The advent of artificial intelligence (AI) and its mass adoption in the tech world has caused a paradigm shift that many companies are feeling -- and NVIDIA Corp. (NVDA 4.43%) has been one of the greatest beneficiaries of these developments. Researchers quickly learned that the same functionality that graphics processing units (GPUs) offer for rendering images in video games makes them ideal for AI applications. The ability to process vast quantities of data at lightning speeds made GPUs critical to early AI research.
NVIDIA has consistently dominated the market for high-end processors, making it the obvious choice for training AI systems. Search giant Baidu (BIDU 0.51%) has one of the most advanced AI programs in China, and has long been a user of NVIDIA GPUs. The companies recently announced a far-reaching AI partnership covering everything from cloud computing to self-driving cars.
That's why it came as something of a surprise when Baidu also announced a collaboration with NVIDIA rival Advanced Micro Devices (AMD 1.71%) "on optimizing software for AMD Radeon Instinct GPUs in Baidu datacenters."
The NVIDIA DGX-1, the world's first deep-learning AI supercomputer in a box. Image source: NVIDIA.
Is the honeymoon over?
While this is AMD's first serious foray into the AI space, NVIDIA has been aggressively courting the market for years. As a result, NVIDIA's data-center revenue has increased 174% year over year, on average, in each of the past five quarters. With a lack of any serious competition, NVIDIA positioned itself in the majority of large data centers running AI applications.
Its customers are a veritable Who's Who of companies in the big tech world, and its dominance continues today. NVDIA GPUs provide the foundation for all the top cloud computing services, and it reaped the significant financial rewards that came with its early lead.
Taking the fight to NVIDIA
AMD has been aiming for a piece of the high-end GPU market with the release of its new Radeon Instinct line of GPUs, which are geared directly toward high-performance computing and a wide range of machine learning and deep learning applications. This collaboration with Baidu gives AMD some much-needed credibility and establishes the company as a viable competitor in the burgeoning space of AI against a much more established rival.
As my colleague Harsh Chauhan pointed out, tests performed on Baidu's open-source DeepBench deep-learning benchmarking tool showed that AMD's Vega GPUs already compare favorably with NVIDIA's Telsa P100 offering. It remains to be seen if that will hold up with the release of NVIDIA's upcoming Tesla Volta V100, which is scheduled to make its debut later this year.
The opportunity is staggering
According to a report by Markets and Markets, the AI chipsets market is expected to grow at a compound annual growth rate of 62.9% over the next five years, to $16 billion by 2022.
All in all, a collaboration between Baidu and AMD doesn't spell immediate trouble for NVIDIA, and its GPUs will still have a prominent position in Baidu's data centers. NVIDIA has had the field to itself for some time, while others were playing catch-up. Baidu also has a history of partnering with all the top players. With the improvements to its latest high-end GPU's designed specifically for AI applications, AMD just got a foot in the door.
This is definitely a win for AMD, giving it an official debut in the nascent field of AI, which could open the door for more companies to adopt AMD processors. This could, in turn, put pressure on NVIDIA's data-center business, which has been the growth engine behind the company's recent gains. AMD still has a long way to go, however, before it could unseat the reigning champ.
