Morgan Stanley says it is keeping its hyperscaler capex forecasts unchanged after the US midterms, at $1.6 trillion in 2027 and $1.8 trillion in 2028, according to its research note of 8 November, as shown in images posted on X by Jonah Lupton. Google is the largest spender in the note’s exhibit, at $375 billion in 2027. The note shifts 5 gigawatts of US data center projects from 2028 to 2029.
For the same five companies our June coverage tracked, Google, Amazon, Microsoft, Meta and Oracle, the note’s exhibit puts 2027 capex at $1.32 trillion. That is about $205 billion, or 18%, above the $1.116 trillion Morgan Stanley figure we reported in June, by our arithmetic.
Business Pill · GPUs Explained
An evergreen Business Engineer explainer on GPUs and AI hardware. It teaches the general idea only and does not discuss Morgan Stanley’s note or capex forecasts.
The key insight: Morgan Stanley’s note keeps $1.6 trillion of hyperscaler capex in 2027 after the midterms. For Google, Amazon, Microsoft, Meta and Oracle, its exhibit puts 2027 at $1.32 trillion, about $205 billion above the $1.116 trillion figure we reported in June, by our arithmetic.
The note, titled “Buy the Drop: AI Investment Opportunities Post the US Midterms”, says “our AI capex projections are unchanged in aggregate, with a partial shift in data center builds from ’28 to ’29.” Lupton’s post says the note runs to 80 pages; the four images show the opening pages and two exhibits.
Morgan Stanley writes that it believes “the market’s reaction to the midterms is overdone”. It expects newly elected governors to leave projects under construction alone, so 2027 projects stay unchanged, but says some 2028 projects in Texas and Ohio will be delayed to 2029 as new requirements for data center developers are written.
The 5 gigawatts it moves are 20% of US projects and 10% of global projects, the note says. It puts the result at US data center additions down 4% in 2028 and up 38% in 2029, and global additions up 9% and 20%. It names the risk for suppliers as “an air-pocket across the tech supply chain in 2028”.
The bank says it still expects the hyperscalers to “spend and build as much capacity as they can”, and that delays “could put downward bias on our ’28 capex”. Exhibit 1 is headed “Our Hyperscaler Capex Forecasts Remain Unchanged at $1.6T/$1.8T in ’27/’28…with Potential Downward Bias in ’28 If Builds Are Delayed”.
Capex by company in Morgan Stanley’s 2026 and 2027 forecasts, US$ billions. Data: Morgan Stanley Research note of 8 November 2026, Exhibit 1, as shown in images posted on X by Jonah Lupton; bar labels read from the exhibit, not independently verified. The 3.3x multiple is our arithmetic.
What Was Posted on X
Morgan Stanley out with a long note this morning (80 pages)โฆ they still expect hyperscaler capex to hit $1.6 trillion in 2027
Jonah Lupton posted four images of pages from Morgan Stanley’s 8 November research note, saying the bank still expects hyperscaler capex to hit $1.6 trillion in 2027. The figures and quotations in this piece come from those images.
Morgan Stanley: The marketโs reaction to midterms is overdone.
Morgan Stanley sees no decrease in their CapEx projections; โfinancing remains available.โ
Similar to Nomuraโs note, they see data center delays rather than a stop (cancellations) to the AI buildout.
AI is a bipartisan trade that is in the best. interests of both parties.
Source: Morgan Stanley Research Note [found by @JonahLupton]— Legend Capital (@legendtrades45) Nov 09 14:35
A second post on X, published at 14:35 UTC on 9 November, summarised the same Morgan Stanley note and attached an image of Exhibit 1. The Update box below sets out what that image shows and which of the post’s lines it does not show.
Update (9 Nov): Morgan Stanley on the midterms. A second post on X, published at 14:35 UTC on 9 November and crediting Jonah Lupton for finding the note, carries one image: Morgan Stanley’s Exhibit 1, headed “Our Hyperscaler Capex Forecasts Remain Unchanged at $1.6T/$1.8T in ’27/’28…with Potential Downward Bias in ’28 If Builds Are Delayed”, with the source line “Company data, Morgan Stanley Research”. For 2028 its bar labels read: Google $400 billion, Amazon $365 billion, Microsoft $323 billion, Meta $250 billion, SpaceX $240 billion, Oracle $93 billion, CoreWeave $47 billion and Nebius $41 billion, against an exhibit total of $1,760 billion, up 7%. Source: Morgan Stanley research note, as shown in an image posted on X; the two smallest segments are matched to companies by their leader lines.
What the post says. The post’s own text summarises the note: that the market’s reaction to the midterms is overdone, that Morgan Stanley sees no decrease in its capex projections, that “financing remains available”, that it sees data center delays rather than cancellations, and that AI is a bipartisan trade in the interests of both parties. Those lines are the poster’s summary and do not appear in the image. The note pages quoted above contain the overdone, financing and delay points; we did not find the bipartisan line in them and have not seen the note’s own wording for it.
