AnswersPublished18 sources
Why Power and Data-Center Capacity Could Constrain the AI Buildout
Morgan Stanley’s reported estimate is a 32 GW U.S. data center power shortfall through 2028 even after mitigation, or 34% of projected demand.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Morgan Stanley’s reported estimate is a 32 GW U.S. data center power shortfall through 2028 even after mitigation, or 34% of projected demand.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What does Morgan Stanley’s report “What Does the Power Shortfall Mean for Semis?”, first published on October 5 and resurfaced by Indian out. Article summary: Morgan Stanley’s central warning is that AI expansion is becoming constrained by the ability to finance, house and power chips—not simply acquire them. That creates a distinction between strong demand for AI semiconducto. Topic tags: general, general web, documentation, news. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts wi
The AI buildout may be constrained less by whether companies can buy advanced chips than by whether they can finance, house and power them. Morgan Stanley estimates that U.S. data-center developers could face a 57-gigawatt shortfall before mitigation and a net gap of about 32 GW—34% of projected demand—through 2028 after accounting for additional power solutions. 10
16
That is a forecast of infrastructure constraints, not proof that demand for AI chips has disappeared. The distinction matters: a chip can be ordered before the data center, electricity supply and funding needed to put it to work are ready. 3
49
The bottleneck extends beyond electricity
Power is only one piece of the deployment challenge. Data centers need sites and facilities as well as financing to build and equip them. October reporting on Morgan Stanley’s research describes access to electricity, financing and data-center infrastructure as increasingly important to the AI race. 3
18
That creates a gap between buying computing hardware and bringing it online. The Economic Times reported Morgan Stanley’s view that more semiconductors are being built than there are powered locations to deploy them. That is a reported characterization of the bank’s thesis, not an independently verified statement from Nvidia. 50
The competitive implication is that companies with stronger access to capital and infrastructure may be better positioned to secure usable computing capacity. Reporting on the analysis warns that infrastructure constraints could make it harder for smaller companies to keep pace with financially stronger technology groups. 3
4
Why Nvidia and Broadcom may be less exposed
Morgan Stanley reportedly sees Nvidia and Broadcom as relatively insulated from the power crunch, with no material risk to their 2027 forecasts. Reuters attributed that view to the companies’ visibility into chip placement, customer expansion across regions, and coordination among data centers, chip suppliers and the power supply chain. 49
That is a relative advantage, not immunity from delays. It suggests that the largest platform suppliers may have more visibility into where their chips can be deployed, while companies supplying supporting components can be more exposed to changes in data-center schedules.
Reuters reported that delays to AI deployments could affect makers of memory, optical and other secondary chip components. If a facility is not ready to operate, orders for equipment that supports its computing systems may be pushed back too. 49
Overseas growth does not erase the U.S. power gap
Morgan Stanley lowered its projected U.S. share of global computing capacity from 60% to 55%, while keeping its estimate of the U.S. power shortfall at 32 GW for 2026–2028. The October 16 report says overseas expansion is not expected to fully offset that domestic shortfall. 13
A lower U.S. share of global capacity is not the same as resolving U.S. power constraints. Building more capacity abroad may help meet global computing needs, but it does not make U.S. projects power-ready.
What the figures do—and don’t—show
The 57-GW and 32-GW figures describe different stages of the estimate: the larger shortfall is before mitigation, while the smaller net gap accounts for additional power solutions. The reported net figure is approximately 34% of projected demand through 2028. 10
16
The available reporting supports the central point—that power and deployment infrastructure could constrain the AI buildout and affect chip suppliers unevenly. It does not establish every more specific figure raised in discussion, including projected capacity-addition shares, a 14-GW shift of demand to Asia, or a possible 68-GW shortfall by 2029. Those claims should not be treated as verified by the sources cited here.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Why Power and Data-Center Capacity Could Constrain the AI Buildout"?
Morgan Stanley’s reported estimate is a 32 GW U.S. data center power shortfall through 2028 even after mitigation, or 34% of projected demand.
Sources
news.futunn.comWhat do power shortages mean for the semiconductor industry?
caifuhao.eastmoney.com摩根斯坦利10月5日最新报告：AI的最新瓶颈变了，2028年美国缺电32GW....
bignewsnetwork.comAI race may depend more on access to power ...
reuters.comNvidia, Broadcom shielded as AI power crunch hits chip supply chain, says Morgan Stanley