Microsoft Set to Launch Maia 300 AI Chip as It Ramps Up Custom Silicon Strategy
Microsoft's Custom AI Chip Development and Industry Context
Upcoming Launch of Maia 300 AI Chip
Aug 10 (Reuters) - Microsoft is planning to unveil its new Maia 300 AI chip this fall, potentially as soon as next month, The Information reported on Monday, citing people with direct knowledge of the plans.
Background: Maia AI Chip and Competitive Landscape
The company introduced its Maia AI chip in November 2023 but has lagged rivals such as Alphabet and Amazon in scaling up its in-house chip efforts as it seeks to reduce its reliance on Nvidia's costly processors.
Rival Efforts in Custom AI Chips
Google began recognizing revenue from direct sales of its custom AI chips, called Tensor Processing Units, in the quarter ended June, while Amazon has also seen growing adoption of its processors, including its Trainium chips.
Production Plans and Partnerships
Manufacturing with TSMC
Microsoft has been in talks with chipmaker TSMC to secure manufacturing capacity for more than 300,000 units of the chip for delivery in 2027, according to the report. It is also looking to significantly ramp up production and persuade major cloud customers such as Anthropic to adopt the chip.
Official Statement from Microsoft
"Microsoft continues to invest in custom silicon as part of our long-term AI infrastructure strategy. While we don't share production volumes, the figures reported don't reflect the scale of our program," said Andrew Wall, general manager for Microsoft's Azure Maia.
TSMC could not be reached for comment outside regular business hours.
Future Production Goals and Technical Details
Microsoft ultimately aims to secure capacity for more than 1 million Maia 300 chips, though component supplies and ongoing capacity negotiations with TSMC could constrain its plans, according to the report.
It unveiled its second-generation Maia 200 in January, built by TSMC using 3-nanometer technology.
Technical Advancements in Maia Chips
Microsoft packed the chip with a significant amount of SRAM, a type of memory that can provide speed advantages for AI systems handling large numbers of user requests.
(Reporting by Harshita Mary Varghese in Bengaluru; Editing by Anil D'Silva and Pooja Desai)