Latest

Solid AI. Smarter Tech.

SiMa.ai Raises $150 Million to Scale Physical AI for Robots, Drones and Cars

$150M RAISE SiMa.ai · Physical AI · Edge Inference · Robotics · Drones · Automotive · September 2026

I’ve spent enough time around AI hardware to notice a pattern: the loudest part of the market is usually the data center, while a huge amount of the actual work eventually has to happen somewhere much smaller.

A drone cannot carry a data center.

A mobile robot cannot afford unlimited power.

A vehicle cannot assume a perfect cloud connection every second.

That is the market SiMa.ai is chasing, and the San Jose-based company has just raised $150 million to push that strategy much further.

The financing, reported on September 28, 2026, values the company at about $1.4 billion and brings its reported total capital raised to roughly $500 million.

The important distinction: The new $150 million is for SiMa.ai's next generation of Physical AI silicon. The company's current second-generation Modalix platform is already in volume production; the planned third-generation platform is a future product targeted for the first half of 2028.
$150M
New Capital Reported
$1.4B
Reported Valuation
50 TOPS
Current Modalix Class
1,000 TOPS
Gen 3 Target at 80W
SiMa.ai Physical AI processor for robotics drones and autonomous systems
SiMa.ai is building specialized edge-AI hardware and software for machines that need to perceive, reason and act locally.

What SiMa.ai Actually Does

SiMa.ai is not trying to build another general-purpose data-center GPU.

The company's core idea is to build purpose-built machine-learning system-on-chips that combine AI acceleration, application processing and other elements of the edge-compute pipeline in a power-constrained package.

Its current Modalix platform is designed for computer vision, transformers, large language models and multimodal workloads.

That combination is aimed at physical machines rather than chat windows.

Think warehouse robots, inspection systems, intelligent cameras, autonomous vehicles, drones and industrial equipment.


Why Physical AI Needs Different Hardware

A cloud server can draw hundreds of watts, use large cooling systems and access virtually unlimited external power.

A robot cannot.

Every watt consumed by onboard AI competes with motors, sensors, communications and battery life.

A drone has an even harsher constraint: additional compute can directly reduce flight endurance.

This is the reason performance-per-watt matters so much. In physical AI, the fastest chip on paper is not necessarily the best chip if its power consumption makes the product impossible to deploy.

SiMa.ai says its Modalix platform is designed to deliver multimodal AI under 10 watts, with the full application pipeline running on-device.


The Current Modalix Platform Is Already Shipping

The new financing is easier to understand once you separate today's product from tomorrow's roadmap.

SiMa.ai announced its second-generation Modalix platform in 2024 and began moving it into production in 2025.

The company now describes Modalix as a 50-TOPS multimodal AI platform operating below 10 watts, with support for CNNs, vision transformers, LLMs and language-and-vision models.

The Modalix system is available in several form factors, including System-on-Module products, development kits and a PCIe accelerator card.

That flexibility is important because many industrial customers cannot redesign an entire machine around a new processor.


The Pin-Compatible Detail Is More Important Than It Sounds

One of SiMa.ai's strongest practical selling points is that its Modalix SoM is designed to be pin-compatible with leading NVIDIA Jetson SoM form factors.

That means an OEM can potentially change the compute module without rebuilding the entire carrier board.

It does not mean the two platforms are electrically or software-identical in every respect.

It means SiMa.ai is attacking one of the biggest barriers in embedded hardware: the cost and risk of redesigning a product around a new processor.

Why OEMs care

A chip is only valuable when a company can get it into production.

For a robot or drone manufacturer, the engineering bill can include carrier-board changes, thermal changes, power redesign, drivers, software porting and validation.

Reducing the number of those changes can be just as important as improving raw AI throughput.


SiMa.ai's New Target Is Far More Ambitious

The newly funded company says the money will help build its third-generation platform.

The target is dramatic: approximately 1,000 TOPS at 80 watts, with chiplets, machine-learning IP and SoCs aimed at drones, humanoid robots, automotive ADAS and AI-powered vehicle cockpits.

The schedule is currently the first half of 2028.

That number should therefore be read as a development target, not a product specification consumers can buy today.

It also highlights how quickly edge AI is changing.

What once sounded like data-center-class AI compute is increasingly being discussed in machines that may have strict thermal and energy limits.


Why Chiplets Matter to SiMa.ai's Roadmap

The move toward chiplets gives SiMa.ai another architectural option.

