Chipmaker AMD has announced plans to acquire open source machine learning and AI software provider Nod.ai as the chipmaker looks to expand its AI capabilities in a bid to shore up competition against AI chip market leader Nvidia.
The acquisition, whose financial details have not been disclosed, is expected to bring AMD a team that can help accelerate the deployment of AI-based offerings optimised for the company’s Instinct data center accelerators, Ryzen AI processors, EPYC processors, Versal SoCs, and Radeon GPUs, AMD said in a statement.
Nod.ai, which specialises in AI and machine learning, provides open source software for futuristic AI systems, such as generative AI, using advanced compiler-based approaches instead of legacy handwritten kernels.
“We leverage machine learning-based methods (reinforcement learning in particular) for code generation and auto-scheduling of compute and communication over a heterogeneous set of resources (such as CPUs, GPUs, Accelerators etc),” a brief description on Nod.ai’s portal read.
Nod.ai’s acquisition aligns with AMD’s AI growth strategy centered on an open source software ecosystem that lowers the barriers to entry for enterprises through developer tools, libraries, and models, AMD said.
AMD’s main competition in the AI software segment has been Nvidia’s CUDA software stack, which many analysts believe acts as the protective moat for the dominant chipmaker.
“Nvidia is differentiated by its CUDA stack, and ecosystem strategy which ensures that any advanced AI technique that requires GPUs is already pre-supported by Nvidia’s CUDA stack, which has no meaningful competition in the space yet,” said Chirag Dekate, vice president and analyst at Gartner.
Also, since every cloud provider and systems vendor has partnered with Nvidia to deliver AI solutions, developers have been forced to adopt Nvidia-driven software options, further leading to Nvidia's dominance and allowing it to drive pricing and margins, Dekate explained.
Nod.ai’s acquisition could allow AMD to develop an open source alternative to CUDA that could see massive adoption.
“Scale of advanced generative AI models is growing faster than compute capabilities needed to train them. With next-generation models ranging in trillions of parameters, and next-generation models that are multi-modal and multi-model, it is no surprise that leading technology innovators are exploring diverse computational accelerators to accelerate training while reducing the time and cost of training involved,” Dekate said.
AMD said it plans to use talent from Nod.ai to accelerate its ability to advance open source compiler technology.
“The acquisition of Nod.ai is expected to significantly enhance our ability to provide AI customers with open software that allows them to easily deploy highly performant AI models tuned for AMD hardware,” Vamsi Boppana, senior vice president of Artificial Intelligence Group at AMD, said in a statement.
The Mountain View, California-based startup, which was founded in 2013 by Anush Elangovan and Harsh Menon, provides AI repositories, such as the SHARK Machine Learning Distribution that is built on LLVM, MLIR, and OpenXLA’s IREE.
The compiler-based automation software capabilities of Nod.ai’s SHARK software reduce the need for manual optimisation and the time required to deploy highly performant AI models to run across a broad portfolio of data center, edge, and client platforms powered by AMD processors and architectures, the startup said in a statement.
The startup, which boasts investors such as Menlo Ventures, Atlantic Bridge, Walden Riverwood Ventures, Alameda Research, and Sequoia Capital, said it already delivers optimised AI solutions to top hyperscalers, enterprises, and startups.
The deal is expected to close by the end of the quarter, AMD said.