Morgan Stanley Releases 2027 AI Capital Expenditure Flow Chart: Nvidia Alone Captures Two-Thirds of S&P 500 Benefits

Morgan Stanley has for the first time mapped the full flow of 2027 AI capital expenditure, revealing a highly concentrated revenue structure: of the hyperscalers' $1.58 trillion in spending, about two-thirds of the S&P 500 benefit goes to Nvidia alone, while the remaining 499 constituents together receive only about $339 billion.

Morgan Stanley strategists Nick Lentini, Mike Wilson and more than twenty industry analysts jointly published the report "Tracking the AI $1.6 Trillion Money Flow: Mapping the Revenue Created by the Capex Boom," the first to trace hyperscaler capital expenditure line by line to specific companies' income statements. The report's core conclusion: AI capital expenditure will directly drive about 25% of S&P 500 sales growth, and AI-related companies will contribute 78% of the index's earnings growth in 2027.

The report also reveals a long-overlooked hidden risk: hyperscalers have disclosed more than $1.1 trillion in lease commitments not yet commenced. This "synthetic capital expenditure" does not appear under the capex line of any cash flow statement, yet is economically equivalent to debt financing. Morgan Stanley credit strategists expect 2027 to be the year with the largest free cash flow shortfall for hyperscalers, when debt financing pressure will rise significantly.

The $1.6 Trillion Panorama

At the heart of the Morgan Stanley report is a Sankey diagram that traces the nine major buyers—Alphabet, Amazon AWS, Microsoft, Meta, the lease pool, SpaceX, Oracle, CoreWeave and Nebius—and their combined $1.58 trillion in 2027 spending, layer by layer, to semiconductors and servers, networking and optics, data center infrastructure, storage and international suppliers, ultimately landing on specific tickers.

Among the nine buyers, semiconductors and servers are the largest single spending category, absorbing $672 billion, the vast majority of which flows to Nvidia. The second-largest category is "international" suppliers at $511 billion, covering Samsung, SK Hynix, TSMC-related infrastructure and Taiwanese ODM manufacturers, none of which are S&P 500 constituents.

The report also notes that the map as a whole maps $1.6 trillion in capital expenditure to about $1.7 trillion in supplier revenue; the difference stems from distortions in customer splits by private emerging cloud providers and bitcoin miners, timing differences, and supplier price increases not yet reflected in capex forecasts. Morgan Stanley's interpretation: the next round of capex upgrades is already embedded in supplier models.

Nvidia Takes Two-Thirds Alone, the Other 499 Split the Remainder

Of the roughly $1.03 trillion in revenue the S&P 500 absorbs from AI capital expenditure, 67% flows to Nvidia—about $690 billion—meaning 44% of the entire $1.58 trillion capex budget lands on one company's income statement. Excluding Nvidia, the entire AI capex boom contributes only 1.6% to S&P 500 sales.

A rough breakdown: about $700 billion goes to Nvidia, about $340 billion is shared by the other 499 S&P constituents, and another roughly $550 billion flows outside the index to Korean memory makers, Taiwanese ODMs and various private companies.

At the individual stock level, Morgan Stanley examined the share of each company's 2027 revenue tied to AI infrastructure buildout. Nvidia leads at 91.7%, followed by Erock (EROC) at 91%, Vertiv (VRT) at 90%, Applied Opto (AAOI) at 83%, Broadcom (AVGO) at 81% and Bloom Energy (BE) at 80%. Even Caterpillar (CAT) has 21% of its revenue tied to the AI capex trend; Seagate (STX) and Western Digital (WDC) both reach 60%.

This structure means that for at least a dozen large companies, "AI capex" and "revenue" are nearly synonymous—a boon on the way up, and on the way down any capex revision will translate directly into an earnings revision, with no other business to buffer it.

The $195 Billion Hidden Liability

One of the report's most important methodological adjustments is incorporating leases into capex statistics. Using companies' disclosed signed-but-not-yet-commenced leases, analyst forecasts for new capacity, and the lease capex ratio per GW of new capacity, Morgan Stanley adds $195 billion to 2027 capex and $246 billion to 2028.

This raises total 2027 AI capital expenditure from $1.39 trillion (cash capex) to $1.58 trillion, year-on-year growth of about 60%; for 2028 it rises from $1.57 trillion to $1.81 trillion.

In terms of lease scale, Oracle leads at $57 billion, followed by Microsoft ($45 billion), Meta ($37 billion), Amazon ($28 billion) and Alphabet ($27 billion). As of the second quarter of 2026, the pipeline of signed-but-not-yet-commenced leases had swelled to $1.118 trillion, more than doubling in just two quarters, with Microsoft ($329 billion), Oracle ($288 billion) and Meta ($279 billion) the main signatories.

Morgan Stanley's qualitative judgment: leasing data centers is economically identical to debt-financed capital expenditure, merely bypassing the GAAP capex line.

Debt: The Underestimated First Link

Morgan Stanley credit strategists Fernanda Lima and Vishwas Patkar expect free cash flow at hyperscalers and neoclouds to hit its most negative level in 2027, with the funding gap even larger if lease payments are included.

Hyperscalers have issued about $255 billion of unsecured bonds year to date, roughly 30% of their estimated 2026 capital expenditure, led by Amazon ($98 billion) and Alphabet ($81 billion), while blended yields have risen 100 to 180 basis points since the start of the year, with Oracle's 10-year bond spread widening the most, by 180 basis points.

Morgan Stanley's stress test shows that if AA-rated hyperscalers issued $400 billion of bonds at a 7% rate, interest coverage would fall from 30.7x to a still-comfortable 11.9x; but if Oracle and SpaceX raised $120 billion at an 8.5% rate, their combined interest coverage would plunge from 4.0x to 1.8x.

The report explicitly states that neoclouds are the most dependent on debt financing, "relying almost entirely on debt issuance to support their capex ambitions." Days before the report was published, CoreWeave's share price had fallen sharply, and the company corresponds to $47 billion of spending in Morgan Stanley's capex map.

Slowing Growth and the "Power Wall"

Morgan Stanley expects AI capex growth to slow sharply from about 60% in 2027 to about 14% in 2028. Against this backdrop, the report screens for names that retain pricing power when capex growth slows; power equipment dominates, with memory, hard drives and cooling equipment also included.

Power supply is another structural constraint. Morgan Stanley's own "AI power supply" model shows that even accounting for various rapid power solutions (covering 34% of the 97GW needed), the US still faces a net shortfall of 32GW before 2028, widening to about 72GW before 2029. The report characterizes this as a timing mismatch rather than a collapse in demand, but notes that memory, networking, optics and ODM construction are most sensitive to power supply pauses.

On downside risk, Morgan Stanley is unusually blunt: there is currently a "fear of missing out on procurement" mentality in the market, with buyers rushing to buy at high prices today because they expect it to be more expensive tomorrow; once supply catches up with demand, this logic will reverse. The report cites the 2022 NAND memory cycle as an example—shipments fell only about 5%, but prices fell 20% to 30%—and concludes: "We believe the market is currently underestimating the risk of simultaneous declines in volume and price."

Morgan Stanley's report provides the market for the first time with a dollar-by-dollar map of where AI capex flows, clearly showing that about 25% of S&P 500 sales growth and 78% of earnings growth depend on a $1.6 trillion capex budget—and within that budget, two-thirds of the index-level benefit is concentrated in a single company, with a growing share supported by leases that are not counted as capex and by debt that only becomes visible upon refinancing.