The AI‑driven edge market is gaining momentum as enterprises look to run sophisticated models on power‑constrained devices. In a move that could reshape how manufacturers and smart‑city projects handle on‑device inference, Nota AI announced a strategic partnership with SiMa.ai on March 25, 2026. The collaboration pairs Nota AI’s NetsPresso® model‑compression platform with SiMa.ai’s Modalix™ MLSoC family, aiming to deliver higher inference throughput while keeping energy draw low enough for rugged, real‑world environments.
Why the partnership matters
Edge AI deployments have long wrestled with a trade‑off: squeezing more compute out of a chip typically means higher power consumption, which in turn limits battery life or cooling options. The two firms bring complementary expertise—Nota AI specializes in software‑level model optimization, while SiMa.ai designs hardware and low‑level SDKs expressly for “Physical AI,” the term the company uses for AI that interacts directly with the physical world (e.g., robotics, autonomous vehicles, surveillance).
By integrating NetsPresso®’s compression algorithms with SiMa.ai’s Palette™ SDK, developers can expect model size reductions exceeding 90 % without a noticeable loss in accuracy. In practice, this translates to faster inference on SiMa.ai’s Modalix™ MLSoC, a silicon solution touted for its multimodal capabilities and energy efficiency at the edge. The partnership also opens a joint go‑to‑market channel, leveraging SiMa.ai’s global sales network to surface pilot projects and smart‑city projects.
Technical highlights
– **Model compression meets MLSoC hardware:** Nota AI’s AI optimization technology will be applied to SiMa.ai’s MLSoC product line, aiming to maximize on‑device performance for physical AI workloads.
– **Joint development of industry‑focused solutions:** The companies will co‑create AI applications for sectors such as intelligent transportation systems (ITS), safety monitoring, and security.
– **Broader collaboration across physical AI domains:** Beyond the initial use cases, the partnership intends to explore opportunities in robotics, mobility and other sectors that require low‑latency, high‑throughput edge inference.
- Model compression meets MLSoC hardware: Nota AI’s AI optimization technology will be applied to SiMa.ai’s MLSoC product line, aiming to maximize on‑device performance for physical AI workloads.
- Joint development of industry‑focused solutions: The companies will co‑create AI applications for sectors such as intelligent transportation systems (ITS), safety monitoring, and security.
- Broader collaboration across physical AI domains: Beyond the initial use cases, the partnership intends to explore opportunities in robotics, mobility and other sectors that require low‑latency, high‑throughput edge inference.
The joint effort will focus on three core activities:
- Co‑development and commercialization of on‑device AI solutions – engineering teams from both firms will work together to adapt and fine‑tune models for SiMa.ai’s hardware stack.
- Expansion of the technology partnership – the alliance will evolve beyond a single integration, potentially incorporating additional SDKs and toolchains.
- Customer identification and pilot execution – SiMa.ai’s sales and partner ecosystem will help locate enterprise customers, while technical teams will run joint proof‑of‑concepts.
Targeted applications
One of the first offerings emerging from the collaboration is Nota AI’s generative‑AI‑based video‑intelligence platform, NVA (Nota Vision Agent). NVA will be re‑engineered to run on Modalix™ MLSoCs, positioning it for deployment in:
- Intelligent Transportation Systems (ITS): Real‑time video analytics for traffic flow optimization and incident detection.
- Safety and security: Edge‑based monitoring that can process multimodal sensor inputs without relying on cloud connectivity.
- Robotics and mobility: Low‑latency perception stacks for autonomous platforms that need to make split‑second decisions.
By compressing models to a fraction of their original size while preserving accuracy, enterprises can embed advanced analytics into devices that previously lacked the compute headroom for such workloads.
Market context
Edge AI is increasingly viewed as a cornerstone of the broader AI ecosystem, especially as data‑privacy regulations tighten and latency becomes a competitive differentiator. According to recent analyst reports, the market for edge‑focused machine‑learning chips is projected to surpass $30 billion by 2028, driven largely by automotive, industrial automation and smart‑city initiatives. Solutions that combine aggressive model compression with hardware optimized for multimodal inference could therefore capture a sizable share of that growth.
The partnership also reflects a broader industry trend where software‑centric AI vendors team up with silicon specialists to deliver end‑to‑end stacks. Companies such as NVIDIA, Intel and Qualcomm have pursued similar strategies, but the Nota AI–SiMa.ai alliance is notable for its focus on “Physical AI,” a niche that emphasizes direct interaction with the environment rather than purely cloud‑based services.
Executive perspectives
“We see great significance in the fact that Nota AI’s AI optimization technology, combined with SiMa.ai’s platform, can accelerate our expansion into physical AI. We look forward to working together with SiMa.ai to build on‑device AI solutions that can be practically deployed in real‑world industrial environments,” said Myung‑su Chae, CEO of Nota AI.
“Software optimization is not optional — it is essential for AI models to run reliably in physical AI environments. We are confident that Nota AI’s AI optimization expertise, which maximizes AI model performance for specific hardware environments, will play an important role in realizing SiMa.ai’s physical AI strategy,” added Krishna Rangasayee, Founder and CEO of SiMa.ai.
What this means for developers and enterprises
For engineers building edge solutions, the combined stack promises a more streamlined workflow:
- Single‑toolchain integration: NetsPresso® can handle compression, quantization and deployment steps, while Palette™ SDK abstracts hardware specifics, reducing the need for low‑level code adjustments.
- Reduced time‑to‑market: By leveraging pre‑validated hardware‑software pairings, development cycles can be shortened, allowing enterprises to pilot projects faster.
- Scalable deployment: The ability to shrink models by over 90 % while keeping accuracy intact means a single MLSoC can support multiple concurrent workloads, improving ROI on edge hardware.
Enterprises that have struggled with the cost and complexity of bringing AI to the edge—particularly in sectors like manufacturing, logistics and public safety—may find the partnership’s offerings a viable path forward.
Looking ahead
While the initial focus is on industrial and transportation use cases, both companies have indicated plans to broaden the collaboration into robotics and mobility. As the edge AI market matures, the capacity to efficiently run generative models and multimodal perception pipelines on low‑power silicon could become a decisive factor for vendors competing for enterprise contracts.
The success of the Nota AI–SiMa.ai alliance will hinge on how quickly joint solutions can be demonstrated in real deployments and whether the performance gains hold up across a diverse set of hardware configurations. If they do, the partnership could set a new benchmark for how model compression and specialized MLSoCs are combined to meet the demanding requirements of physical AI.
