AMD: Meta Adopts Its Cloud Chips to Support New Artificial Intelligence Strategy
On Tuesday, June 14, local time in the United States, chip giant AMD announced that it will ramp up production of its new artificial intelligence chip MI300X in the fourth quarter to challenge market leader Nvidia. Meanwhile, Meta is adopting AMD's cloud chip to support its new artificial intelligence strategy.
With 192 GB of memory, the MI300X chip exceeds all of Nvidia's current chips, AMD CEO Lisa Su said at a conference in San Francisco. For processing large AI systems such as ChatGPT, memory is a key performance metric for the chip.
Soo said, "There is no doubt that AI will be a key driver of chip consumption in the foreseeable future." She added that customers will get sample chips in the third quarter and production volumes will ramp up by the end of the year.
Soo also highlighted a system that integrates eight MI300X chips into a single computer to compete with similar products from Nvidia.
AMD also revealed that it has begun shipping a large number of general-purpose CPU chips called "Bergamo" to Facebook parent Meta, among others.
Meta computing infrastructure executive Alexis Black Bjorlin (Alexis Black Bjorlin) also confirmed that the company does use the Bergamo chip. The chip is part of AMD's data center business and is designed to meet the needs of cloud computing service providers and other large chip buyers.
However, it won't be easy to challenge Nvidia's dominance, as it has few too strong competitors. While Intel has competing products with several startups such as Cerebras Systems and SambaNova Systems, by far the biggest sales threat to Nvidia is the in-house chip business of Google and Amazon's cloud computing unit, both of which lease their own custom chips to outside developers.
Nvidia's leadership comes not only from its chips, but also from more than a decade of providing software tools for artificial intelligence researchers and learning to predict what support they will need in chips that take years to design. AMD on Tuesday updated its ROCM software, which competes with Nvidia's Cuda software platform.
Moor Insights & Strategy analyst Anshel Sag said, "Even though AMD is competitive in terms of hardware performance, people still don't believe its software solution can compete with Nvidia."
Meta Vice President Somith Chintala (Soumith Chintala) has helped develop open source software for artificial intelligence. In a recent speech, he said he worked closely with AMD to make it easier for AI developers to use free tools and to shift the "single dominant supplier" of AI chips to other companies, such as AMD.
Chintala added: "It's easy to switch from one platform to another, and you don't actually have to do that much work. In many cases, there's hardly even any work to do."
Nvidia's shares have soared 170 percent so far this year and hold 80 to 95 percent of the artificial intelligence computing market, according to analysts. At the close of trading Tuesday, Nvidia shares rose 3.9 percent to close at $410.22 per share, making it the first chipmaker to close with a market capitalization of more than $1 trillion.
By contrast, AMD failed to impress investors despite releasing an update on its artificial intelligence strategy, with the stock closing down 3.6 percent on the day. Nonetheless, AMD shares have doubled since the start of the year and touched a 16-month high earlier in the day, driven by optimism that the company is investing in artificial intelligence.
With 192 GB of memory, the MI300X chip exceeds all of Nvidia's current chips, AMD CEO Lisa Su said at a conference in San Francisco. For processing large AI systems such as ChatGPT, memory is a key performance metric for the chip.
Soo said, "There is no doubt that AI will be a key driver of chip consumption in the foreseeable future." She added that customers will get sample chips in the third quarter and production volumes will ramp up by the end of the year.
Soo also highlighted a system that integrates eight MI300X chips into a single computer to compete with similar products from Nvidia.
AMD also revealed that it has begun shipping a large number of general-purpose CPU chips called "Bergamo" to Facebook parent Meta, among others.
Meta computing infrastructure executive Alexis Black Bjorlin (Alexis Black Bjorlin) also confirmed that the company does use the Bergamo chip. The chip is part of AMD's data center business and is designed to meet the needs of cloud computing service providers and other large chip buyers.
However, it won't be easy to challenge Nvidia's dominance, as it has few too strong competitors. While Intel has competing products with several startups such as Cerebras Systems and SambaNova Systems, by far the biggest sales threat to Nvidia is the in-house chip business of Google and Amazon's cloud computing unit, both of which lease their own custom chips to outside developers.
Nvidia's leadership comes not only from its chips, but also from more than a decade of providing software tools for artificial intelligence researchers and learning to predict what support they will need in chips that take years to design. AMD on Tuesday updated its ROCM software, which competes with Nvidia's Cuda software platform.
Moor Insights & Strategy analyst Anshel Sag said, "Even though AMD is competitive in terms of hardware performance, people still don't believe its software solution can compete with Nvidia."
Meta Vice President Somith Chintala (Soumith Chintala) has helped develop open source software for artificial intelligence. In a recent speech, he said he worked closely with AMD to make it easier for AI developers to use free tools and to shift the "single dominant supplier" of AI chips to other companies, such as AMD.
Chintala added: "It's easy to switch from one platform to another, and you don't actually have to do that much work. In many cases, there's hardly even any work to do."
Nvidia's shares have soared 170 percent so far this year and hold 80 to 95 percent of the artificial intelligence computing market, according to analysts. At the close of trading Tuesday, Nvidia shares rose 3.9 percent to close at $410.22 per share, making it the first chipmaker to close with a market capitalization of more than $1 trillion.
By contrast, AMD failed to impress investors despite releasing an update on its artificial intelligence strategy, with the stock closing down 3.6 percent on the day. Nonetheless, AMD shares have doubled since the start of the year and touched a 16-month high earlier in the day, driven by optimism that the company is investing in artificial intelligence.
