Decentralized and autonomous physical AI "without the cloud"? SiMa.ai covers everything from robotic lawnmowers to smart machines

Cars and robots in real time: The secret key to the next generation of AI

Attack on the AI ​​market: How Nota AI and SiMa.ai are paving the way for smart machines

Artificial intelligence is increasingly leaving the gigantic data centers and conquering our physical world. Whether in autonomous vehicles, intelligent traffic lights, or industrial robots – so-called "physical AI" must process highly complex data sets directly on-site in milliseconds. But here, the industry is encountering a physical limit: Conventional AI models are simply too large and power-hungry for the tiny, energy-efficient chips in these devices. A permanent connection to the cloud is often not an option due to high latency and security concerns. A groundbreaking partnership is now addressing precisely this technological bottleneck: The South Korean software optimizer Nota AI and the Californian chip specialist SiMa.ai are joining forces. By combining extreme model compression with highly specialized edge AI chips, they aim to solve the industry's fundamental efficiency problem. Read on to find out why this strategic alliance goes far beyond a typical collaboration and how it could redefine the rules of the game in the multi-billion-dollar edge AI market.

SiMa.ai's MLSoC platform is designed for the embedded edge: AI models run directly on the chip, inference takes place locally on the device – without image or sensor data having to be transferred to the cloud for every decision.

When optimization software meets ML SoCs: Nota AI and SiMa.ai join forces for Physical AI

On March 25, 2026, Nota AI and SiMa.ai signed a strategic partnership in San Jose, California, with the stated goal of jointly developing the physical AI market. What at first glance appears to be just another collaboration announcement in the overheated AI sector reveals itself upon closer inspection as a strategically sound merger of two highly specialized players who complement each other in a complementary way – one on the software side, the other in the area of ​​dedicated AI hardware. To understand the significance of this alliance, it is worthwhile to first take a closer look at both companies and the market environment in which they operate.

The foundation: Who are Nota AI and SiMa.ai actually?

Nota AI was founded in Seoul, South Korea, in 2015 and has since established itself as a leading company in the field of AI model optimization and compression. The company's core product is the NetsPresso platform, a hardware-aware AI optimization platform comprised of three modules: Model Searcher (automated model search and neural architecture search), Model Compressor (compression, structured pruning, and filter decomposition), and Model Launcher (quantization, conversion, and cross-device deployment). NetsPresso's key strength lies in its ability to automatically optimize AI models without requiring in-depth expert knowledge—a significant advantage in a market where the shortage of highly skilled AI engineers is a structural bottleneck.

Nota AI claims to be able to reduce the size of AI models by over 90 percent without significantly compromising model accuracy. The company has raised approximately $42.6 million in funding by 2024, including investments from the Korea Development Bank, Mirae Asset Securities, and strategic investors in the semiconductor industry. This investor structure—with Samsung SDS and LG CNS as early strategic partners—demonstrates that Nota AI has been positioned at the intersection of software optimization and the semiconductor industry from the outset.

SiMa.ai, on the other hand, was founded in 2018 in San Jose, California, by former Groq COO Krishna Rangasayee and specializes in developing dedicated machine learning system-on-chips (MLSoCs) for the edge market. The company has raised approximately $355 million in venture capital by 2025, including $85 million in an oversubscribed funding round in July 2025 led by Maverick Capital. The company's current valuation is around $960 million – just below unicorn status. Investors include Maverick Capital, Amplify Partners, Dell Technologies Capital, and prominent chip executive Lip-Bu Tan.

SiMa.ai's flagship product is the second-generation Modalix MLSoC, a system-on-chip based on TSMC's N6 process and available in configurations ranging from 25 to 200 TOPS (Tera Operations Per Second). The chip supports CNNs, transformers, LLMs, LMMs, and generative AI at the edge and, according to the manufacturer, achieves more than ten times the performance-per-watt efficiency of alternative solutions. SiMa.ai delivers not only hardware but also a complete, software-centric platform, including the Palette SDK, designed to simplify the development and deployment of complex edge AI applications without performance degradation.

The core problem that this partnership is intended to solve

To understand the strategic core of this alliance, one must first grasp the fundamental technical dilemma facing the entire physical AI industry. AI models trained in the cloud or on high-performance data centers are typically large, computationally intensive, and power-hungry. They run exceptionally well on GPUs with access to ample cooling and power. However, at the network edge—directly within robots, vehicles, surveillance cameras, production machines, or transportation systems—the conditions are entirely different: limited computing power, tight energy budgets, often less than 10 watts of system power, and the need to react in real time.

