Microsoft Maia 300, Amazon-Decart Talks, Navitas-Nvidia Power Board

Custom silicon dominated the AI tape on Oct. 12. Microsoft detailed its Maia 300 chip, Amazon reportedly explored a $7 billion software acquisition, and Navitas showed a power board built with Nvidia.

Key points

• Microsoft unveiled Maia 300 and is targeting over 1 million deployments next year, with a claim of 40% more useful computing per megawatt than leading GPU systems, per The Hindu BusinessLine.
• Amazon is reportedly eyeing a $7 billion acquisition of Decart, software that helps AI models run efficiently across chips, according to Business Standard.
• Navitas is collaborating with Nvidia on an 800 V-to-12 V power delivery board for AI infrastructure, per Manila Times.
• Semiconductors now make up about 15% of U.S. market capitalization, according to GuruFocus.

What does Microsoft’s Maia 300 change for Nvidia?

Microsoft said it will lean more heavily on its own AI chips. The Maia 300 targets more than 1 million deployments next year, and the company claims 40% more useful computing per megawatt than leading GPU systems, as reported by The Hindu BusinessLine. A GuruFocus summary also describes the plan as expanding use of proprietary chips.

The efficiency figure is Microsoft’s own claim, and the sources provided do not include independent benchmarks or details on which GPU systems were compared. Investors should treat it as a company target. The deployment goal, if met, would be meaningful volume for an in-house chip. But the sources do not say how much Nvidia purchasing Microsoft plans to displace, so the effect on Nvidia demand is unclear.

Why would Amazon pay $7 billion for Decart?

Business Standard reports that Amazon is in talks to acquire Decart at a roughly $7 billion valuation. The appeal is software that makes AI models run efficiently across different chips. That would support Amazon’s push for its own Trainium processors against Nvidia, since software compatibility is a major barrier to switching away from Nvidia hardware.

This is a report of talks, not a signed deal. The summary provided does not give terms, timing, or confirmation from either company, so the outcome remains uncertain.

Taken with Microsoft’s announcement, the pattern is clear: two of the largest cloud buyers are investing in alternatives to merchant GPUs, on both the silicon and the software side.

Where does Nvidia still show up in the supply chain?

Navitas said its 24 kW board converts 800 VDC directly to 12 VDC in a single stage. It removes the 50 V intermediate bus converter stage, which the company says improves efficiency and saves board space. The design will be shown in the NVIDIA AI Factory MGX Ecosystem partner area at OCP, per the press release. Power delivery is a growing constraint in AI data centers, and this shows Nvidia’s ecosystem extending to suppliers beyond the chip itself. The efficiency gains are vendor claims.

What else moved the AI narrative?

Meta announced two models, Muse Wonder and Muse Wonder Full, aimed at complex legal and software tasks. The sources give few details on performance or availability.

Elon Musk said Tesla and SpaceX could become strong chip manufacturers, according to GuruFocus, and a separate report says he announced a Texas fab called Terrafab. These are announced ambitions, and the sources show no timeline or funding detail.

On the market backdrop, GuruFocus says semiconductors are about 15% of U.S. market cap. That concentration raises sensitivity to any AI spending doubts.

This site is independent and not affiliated with any company mentioned. Nothing here is investment advice.