[Edaily Reporter Yun Junghoon ] #Nota Inc., a company specializing in on-device AI (personalized artificial intelligence) and model lightweighting, is set to fully join the proprietary AI semiconductor ecosystem of Amazon Web Services (AWS), the world’s leading cloud provider. Having collaborated with top-tier global companies across all hardware sectors—including SamsungElectronics(005930), Arm, and FuriosaAI—Nota Inc. plans to aggressively expand its business into the cloud infrastructure sector through this partnership with AWS.

On the 7th, Nota Inc. announced that it has obtained AWS Authorized Partner status and launched its “NetsPresso®-based model tuning service,” which maximizes the performance of enterprise customers’ AI models on AWS’s in-house designed AI-specific chips.

Amid a sharp rise in AI infrastructure costs for enterprises due to the recent acceleration in the adoption of high-performance large language models (LLMs), this service is expected to provide a practical breakthrough in “cost efficiency” for companies looking to adopt AWS’s high-performance, low-cost AI chips.

“AWS’s In-House Chips” Emerge as a Rival to NVIDIA… Nota Inc. Takes Charge of “One-Stop” Optimization

The core of this collaboration lies in significantly lowering the barriers to adopting “Trainium,” an AI training-specific chip, and “Inferencia,” an inference-specific chip—both designed in-house by AWS to reduce its reliance on NVIDIA.

While AWS’s in-house semiconductors boast world-class value for money and energy efficiency, enterprise customers lacking specialized AI optimization expertise have faced technical challenges in porting and tuning their models to this environment.

Nota Inc. leverages its flagship platform, “Netpresso,” to provide one-stop support for the entire process—from proof of concept (PoC) and diagnostics to model porting, lightweighting, and optimized deployment across different hardware. Netpresso possesses unique lightweighting and deployment automation technologies that can reduce the size of AI models by up to 90% while maintaining existing accuracy.

In fact, Nota Inc. has already demonstrated its overwhelming technical prowess by tuning a high-performance language model with 32 billion parameters in AWS Trainium and Inferencia environments, reducing the model size by 68% while limiting accuracy loss to less than 1%.

AWS’s Trainium-related business is growing into a multi-billion-dollar market, with quarter-over-quarter growth of 150% driven by increasing adoption by big tech companies such as Anthropic, Apple, and Databricks. Nota Inc.’s new service is expected to serve as a catalyst for rapidly absorbing this exploding demand for AWS AI chips.

From Mobile to the Cloud… Leaping Forward as an All-Around Partner for “AI Lightweighting”

Nota Inc. has built a track record of major global partnerships across the entire hardware ecosystem where AI model optimization is essential—ranging from mobile (SamsungElectronics) and semiconductor IP design (Arm) to data center AI accelerators (FuriosaAI) and edge devices (Mobilint).

This partnership with AWS and the launch of dedicated tuning services signify that Nota Inc.’s converged technological capabilities now fully cover the megascale cloud infrastructure domain, extending beyond the edge device level.

Chae Myeong-soo, CEO of Nota Inc., stated, “AWS AI chips offer outstanding price-performance, and Nota Inc. supports model tuning so that customers can fully leverage the performance of these AWS chips in their own models.” He added, “We will deliver tangible cost-efficiency benefits to enterprise customers considering the adoption of AWS AI chip-based infrastructure.”