Exhibit 1 shows eight companies. For 2027 its bar labels read: Google $375 billion, Amazon $355 billion, Microsoft $269 billion, SpaceX $236 billion, Meta $225 billion, Oracle $97 billion, CoreWeave $46 billion and Nebius $37 billion. The exhibit total is $1,641 billion, up 73% from $951 billion in 2026, and $1,760 billion in 2028, up 7%.
The biggest multiple in the exhibit is SpaceX’s, listed by its SPCX ticker: $72 billion in 2026 and $236 billion in 2027, 3.3 times as much, by our arithmetic. Google nearly doubles in the exhibit, from $196 billion to $375 billion.
The exhibit puts Meta at $144 billion for 2026, inside the $135-145 billion range Meta gave in its third-quarter earnings release. Its $225 billion for Meta in 2027 is 1.55 to 1.67 times that guidance range, by our arithmetic.
From $1.1 Trillion to $1.6 Trillion
In June we reported Morgan Stanley figures of $805 billion for 2026 and $1.116 trillion for 2027 across the five largest cloud companies. The new exhibit puts those five at $819 billion for 2026 and $1,321 billion for 2027, by our arithmetic from its bar labels.
The rise to the $1.6 trillion level came before this note. Fortune reported on 19 October that an earlier Morgan Stanley note estimated hyperscalers and AI cloud providers “will spend $1.58 trillion in 2027”. What the new note adds is that the figure held through the election, while 2028 carries what the exhibit calls potential downward bias.
The note follows other recent readings of AI capex and financing. A Goldman Sachs trading note said tech giants may pull AI capex forward into 2027, and Reuters reported that some lenders balk at AI debt even with backstops. Morgan Stanley’s note says “financing remains available, but the midterms should deepen the quality divide.” On Monday morning, Nvidia shares fell while four AI chip rivals added $376 billion of market value, by our earlier arithmetic.
What GPU Rents Show
Morgan Stanley lists “higher lease rates for compute” among the positives of a longer capex cycle. The spot rental market for Nvidia’s B200 has already moved. Ornn’s B200 rental price index, built from executed compute trades, settled at $9.94 per GPU-hour on 8 November, against $5.66 on 10 August, the first day of its public three-month window, according to Ornn’s data page.
That is a rise of 75.6% in three months, by our arithmetic. Ornn’s monthly table shows the index ending September at $8.05 and October at $9.54. Forecasts for capex measure what the buyers plan to build; the rental index measures what users pay today for chips already running.
The Structural Read
As we read the note, the midterm effect lands on timing rather than on the total: the bank leaves 2027 projects unchanged, moves 5 gigawatts of US builds from 2028 to 2029, and flags potential downward bias for 2028 only.
Morgan Stanley Research, note of 8 November 2026, as shown in images posted on X
“This again puts a higher value on compute that is already built/ functional, which favors the hyperscalers.”
Three Implications
FOR CHIP AND EQUIPMENT SUPPLIERS The note says an air-pocket across the tech supply chain in 2028 could lead to a reduction in pricing power or a build of inventory.
FOR THE LARGEST SPENDERS Morgan Stanley says it still expects the hyperscalers to spend and build as much capacity as they can, and expects them to adjust build locations toward states such as Tennessee and Louisiana.
FOR COMPUTE BUYERS Ornn’s B200 rental index rose from $5.66 to $9.94 per GPU-hour between 10 August and 8 November, 75.6% by our arithmetic.
The framework argues that “the capex race is no longer a corporate spending cycle. It is a sovereign-scale financing problem dressed in tech-company clothing”, and says: “This is not a continuation of the cloud capex cycle.”
As we read it, Morgan Stanley’s note sits on the financing side of that frame: it expects higher AI-related debt issuance in 2027, led by hyperscalers, and says the midterms should deepen the quality divide. The framework predates the note and does not discuss it.
What Is Not Established
We read four pages of the note as images posted on X, not the full document, and we have not seen it published. The per-company figures are bar labels read from Exhibit 1; the smaller segments are matched to companies by leader lines and the stated totals. The June comparison uses the figures as our June piece reported them, and the company set in the new exhibit is wider. Ornn’s index is one venue’s transaction prices, not the whole rental market.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Morgan Stanley’s note keeps $1.6 trillion of hyperscaler capex in 2027 after the midterms and moves 5 gigawatts of US builds to 2029. For the five companies we tracked in June, its 2027 figure is about $205 billion higher than the one we reported then, by our arithmetic.
A note on sourcing. The Morgan Stanley figures and quotations come from four images of the bank’s 8 November research note posted on X by Jonah Lupton; we have not seen the full note and the figures are not independently verified. Meta’s guidance is from its third-quarter earnings release. The B200 rental prices are Ornn’s published index settles. Totals for the five companies, the $205 billion gap, the multiples and the 75.6% rise are our arithmetic.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
