Microsoft Unveils Maia 300 AI Chip to Boost Infrastructure and Reduce Nvidia Reliance
Microsoft has unveiled Maia 300, its third-generation AI chip built on a chiplet architecture to rival Nvidia. Alongside massive infrastructure investments, the tech giant is restructuring its Israeli R&D center.
Microsoft continues to invest heavily in independent artificial intelligence infrastructure with the introduction of Maia 300, the company's third-generation AI chip. According to recent plans, the tech giant may deploy more than one million units of the new processor next year, aiming to support its internal artificial intelligence services and expand processing capabilities.
Advanced Architecture and Production
According to reports, the new chip is designed to handle complex AI tasks, including code generation and enterprise business operations for users. Previous generations of the processor already power Microsoft's internal models alongside OpenAI models offered through the Azure cloud platform.
A standout feature of the Maia 300 is its chiplet architecture. Unlike Nvidia's approach, which relies on a large central processing die alongside memory components, Microsoft utilizes a chiplet design that integrates several specialized dies connected together. The primary processing components are manufactured using TSMC's advanced 2-nanometer process, while supplementary components for data transfer utilize a more cost-effective 3-nanometer technology.
Energy Efficiency and Performance
Power consumption and energy efficiency are also central to the hardware design. Maia 300 server racks are projected to house 96 chips and consume approximately 167 kilowatts, compared to 142 kilowatts for Nvidia's GB300 racks, which contain 72 graphical processing units. Microsoft claims the new system delivers roughly 40% more effective processing per megawatt compared to leading GPU systems on the market. Rani Borkar, president of hardware systems and infrastructure at Azure, described the objective as delivering "more intelligence per watt, per dollar, and per byte of memory."
The launch aligns with a broader industry effort by cloud and artificial intelligence providers to develop proprietary silicon, reduce reliance on Nvidia, and tailor hardware specifically to internal service requirements. Microsoft plans to begin rolling out gigawatt-scale deployments next year, amid forecasts of capital expenditures exceeding $175 billion for the current calendar year.
Restructuring at the Israel Development Center
Parallel to its infrastructure investments, Microsoft implemented organizational changes at its research and development center in Israel. Approximately 50 employees, primarily from the Copilot Studio group developing AI agents, were summoned for pre-termination hearings as part of a global corporate restructuring. Company estimates suggest that some employees are expected to remain within the R&D center, and most will be offered opportunities to transition to alternative roles internally.
"We remain deeply committed to our teams in Israel while aligning our global resources to meet strategic priorities in artificial intelligence infrastructure and software development." — Microsoft Spokesperson
Microsoft currently employs approximately 3,000 R&D personnel across six sites in Israel, with the largest facility located in Herzliya. The reductions coincide with leadership changes at the Israeli center, where Ohad Yassin is scheduled to assume the role of general manager in early November, succeeding Michal Braverman-Blumenstyk, who is stepping down from the position.