Microsoft’s new AI chip makes it less reliant on Nvidia’s GPUs
Microsoft has announced Maia 300, its third-generation in-house AI chip, which it plans to make the main system for running AI responses across its products and cloud services. The move could reduce its reliance on Nvidia and AMD GPUs while supporting newer workloads such as long-context reasoning and multi-step AI agents.
The chip uses a chiplet design built on TSMC’s N2P process, separating computing, memory control and input-output functions. Microsoft says it expects 40 per cent more tokens per megawatt than leading GPU systems, and up to seven times the rack-scale token throughput of Maia 200 in a 96-accelerator configuration. Production rollout is planned for 2027; Maia 200 already handles workloads including Copilot and GitHub Copilot.
- Microsoft plans to roll out its Maia 300 AI chip in 2027.
- It expects 40 per cent more tokens per megawatt than leading GPU systems.
- The chip is intended to reduce reliance on Nvidia and AMD.
New here? Start with this
Artificial intelligence systems need powerful specialised hardware to run. Microsoft and other large technology companies currently rely on processors made by Nvidia and AMD, meaning they must pay external suppliers for the chips their AI services require.
To reduce costs and dependence on these suppliers, some technology companies have begun developing their own AI chips internally. Microsoft has been working on this for several years and has already started using an earlier version of its own chip to power some of its AI services.
Building and controlling these chips matters because it influences how a company's AI systems develop and operate. For Microsoft, developing its own hardware would reduce costs and provide independence from outside suppliers.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Those favouring the move argue that vertical integration allows Microsoft to optimise hardware precisely for its unique workloads and strategic requirements, whilst reducing vulnerability to supplier pricing and supply-chain pressures. For a company of Microsoft's scale, the investment required is proportionate and well-justified by the potential gains in efficiency, cost reduction, and long-term independence.
The case against
Sceptics argue that Nvidia's deep expertise, proven ecosystem, and sustained research investment represent formidable competitive moats that newcomers cannot quickly overcome. They contend Microsoft should concentrate its engineering efforts on software and services, where its genuine advantages lie, rather than venture into capital-intensive semiconductor manufacturing where dominant incumbents have structural advantages.