $960 million after an $85 million Series B in July 2025 ^[4]. Against that mark the step-up is about 51%. Or rather, for the shares that already existed, it is closer to 35% ^[1]. Treat the $1.45 billion as post-money and subtract the $150 million of new cash, and the pre-money value is $1.30 billion ^[1]. Against the $500 million SiMa.ai has raised in total, the new price is 2.9 times every dollar ever invested ^[2].

No price-to-sales multiple can be built from the record. SiMa.ai reports its growth as a quadrupling of revenue from 2024 to 2025 and does not give the base ^[6].

The chip dates say more. SiMa.ai rates its next generation at 1,000 dense TOPS, due in the first half of 2028 as licensable designs, chiplets and full systems on a chip ^[7]. Nvidia lists Jetson Thor, already on sale, at up to 2,070 TOPS at 4-bit precision and up to 130 watts ^[8]. Third-party spec listings describe that as a sparse figure, with a dense equivalent of roughly half ^[8]. Half of 2,070 is about 1,035, so on the dense measure SiMa.ai's 2028 part arrives at about 97% of what Thor lists today ^[3].

Until then the company sells Modalix, its second-generation chip, in production on TSMC's 6-nanometre process ^[9]. Rangasayee pitches it at Nvidia's Jetson Orin family, claiming about twice the performance at one-fifth the power ^[12]. Drone makers are offered a live port of their Nvidia code during the sales meeting ^[11]. The founder, a former Groq chief operating officer who spent 18 years at Xilinx ^[16], put the prize in larger terms. "Physical AI in humanoids, automotive, and drones is the gateway to a $50 trillion market that has remained largely untouched by modern innovation," Rangasayee said ^[13]. The forecast the announcement relies on, from Counterpoint Research, is 145 million cumulative physical AI device shipments across vehicles, robots and drones between 2025 and 2035 ^[15].

Where the price goes depends mostly on what Orin customers do next. If they swap modules through 2027, the quadrupling compounds on Modalix and the 2028 part becomes an upgrade for customers SiMa.ai already has. Should they move up to Thor instead, SiMa.ai spends two years selling against the tier its buyers are leaving. A third path runs through the licensable designs and chiplets ^[7]. On that path, part of the 2028 revenue would come as licences on other companies' silicon.

I think the round is priced on the first path. The new money goes to software and the 2028 part ^[14], so the Orin-replacement sale has to run on Modalix as it stands. The counter-case is that Palette Neat is the asset. It is open source, and SiMa.ai says it cuts deployment from months to days or hours ^[10]. Software that ports Nvidia code ^[11] is worth something whichever chip wins. If customers adopt it and stay on Nvidia boards, I have misjudged what Fidelity and Amplify paid for ^[2].