Instead of building every function into one monolithic piece of silicon, companies can combine specialized components into a larger package.

For a physical-AI platform, that could allow different processing blocks to be tuned for different workloads while preserving a broader software and system architecture.

It also gives SiMa.ai a potential path to scale performance without abandoning the power-efficiency focus that defines its current products.


Memory Is Becoming a First-Class AI Problem

Another overlooked part of the story is memory.

More capable AI models create more pressure on memory bandwidth, capacity and movement of data.

SiMa.ai has been working with Micron on this exact problem.

In April 2026, Micron made a strategic investment in SiMa.ai and the companies announced the integration of LPDDR5X memory into the Modalix platform.

That is important because edge AI efficiency is not simply about adding more compute.

The system also has to feed data to the accelerator efficiently while keeping power consumption under control.

For developers, watch the memory numbers

When comparing edge-AI hardware, do not stop at TOPS. Look at memory type, bandwidth, capacity, data movement, camera interfaces and the power envelope of the complete application.


Where the Technology Is Being Used

SiMa.ai's customer and partner activity provides a clearer picture than the funding announcement alone.

The company is working across robotics, drones, industrial automation, smart vision, automotive, aerospace and defense.

Recent partnerships include ARK Electronics for autonomous drones, STIGA for robotic lawn mowers, Nota AI for model optimization, and Cerence-related automotive applications.

SiMa.ai also has robotics software that integrates with ROS 2 and supports sensor combinations such as cameras, LiDAR and IMUs.

Those are important signals because they show the company is trying to build an ecosystem around the silicon rather than selling a processor as an isolated component.


Robots Need a Closed-Loop AI Stack

A chatbot can take several seconds to generate an answer without creating a physical problem.

A moving robot does not have the same luxury.

Perception, localization, planning and action have to operate as a loop.

SiMa.ai's robotics materials describe workflows where sensor data enters the system, AI perception runs locally, and the resulting information feeds navigation and control components.

The closer those steps are to the machine, the less the system depends on cloud round trips.


What Generic Coverage Often Misses

1. SiMa.ai is selling a platform, not just a chip

Modalix is paired with software, developer tooling, model compilation and deployment infrastructure.

2. The current product is already in production

The $150 million is primarily about scaling the next generation rather than funding a company that has no hardware in the field.

3. Memory is part of the strategy

The Micron relationship shows that performance-per-watt depends on the compute and memory architecture working together.

4. Non-humanoid robots may matter more in the near term

SiMa.ai says robotics and drones currently make up the largest share of its business and expects non-humanoid robotics to drive near-term volumes.

5. U.S. manufacturing and supply-chain considerations matter

The ARK Electronics partnership gives SiMa.ai a U.S.-based hardware ecosystem for drone deployments, a relevant consideration for commercial and government customers.


SiMa.ai vs. a Conventional Edge GPU Approach

Factor SiMa.ai Approach Conventional GPU Approach
Architecture Purpose-built MLSoC with heterogeneous compute General-purpose GPU accelerator plus supporting processors
Primary emphasis Physical AI and edge inference Broad graphics and AI workloads
Power focus Designed around constrained edge systems Varies significantly by GPU and system
Software Palette platform and physical-AI tooling GPU SDKs, drivers and broader developer ecosystems
Upgrade strategy Pin-compatible SoM options for selected designs Depends on the specific carrier and system architecture

Pros and Cons of SiMa.ai's Strategy

Potential Advantages

  • Purpose-built for power-constrained Physical AI
  • Current Modalix hardware is already shipping
  • Full-stack hardware and software approach
  • Multiple form factors for embedded deployment
  • Strong focus on robotics, drones and industrial systems

Important Challenges

  • Large GPU ecosystems remain deeply established
  • Future Gen 3 performance targets are not current products
  • Industrial customers require long qualification cycles
  • Software migration can still require engineering effort
  • Performance claims need workload-specific evaluation

Andrew Ng's Old Prediction Fits This New Hardware Race

“AI is the new electricity.”

— Andrew Ng, co-founder of Coursera and DeepLearning.AI

That line has been repeated for years, but edge AI gives it a useful hardware interpretation.

Electricity became transformative when it moved beyond centralized generators and into factories, homes, vehicles and everyday machines.

AI is following a similar path.

The interesting question is no longer only how smart the largest model can become.

It is how cheaply and efficiently intelligence can be embedded into millions of physical systems.


