Microsoft announced its next-generation custom AI chip on 13 October. The company intends to deploy over 1,000,000 Maia 300 chips during the coming year.
Representing the third generation of custom silicon from Microsoft, the Maia 300 features a chiplet design using 3-nanometer technology for data movement functions. Each rack containing 96 chips consumes approximately 167 kilowatts of power.
These units will provide roughly 40% more useful computing per megawatt compared to leading GPU-based systems. Microsoft plans to initiate multi-gigawatt deployments of the Maia 300 starting next year while expecting capital expenditures to exceed $175 billion this calendar year.
Safety advocates argue that AI currently faces fewer requirements than milk and suggest global rules requiring an "emergency brake" and victim notification for rogue agent breaches. Legal requirements should also include mandatory incident reports from companies, according to these safety claims.
Britain is positioned to lead AI regulation through the AI Security Institute established under former prime minister Rishi Sunak, which receives pre-deployment access to major models via voluntary agreements. AI Minister Kanishka Narayan stated Britain would consider regulating advanced models if the current system proved insufficient.
"It's basically delivering more intelligence for every watt, every dollar, and every byte of memory by optimizing across the entire system," said Rani Borkar, president of Azure hardware systems and infrastructure at Microsoft.
Rani Borkar, president of Azure hardware systems and infrastructure at Microsoft
