Microsoft Bets on Maia as Banks Cap Risk on $5 Billion AI Loan

Microsoft is scaling homegrown AI silicon as lenders draw a firmer line around infrastructure risk.

🗄️ Microsoft plans gigawatts of Maia chips

Decoded: Microsoft expects to install multiple gigawatts of its own Maia processors in its data centers, with the Maia 300 starting to support production workloads in 2027, Bloomberg reported on Oct. 12. The chips are designed for inference, the stage when trained models answer questions, summarize calls or complete other user tasks.

Why it matters: Bringing more inference silicon in-house gives MSFT another lever over supply, performance and cost as AI usage expands. The key test is whether Maia can reach high utilization across real workloads while reducing reliance on outside accelerators.

💰 Banks limit exposure on a $5 billion AI loan sale

Decoded: JPMorgan and Goldman Sachs are leading a $5 billion financing for AI cloud startup Volta Infrastructure but are not underwriting the $3.7 billion chip-loan portion, Bloomberg reported on Oct. 12. The banks underwrote a separate $1.3 billion term loan as investors demanded higher yields on riskier AI-linked debt.

Why it matters: The split structure shows lenders distinguishing between corporate credit and hardware collateral as AI borrowing grows. For JPM and GS, limiting the chip exposure can protect balance sheets, while higher yields raise the financing hurdle for newer cloud operators.

Stay decoded. See you tomorrow.

— The Get AI Decoded Team