The Gaudi 3 AI chip. Image: Intel
Amid mounting losses at its foundry business, Intel hopes the new AI chip – to be used by the likes of Dell and HPE – will give it a boost in the AI race.
Intel has unveiled a new AI chip called Gaudi 3 that it claims has better performance and energy efficiency than its Nvidia counterpart, the popular H100.
At is Vision 2024 event in Phoenix, Arizona yesterday (9 April), Intel introduced the Gaudi 3 accelerator aimed at enterprise generative AI and announced a spate of new customers as it looks to position itself at the front of an increasingly saturated global AI space.
The US chipmaker claims Gaudi 3 delivers on average 50pc better inference and 40pc better power efficiency than Nvidia H100 – the leading AI chip behind Nvidia’s meteoric rise – at a “fraction” of the cost.
Intel CEO Pat Gelsinger said that innovation is advancing at an “unprecedented” pace thanks to the chips that power it, and that every company is “quickly becoming an AI company”.
“Intel is bringing AI everywhere across the enterprise, from the PC to the data centre to the edge,” he said. “Our latest Gaudi, Xeon and Core Ultra platforms are delivering a cohesive set of flexible solutions tailored to meet the changing needs of our customers and partners and capitalise on the immense opportunities ahead.”
Unlock insights and drive innovation with the #IntelGaudi 3 #AI accelerator. Combined with the latest #IntelXeon processors, the Intel Gaudi 3 accelerator delivers best-in-class performance to scale GenAI and meet demanding LLM workloads. Learn how. https://t.co/j4zomqChqE pic.twitter.com/2mSE1Jy9Hd
— Intel Business (@IntelBusiness) April 9, 2024
According to Intel, Gaudi 3 will power AI systems with up to “tens of thousands of accelerators” connected through the common standard of Ethernet. It is poised to deliver a “significant leap” in AI training and inference for global enterprises looking to deploy generative AI at scale.
The company has some industry heavyweights in its clientele, including original equipment manufacturers such as Dell, Hewlett Packard Enterprise, Lenovo and Supermicro. The chip will be made available to them this quarter.
This news comes in the context of mounting operating losses faced by Intel’s foundry business. The division had revenue of more than $18.9bn in 2023, dropping roughly 31pc from the previous year. The business had operating losses of nearly $7bn last year, compared to $5.2bn in 2022.
Despite this, Intel said last week that its goal is to be the world’s second-largest foundry by 2030.
Google also revealed its latest AI chip yesterday as major tech companies race to make the most advanced AI offerings.
These Axion chips, which are custom central processing units (CPUs) based on technology created by UK-based chip giant Arm, are designed for data centres and will be available to Google Cloud customers later this year. It already powers YouTube ads among other Google services.
Find out how emerging tech trends are transforming tomorrow with our new podcast, Future Human: The Series. Listen now on Spotify, on Apple or wherever you get your podcasts.
