SiMa.ai, a developer of chips and software for running AI directly on devices, has raised a $150 million Series C at a $1.45 billion valuation.
The round was co-led by Fidelity Management & Research Company and Amplify. Alter Venture Partners, Dell Technologies Capital and StepStone Group also participated. The financing brings the company’s total funding to more than $500 million.
For operators building robots, drones and camera-based systems, the raise is another sign that investors see a substantial market beyond cloud-hosted generative AI: devices that must make decisions locally, quickly and within tight power budgets.
What changed
Founded in 2018 by Krishna Rangasayee, previously COO of chipmaker Groq, SiMa.ai sells energy-efficient chips and related software for on-device AI workloads. Its target systems include robots, drones, cameras and other connected physical devices.
The company was valued at $960 million following an $85 million Series B in July 2025, according to PitchBook. The new valuation represents a meaningful step up and gives SiMa.ai more resources to compete in a hardware market dominated by Nvidia while serving customers that may not need—or cannot support—data-center-class GPUs.
Its central proposition is to avoid sending device data to the cloud for every inference. That can reduce latency, limit bandwidth dependence and keep operations functioning where connectivity is intermittent or unavailable.
Why edge inference matters
Many physical systems have requirements that differ from chatbot and data-center workloads. A drone responding to its environment, a robot navigating a facility, or a camera identifying an event may need an answer in milliseconds. Transmitting video or sensor data to a remote service introduces network delay and creates an operational dependency on connectivity.
Local inference can also change the economics. Moving large volumes of sensor data to the cloud incurs networking and compute costs, while a device with a suitably efficient processor can handle repeatable inference workloads on site. For deployments at fleet scale, power consumption, thermal constraints and unit cost can matter as much as peak model performance.
That creates an opening for specialized silicon vendors. SiMa.ai is positioning its chips as lower-cost alternatives to Nvidia GPUs for these use cases, though customers will ultimately evaluate that claim against real-world throughput, model compatibility, software tooling and total cost of ownership.
The software question is as important as the chip
Winning edge AI deployments requires more than silicon. Hardware suppliers need developer tools that let teams convert, optimize, deploy and monitor models across constrained devices. They must also support the computer-vision and sensor-processing pipelines common in industrial and robotic systems.
This is where incumbent ecosystems retain an advantage. Nvidia’s hardware is widely used, but its software platform and developer familiarity are also major buying factors. SiMa.ai’s funding can help it invest in the software layer and customer support needed to lower switching costs for engineering teams.
What to watch next
The important signals now are commercial rather than financial. Watch for named production deployments, evidence that SiMa.ai can support customers from evaluation through fleet operations, and details on how its chips perform on the models customers actually use.
Also watch whether demand for humanoid robots and other “physical AI” systems translates into sustained chip volumes. That market is attracting considerable attention, but it remains dependent on manufacturers moving from pilots to reliable, economical production.
SiMa.ai has new capital and a higher valuation. Its next test is whether it can turn the edge AI thesis into repeatable customer adoption in a market where performance, power, cost and software maturity all determine the winner.
