A significant change is set to happen in the artificial intelligence chip market as Intel has unveiled plans to launch a new artificial intelligence chip called Gaudi 3 to rival Nvidia’s dominance in the market, according to a report from LiveMint.
The updated chip, which will be widely available in the third quarter of this year, aims to enhance performance in two vital areas: training AI systems and running final software. Previous versions of Intel’s Gaudi AI chips have failed to make any progress in market share, and the company is optimistic that the new and improved version of Gaudi will have a positive impact.
What’s new with Gaudi 3?
According to Intel’s assessment, the Gaudi 3 AI chip is considered as an answer to Nvidia’s current H100 accelerator as it can compete with Nvidia’s product in speed and energy efficiency. The chip manufacturer claims that Gaudi 3 will train specific AI models 1.7 times faster and be 1.5 times better at running final software. However, despite Intel’s affirmations, the product is expected to slightly lag behind Nvidia’s upcoming H200 chip in some areas, as revealed by a LiveMint article.
A Reuters article mentioned that Intel stated its new Gaudi 3 chip can train large language models 50% faster than Nvidia’s previous H100 processor. Additionally, it was reported that for certain models tested by Intel, the new smart chip can compute generative AI responses through inference faster than H100 processors.
The Reuters report mentioned that the smart Gaudi 3 chip will be available to server builders like Supermicro and Hewlett Packard Enterprise by the second quarter of 2023, and the chip is a merger of two main processing chips, making it faster than the previous Gaudi version.
Earlier in April this year, Microsoft also revealed that they are working on a specialized AI chip called Athena, aiming for better performance and reducing operational costs compared to externally acquired ones.
Gaudi 3 to Elevate Intel’s Prospects
Gaudi 3 comes in response to the increasing demand for AI services, prompting tech companies to search for specialized acceleration chips. The broader market for corporate spending on productive AI equipment is expected to increase from $40 billion in 2024 to $151 billion in 2027, according to Intel-conducted market research.
Nvidia’s data center revenues exceeded $47 billion in the last 12 months, and analysts estimate the company’s data center revenues will surpass $95 billion in the current fiscal year. Intel aims to capture a significant share of this lucrative market through a new chip called Gaudi 3, as reported by LiveMint.
