Microsoft Tells Wall Street It's Tripling Data Center Capacity to 38 Gigawatts by 2032 — Because It's Already Turning Customers Away 마이크로소프트, 2032년까지 데이터센터 용량 3배 확대해 38기가와트로 — 이유는 '이미 고객을 돌려보내고 있어서'

The Report: 38 Gigawatts, More Than Tripled, by 2032

On September 10, 2026, Bloomberg News reported that Microsoft plans to more than triple its global data center capacity to 38 gigawatts by 2032. The report spread within hours across Reuters, The Information, Seeking Alpha, Yahoo Finance, TipRanks, GuruFocus, TradingView, Dataconomy, Technology Org, and financial outlets across Asia (조선일보, 아시아경제, futunn, moomoo, NAI500, kaohoon international), all citing the same Bloomberg scoop. What makes this more than a routine capacity-planning update is the reported driver: Microsoft has reportedly been capacity-constrained enough to turn away cloud and AI customer business. TradingView’s headline put it directly – “Microsoft’s AI Crunch Forced It to Turn Away Business, Now It Wants to Triple Computing Power” – and techtimes.com and Korean outlet 아시아경제 both independently framed the story the same way: lost customers due to a server shortage, not proactive over-building.

Why This Belongs in the AI CapEx Conversation

Microsoft is one of the five companies this blog’s AI CapEx coverage tracks directly (alongside Alphabet, Amazon, Meta, and Oracle), and this report doesn’t stand alone – it lands on top of a summer of escalating Microsoft cloud data points. On July 30, Microsoft’s fiscal Q4 2026 earnings showed Azure crossing the $100 billion annualized-revenue mark with growth at a four-year high, and the stock had its biggest one-day gain since 2008, adding roughly $480 billion in market value (Fortune, qz.com). On September 3, The Motley Fool reported Microsoft’s total sales backlog had reached $678 billion, up 84% year-over-year. A 38-gigawatt target by 2032 is the infrastructure-side answer to that demand signal: the backlog and revenue numbers say customers want more Azure and AI capacity than Microsoft can currently deliver, and this report says Microsoft is now committing to build multiples of its current footprint to catch up.

The Pattern This Blog Has Tracked Since July

Unlike Broadcom’s or Oracle’s cases, this Microsoft report isn’t an earnings print with a stock-moving guidance number attached – it’s a leaked/reported multi-year infrastructure plan, so there’s no single-day price reaction to grade the way there was for Oracle’s post-earnings swing. What makes it belong on this list anyway is the scale and the framing: tripling a hyperscaler’s entire global data center footprint is a strategic commitment on the order of hundreds of billions of dollars in cumulative capital spending across the rest of the decade, not an incremental guidance tweak.

Investment Read

The throughline connecting Azure’s $100B/84%-backlog-growth numbers, Oracle’s $664B RPO the day before, and now Microsoft’s 38GW target is that on the demand side, this blog has yet to see a hyperscaler data point this quarter that says AI infrastructure demand is cooling – every read is “we can’t build fast enough,” not “we’re pulling back.” That’s a different signal than the capex-anxiety framing that hit Alphabet and Meta in late July, where the market’s objection wasn’t to the existence of AI demand but to whether the spending would show up as returns. A capacity shortage severe enough to turn away paying customers is about as unambiguous a demand signal as a cloud company can produce. The open question this blog keeps returning to (flagged in the Broadcom and Qualcomm posts) is conversion speed: contracted backlog and long-range targets are not delivered megawatts, and 2032 is six years out. Watch whether Microsoft attaches a nearer-term milestone (a 2027-2028 capacity figure) in its next earnings call, and whether the financing mix for this build-out leans on Microsoft’s own balance sheet or on the debt-heavy, off-balance-sheet structures increasingly used across the sector.

Disclaimer: This post is for informational purposes only and does not constitute investment advice. All investment decisions are the reader’s sole responsibility.

Sources: Bloomberg News, “Microsoft Plans Data Center Push to Triple Its Computing Power” (2026-09-10/11, as reported); Reuters, “Microsoft plans 38 gigawatts of data center capacity by 2032, Bloomberg News reports” (2026-09-10); The Information, “Microsoft, Hurt By Server Shortage, Aims to Triple Cloud Capacity by 2032” (2026-09-10); Seeking Alpha, “Microsoft plans to more than triple data center capacity by 2032” (2026-09-10); TradingView, “Microsoft’s AI Crunch Forced It to Turn Away Business, Now It Wants to Triple Computing Power” (2026-09-11); techtimes.com, “Microsoft Data Center Expansion to 38 GW Triggered by Lost Clients and Capacity Crisis” (2026-09-11); 아시아경제, “‘서버 부족으로 고객 잃어’… MS, 데이터센터 용량 3배로” (2026-09-11); Investing.com Australia, “Nvidia stock edges higher on Microsoft data center expansion report” (2026-09-10); Fortune, “Microsoft’s stock has biggest one-day gain since 2008” (2026-07-30); qz.com, “Microsoft Q4 2026 earnings: Azure tops $100 billion” (2026-07-30); The Motley Fool, “Satya Nadella’s Microsoft Now Has a $678 Billion Sales Backlog, Up 84% Year Over Year” (2026-09-03); this blog’s July 26, July 29-30, September 7, September 9, and September 11 posts on hyperscaler AI capex.