Microsoft is preparing to deploy potentially more than one million of its own artificial intelligence chips next year as the technology giant steps up efforts to cut computing costs and reduce its reliance on external suppliers.
The company unveiled its third-generation custom AI chip, Maia 300, on Monday, saying it was designed to power demanding AI workloads, including coding and complex business tasks, across its internal services and Azure cloud platform.
The move shows a growing shift in the global AI race, where success increasingly depends not only on building powerful models but also on running them efficiently at enormous scale.
Microsoft said that Maia 300 could deliver about 40 per cent more useful computing per megawatt than leading GPU-based systems, potentially lowering the energy and infrastructure costs associated with operating AI services.
The chip uses a chiplet-based architecture, combining specialised components manufactured using advanced 2-nanometre and cheaper 3-nanometre technologies from Taiwan Semiconductor Manufacturing Company.
Each Maia 300 rack will contain 96 chips and consume about 167 kilowatts of power, compared with 142 kilowatts for Nvidia’s latest GB300 racks, which house 72 graphics processing units.
The technology company maintains that its efficiency gains come from optimising the entire system rather than relying on raw computing power alone.
The development intensifies competition with Nvidia, whose processors have become central to the AI industry, as cloud providers and AI companies increasingly develop their own silicon to gain greater control over performance, supply and costs.
Microsoft said that it was planning multi-gigawatt deployments of Maia 300 beginning in 2027, following expectations of more than $175 billion in capital expenditure this calendar year.
Meanwhile, the stakes go beyond the chip market. As AI systems become more widely used, the cost of running them at scale is emerging as a critical competitive advantage.
As a result, developing proprietary chips could help Microsoft reduce its dependence on third-party processors, improve energy efficiency and manage the cost of delivering AI services to businesses and consumers.
If its efficiency claims translate into real-world savings, Maia 300 could strengthen Microsoft’s position in the cloud computing market while putting additional pressure on Nvidia and other chipmakers.
In the meantime, the central focus appears to be shifting from who can build the most powerful AI to who can deliver it at the lowest sustainable cost.
