AMD has signed a definitive agreement to acquire Nod.ai, an expert in open-source artificial intelligence software. The financial terms of the acquisition were not disclosed. This move will expand AMD's capabilities in optimizing AI workloads across its broad portfolio of data center, edge and client products.
Founded in 2013, Nod.AI has emerged as a key figure in the AI domain, boasting cutting-edge open-source technologies for next-gen AI systems. Their core strength lies in advanced compiler-based methodologies, a sharp departure from traditional handwritten kernels. Nod.ai is the genius behind the SHARK Machine Learning Distribution, rooted in LLVM, MLIR, and their proprietary tuning.
The SHARK framework, with Nod.ai's pivotal contributions, has been distinguished as one of the fastest runtimes for advanced BERT and Transformer-like models on GPUs.
Nod.ai's personnel roster is a blend of industry veterans and innovators. Anush Elangovan, its Founder and CEO, has a rich history with Google's ChromeOS team, Cisco, and FireEye, to name a few. Alongside, Harsh Menon, the co-founder and CTO, boasts a tenure with Zee.Aero/KittyHawk, focusing on flying cars.
The addition of Nod.ai and its engineering talent will bring industry-leading software technology to AMD that accelerates the deployment of AI models tuned specifically for AMD hardware platforms including Instinct accelerators, EPYC processors, and Radeon GPUs.
Nod.ai's SHARK software reduces the need for manual optimization using compiler-based automation. This enables faster deployment of high performance AI implementations optimized for the advanced compute architectures of AMD's CDNA, XDNA, RDNA, and Zen processor families.
Vamsi Boppana, senior vice president, Artificial Intelligence Group at AMD, highlighted the acquisition's alignment with the company's growth vision in AI. "The acquisition is set to magnify our prowess in providing AI customers with open software, streamlining the deployment of high-performing AI models honed for AMD hardware." Notably, Nod.ai's technologies find extensive applications, from the cloud and edge to a spectrum of end-point devices.
The deal also bolsters AMD's efforts in open-source AI development. Nod.ai leads contributions to key repositories like SHARK, Torch-MLIR, and IREE code generation stacks used by the AI research community.
AMD has invested heavily in recent years in AI-focused silicon designs and is now acquiring software technology and talent to accelerate competitive optimized AI solutions. With the AI market estimated to reach $594 billion by 2032, this acquisition positions AMD to capture more data center and edge AI workloads.
By acquiring Nod.ai's mature software assets and engineering expertise in AI optimization, AMD aims to provide turnkey AI solutions bringing together its own high performance hardware and Nod.ai's tailored software. This further boosts AMD's ambition to be a prime AI silicon and platform provider to rival NVIDIA and Intel.