Cloud solutions are unsuitable for many of these applications for several reasons. Historically, the latency of traditional cloud architectures ranged from 100 to 500 milliseconds; modern edge AI systems, on the other hand, aim for inference times of less than 10 milliseconds, and in safety-critical applications, even in the range of 1 to 10 milliseconds. Collision detection in a vehicle environment or hazard analysis in an industrial plant simply cannot wait for a server response. Furthermore, there are data privacy concerns and the question of connectivity: Who guarantees that a robot in a cold storage facility or a camera module on a bridge will always have a stable internet connection?

The fundamental problem is this: the models are too large for the hardware, and the hardware alone cannot solve the problem. This is precisely where the tension arises from which the partnership between Nota AI and SiMa.ai derives its value. SiMa.ai delivers the most powerful and efficient dedicated AI chip for the embedded edge – but without optimized software, some of this performance remains untapped. Nota AI provides the ability to compress and optimize any AI model so that it is precisely tailored to the specific hardware architecture of the target chip – but without powerful and efficient hardware, the benefits of this optimization remain limited.

Our global industry and economic expertise in business development, sales and marketing - Image: Xpert.Digital

Industry focus areas: B2B, digitalization (from AI to XR), mechanical engineering, logistics, renewable energies and industry

More information here:

• Expert Business Hub

A thematic hub offering insights and expertise:

• Knowledge platform covering global and regional economies, innovation and industry-specific trends
• A collection of analyses, insights, and background information from our key areas of focus
• A place for expertise and information on current developments in business and technology
• A hub for companies seeking information on markets, digitalization, and industry innovations

The interplay between NetsPresso and Palette SDK: More than the sum of its parts

The technical heart of this partnership lies in the integration of the two SDK platforms: NetsPresso from Nota AI and Palette from SiMa.ai. While Palette provides the deployment framework for the Modalix MLSoC and manages the entire software stack for edge AI applications, NetsPresso handles the upstream model optimization phase.

The concept works as follows: A user wants to run a complex computer vision model – for example, for pedestrian detection in an urban traffic system – on an embedded system with low power consumption. In its raw form, the model is simply too large and computationally intensive. NetsPresso analyzes the model architecture, identifies redundant parameters, automatically applies structured pruning and quantization, and thereby reduces the model size to a fraction of the original – while maintaining detection accuracy. The optimized model is then deployed via the Palette SDK on the Modalix MLSoC, which, thanks to its specific hardware architecture, is designed precisely for this type of workload.

The result is a system that infers directly on the device, requires no cloud connection, consumes significantly less energy, and yet still handles high-performance tasks. For industrial environments where maintenance costs, reliability, and energy efficiency are direct economic factors, this is not a theoretical advantage, but a tangible competitive advantage.

The market: Why Physical AI is becoming an economic factor right now

The macroeconomic backdrop of this partnership is anything but accidental. Physical AI – that is, artificial intelligence that operates in the physical world and doesn't just process data – is developing into one of the most significant growth markets in the technology sector. The global market for physical AI was valued at approximately US$4.12 billion in 2024 and is projected to grow to around US$61.19 billion by 2034, representing a compound annual growth rate (CAGR) of 31.26 percent. Other estimates even predict a CAGR of 32.53 percent by 2033, with a market volume of nearly US$50 billion.

The overarching edge AI market, which includes physical AI as a sub-segment, is projected to grow from approximately $24.9 billion in 2025 to over $118 billion by 2033, at a CAGR of 21.7 percent. North America currently dominates with a market share of around 41 percent, while the Asia-Pacific region—and thus Nota AI's home market in South Korea—is considered the fastest-growing segment. This geographic complementarity of the two partner companies—a US hardware company and a South Korean software company—is strategically significant, as it potentially opens access to both important global regions.

As far as growth drivers are concerned, there are essentially three forces: firstly, the rapidly increasing proliferation of IoT devices and networked systems that need to process data at the point of origin; secondly, the growing demand for autonomous systems in robotics, mobility and manufacturing; and thirdly, the increasing regulatory and data protection requirements that favor a shift of data processing away from the cloud and towards the device.

Three markets in focus: ITS, security and robotics

The partnership identifies three specific application areas in which the joint solution will primarily be used: Intelligent Transport Systems (ITS), safety and security applications, and the broader fields of robotics and mobility.

