Key Points
-
Nvidia is poised to introduce a dedicated inference chip at its GTC later this month.
-
The company was at risk of losing market share to competitors developing more energy-efficient solutions.
- 10 stocks we like better than Nvidia ›
Nvidia (NASDAQ: NVDA) has been one of the undeniable winners of the artificial intelligence (AI) revolution. The company’s graphics processing units (GPUs) quickly became the gold standard for training AI models. However, competition has been ramping up as the deployment of AI enters the next phase of adoption — that of inference — when AI models do the jobs for which they were trained. Rivals have been working furiously to develop customized solutions for AI inference, which has been a source of concern for Nvidia investors.
Reports have emerged that Nvidia is tackling the issue head-on, with plans to release a specialized chip specifically designed to speed the process of AI inference — and it could be a game changer.
Will AI create the world’s first trillionaire? Our team just released a report on the one little-known company, called an “Indispensable Monopoly” providing the critical technology Nvidia and Intel both need. Continue »
Image source: Nvidia.
At its GPU Technology Conference (GTC) later this month, Nvidia is poised to launch a newly designed AI chip that addresses the unique needs of AI inference, according to a report in The Wall Street Journal. The next-generation processor will incorporate a new chip developed by AI start-up Groq. This focus on inference will represent “a major shake-up to its business that is poised to reset the AI race,” according to the report.
While Nvidia’s GPUs are unmatched for AI training, competitors have been working to develop more energy-efficient processors. These include Amazon‘s Inferentia 2 chips, which the company claims are 30% to 40% more energy-efficient than Nvidia’s GPUs, making them an attractive option for companies on a budget. Alphabet says its Ironwood Tensor Processing Units (TPUs) offer higher performance per watt than Nvidia GPUS.
By designing a processor specially tailored to the needs of inference processing, Nvidia is responding to the calls for more energy-efficient offerings. Perhaps as importantly, it also expands the company’s existing market opportunity, at a time when developers are increasingly focused on the transition to inference. Nvidia controls an estimated 92% of the data center GPU market, according to IoT Analytics. However, the company was at risk of losing some of that share as its customers sought less costly alternatives.
OpenAI will be one of the first users to deploy the nascent chip, as it has been seeking more energy-efficient alternatives. The start-up has committed to 3GW of capacity from Nvidia, marking a major win for Nvidia’s initial foray into an inference-centric chip.
GTC is scheduled to kick off on Monday, March 16, so this could well be the highlight of Nvidia’s AI-centric conference.
The popular n