Watch SiMa.ai's Modalix Handle 16 Video Streams

The demonstration shows SiMa.ai's Modalix SoM handling a multi-stream intelligent-video workload on the edge, illustrating the type of power-constrained application the platform is designed to target.


Amazon: Edge AI Hardware for Developers

NVIDIA Jetson Orin Nano Super Developer Kit

The Jetson Orin Nano Super is a useful reference point for developers exploring the edge-AI market. Comparing it with purpose-built systems such as Modalix can help clarify the trade-offs between a broad GPU software ecosystem and specialized edge inference hardware.

Check Jetson Orin Nano Super on Amazon →

Raspberry Pi AI Kit

A lower-power development platform is useful for experimenting with computer vision, sensor processing and local AI applications before moving to industrial-grade embedded hardware.

Check Raspberry Pi AI Kit on Amazon →

What the $150 Million Will Actually Buy

Funding a semiconductor company is different from funding a typical software startup.

Money has to support chip architecture, verification, physical design, software, testing, manufacturing, customer qualification and long-term product support.

SiMa.ai says the new money will specifically help develop the third-generation platform.

The roadmap includes chiplets, new ML IP and SoCs for a range of Physical AI systems, with a targeted first-half-2028 launch window.

That means investors are effectively financing the next stage of the company's silicon roadmap rather than simply adding sales capacity to the current product.


Why the Robotics Market Is the Real Test

It is easy to make an impressive AI chip benchmark.

It is much harder to make a robot that can run all day, process multiple sensors, make decisions in real time and survive a harsh environment without exceeding its thermal or power budget.

That is why SiMa.ai's real test will happen in the field.

The company needs its silicon to deliver predictable performance across complete applications, not just isolated model benchmarks.

Its partnerships in drones, robotics and automotive are therefore strategically important because they expose the platform to the messy constraints of real products.


The Bigger Picture

The $150 million financing is a useful signal that the AI infrastructure story is spreading beyond hyperscale data centers.

There is a second AI hardware market emerging underneath the cloud.

It lives inside machines.

Robots need perception.

Drones need navigation.

Cars need real-time assistance.

Factories need machine vision.

Those workloads do not always need the largest possible processor. They need the right combination of compute, memory, software, latency and power efficiency.

SiMa.ai is betting that this market will become enormous.

The $150 million round gives it more resources to pursue that bet, but the real proof will come from production deployments and the success of its next-generation silicon.

The most interesting thing about SiMa.ai is not that it wants to challenge NVIDIA. It is that it is betting AI's next frontier may be measured in millions of intelligent machines, not just millions of GPU servers.

Physical AI 2026: NVIDIA, Tesla Optimus & Market Reality

As edge silicon startups push AI into autonomous machines, the broader battle for the robotics market is rapidly escalating. Read our complete 2026 Physical AI guide to examine NVIDIA's robotics stack, track the progress of Tesla Optimus, and separate trillion-dollar market hype from deployable reality.

Read the Physical AI Guide →


Frequently Asked Questions About SiMa.ai

What is SiMa.ai?

SiMa.ai is a San Jose-based AI semiconductor company building purpose-built hardware and software for Physical AI and edge inference in applications such as robotics, drones, automotive systems and industrial automation.

How much did SiMa.ai raise in September 2026?

SiMa.ai raised $150 million in new funding, according to September 28, 2026 reporting. The round reportedly values the company at about $1.4 billion and brings total capital raised to roughly $500 million.

What will SiMa.ai use the $150 million for?

The company says the new funding will support development of its third-generation platform, including chips, chiplets and machine-learning IP targeted at drones, humanoid robots, automotive ADAS and AI-powered vehicle cockpits.

What is SiMa.ai Modalix?

Modalix is SiMa.ai's second-generation MLSoC platform for multimodal AI at the edge. The company lists 50 TOPS of AI performance in a sub-10-watt platform and offers the technology in multiple form factors.

Is SiMa.ai's 1,000-TOPS chip available now?

No. The approximately 1,000-TOPS, 80-watt figure is a target for SiMa.ai's third-generation platform, which the company says is planned for the first half of 2028. It is not the specification of the current shipping Modalix product.

Disclosure: Some of the product links in this article are Amazon affiliate links. If you purchase a product through one of these links, we may earn a small commission at no additional cost to you. Our recommendations are based on the product's relevance and usefulness to our readers.

No comments:

Post a Comment

Explore More