Undoubtedly, NVIDIA is dominating the generative AI industry in both the software and the hardware realm. Its rival AMD, on the other hand, has been trying several ways to come up in both the realms. On the hardware front, it has been trying to maintain the lead with MCM design, which NVIDIA recently decided to adopt as well. But when it comes to software, AMD has made a huge leap. 

That leads us to AMD making a recent bet on Nod.ai, an open source AI software firm. “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,” said Vamsi Boppana, senior vice president, AI Group at AMD. 

There has been an urgent need for open source GPU architecture for a very long time. All the hardware companies have been trying to build alternatives and close the gap with NVIDIA, which includes AMD and Intel.

For AMD, the deal might be the easiest and cheapest way forward as well. Founded in 2013 by Anush Elangovan and Harsh Menon, Silicon Valley-based Nod.ai, had raised $20 million in its last funding. According to reports, the AI software company might just be purchased by AMD, which is currently valued at $36.5 million. 

Elangovan, one of the co-founders has worked on the first Google ARM Chromebook and has worked as the foundation of the ChromeOS team at Google. Menon, the CTO of Nod.ai, was an early employee at Zee.Aero, and has an MS in Aerospace Engineering, and is said to have worked on flying cars.

Betting on the open source horse

Nod.ai has been known for developing a portfolio of tools and systems for boosting AI applications on AMD hardware. The team at Nod.ai has agreed to develop leading software for Instinct data centres accelerators, the consumer grade Ryzen AI processors, and Radeon GPUs. 

The best part about the company is its open source approach. AMD is reportedly also too interested in working on open source solutions for lowering “the barriers of entry for customers through developer tools, libraries and models,” as it said in its AI strategy announcement. This is in contrast to what NVIDIA has been doing with its software made for GPUs. 

For the longest time, NVIDIA’s compute unified device architecture (CUDA) has been the biggest moat for the company. The only problem is that it is closed source and only works for NVIDIA GPU workloads. Though people have been finding several solutions to work around this restriction, CUDA still remains the best on NVIDIA GPUs, and since everyone is using its GPUs, the moat becomes even bigger. 

Arguably, the biggest moat for hardware companies is software. Bryan Catanzaro, VP of applied deep learning research, NVIDIA, has stated in an interview that even though NVIDIA is known as a hardware company, “many people don’t know this, but NVIDIA has more software engineers than hardware engineers.” Thus, even though CUDA is closed source, it is still free to use. 

But AMD is hell-bent on open source, and Nod.ai’s Elangovan, agrees. “Our journey as a company has cemented our role as the primary maintainer and major contributor to some of the world’s most important AI repositories, including SHARK, Torch-MLIR and OpenXLA/IREE code generation technology,” he said. Added to all this is AMD’s Radeon open compute (ROC), the partnership might actually prove to be an alternative to NVIDIA’s moats.

The trillion dollar rival

The software bet has been going on at AMD for some time now. In August, the company also announced the acquisition of Mipsology, a French AI startup, which has also been a long-standing AMD partner and developing AI software for the chip maker, similar to Nod.ai.

In August, Boppana wrote, “The team will help develop our full AI software stack, expanding our open ecosystem of software tools, libraries, and models to pave the way for streamlined deployment of AI models running on AMD hardware.”

Both the acquisitions, Nod.ai and Mipsology, highlight AMD’s dedication towards challenging NVIDIA’s monopoly in the AI market. Nod.ai team members are also going to join the Mipsology and AMD AI group, which has around 1,500 employees at the moment. 

Finally, it seems like AMD might be able to crack the moat that NVIDIA has in the AI market.

The post AMD’s Attempt to Break NVIDIA’s CUDA appeared first on Analytics India Magazine.