In the field of intelligent transportation systems, the technology enters a market valued at US$9.84 billion in 2025 and projected to grow at a CAGR of over 10 percent by 2033. The requirements of the ITS environment—real-time detection of vehicles, pedestrians, traffic signs, and hazardous situations, combined with high system availability and low energy consumption—perfectly illustrate the strengths of the combined solution. AI-powered traffic management solutions have demonstrably reduced congestion by 25 to 30 percent in major cities. Nota AI's Nota Vision Agent (NVA) solution, specifically designed for video-based AI intelligence, is optimized for precisely this application scenario and is being adapted for the SiMa.ai hardware.

In the area of ​​security and protection – meaning classic video surveillance, access control, and perimeter monitoring – edge deployment also offers a clear advantage over cloud-based approaches, both in terms of data protection and response speed. And in the field of robotics and autonomous mobility, the demand for chips capable of processing multimodal AI models in real time is growing rapidly with the increasing prevalence of cobots in industry and autonomous vehicles.

The strategic logic behind the cooperation: Why now and why these partners?

From a business perspective, this partnership follows a clear logic. SiMa.ai has a technically compelling product on the market with the Modalix MLSoC and boasts a broad global sales network as well as an established partner network. What the company lacks is a seamless software layer that supports customers in model adaptation and accelerates the transition from proof of concept to production application. Because the most frequent bottleneck in edge AI deployment is not the hardware, but rather the question: How do you efficiently get the model onto the chip?

Nota AI, in turn, possesses a sophisticated optimization platform and many years of experience collaborating with semiconductor companies, but naturally has a limited sales reach outside of South Korea. Leveraging the global SiMa.ai network for joint customer acquisition and pilot projects offers Nota AI significant leverage for international expansion. For both parties, the partnership reduces go-to-market costs and shortens the path to commercialization.

Furthermore, this partnership sends a clear signal to potential customers and investors: those who invest in the SiMa.ai ecosystem automatically gain access to best-in-class model optimization. Those using NetsPresso can deploy their optimized models on the most powerful embedded edge chip on the market. This flywheel argument – ​​the more customers, the stronger the ecosystem; the stronger the ecosystem, the more customers – is a classic characteristic of successful platform strategies.

What this partnership means for industry

From a competitive strategy perspective, the alliance can be understood as a response to a clear market trend. The convergence of hardware and software in the AI ​​value chain is not a coincidence, but a structural necessity. Large chip companies like Nvidia learned this lesson early on and built their value not least through the CUDA ecosystem – hardware can only be used to its full potential when the software layer is perfectly aligned with it. In the edge AI segment, where resources are significantly scarcer, this hardware-software stack is even more critical.

It is telling that Nvidia acquired OctoML (now OctoAI) for an estimated $165 to $250 million and took over Red Hat Neural Magic in January 2025 – both players in the field of model optimization and compression for edge deployments. The market is sending a clear signal: software optimization is not a commodity, but a strategic differentiator. Nota AI and SiMa.ai are responding to this trend with a partnership rather than an acquisition, which gives both companies greater flexibility.

For industrial customers in manufacturing, logistics, autonomous vehicles, and smart infrastructure, this partnership translates into concrete benefits: increased computing power at the device, lower energy consumption, shorter deployment cycles, and ultimately, lower total cost of ownership. In an economic environment where the return on investment of AI investments is increasingly scrutinized, this is not a marginal advantage, but a decisive one.

Whether the partnership can reach its full potential ultimately depends on three factors: first, the technical quality of the SDK integration, which must go beyond mere marketing compatibility; second, SiMa.ai's sales execution in joint customer acquisition; and third, the ability of both companies to quickly transform real-world pilot projects into scalable production products. The announcements are promising – the proof lies in the deployment.

☑️ Our business language is English or German

☑️ NEW: Correspondence in your native language!

Konrad Wolfenstein

I and my team are happy to be available to you as your personal advisor.

You can contact me by filling out the contact form here or simply call me at +49 7348 4088 965. My email address is : [email protected]

I'm looking forward to our joint project.

☑️ SME support in strategy, consulting, planning and implementation

☑️ Creation or realignment of the digital strategy and digitization

☑️ Expansion and optimization of international sales processes

☑️ Global & Digital B2B trading platforms

☑️ Pioneer Business Development / Marketing / PR / Trade Fairs

The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business - Image: Xpert.Digital

Xpert.Digital is a data-driven B2B industry hub led by Konrad Wolfenstein . The company acts as an external, quasi-in-house solution for industrial partners, closing operational gaps in marketing, content, and sales – without requiring additional resources on the client side.

More information here:

• The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business