AI Capital Flow Chart: Nvidia Takes 690 Billion, Remaining 499 Companies Share 340 Billion

Morgan Stanley has, for the first time, mapped out a comprehensive capital expenditure flow chart for AI spending in 2027, revealing a highly concentrated profit structure: of the $1.58 trillion spent by hyperscale cloud providers, about two-thirds of S&P 500 beneficiary revenue flows to Nvidia alone, while the other 499 component stocks together receive only about $339 billion.

Morgan Stanley strategists Nick Lentini, Mike Wilson, and more than 20 industry analysts jointly issued the report "Tracking $1.6 Trillion in AI Capital Flows: Mapping the Revenue Created by a CapEx Boom," which, for the first time, traces every dollar of the hyperscalers' capex to specific company revenue statements. The core finding: AI capital expenditure will directly drive around 25% of S&P 500 sales growth, with AI-related companies contributing 78% of S&P Index profit growth by 2027.

The report also exposes a long-overlooked hidden risk: hyperscale cloud providers have disclosed more than $1.1 trillion in pending lease commitments, this "synthetic capex" not being reflected in any cash flow capital expense line, yet economically equivalent to debt financing. Morgan Stanley's credit strategists estimate that 2027 will be the year with the largest free cash flow gap for the hyperscalers, at which point debt financing pressure will rise significantly.

$1.6 Trillion Panoramic Map

The core of the Morgan Stanley report is a Sankey diagram that traces the total $1.58 trillion 2027 spending of nine top buyers—Alphabet, Amazon AWS, Microsoft, Meta, lease pools, SpaceX, Oracle, CoreWeave, and Nebius—layer by layer to semiconductors and servers, networking and optics, data center infrastructure, storage, and international suppliers, eventually down to specific stock tickers.

Among these nine buyers, semiconductors and servers are the largest single expenditure, absorbing $672 billion, with most of it going to Nvidia. The second largest is "international" suppliers, totaling $511 billion, covering Samsung, SK Hynix, TSMC infrastructure, and Taiwan ODM vendors, none of which are S&P 500 constituents.

The report also notes that the whole map matches $1.6 trillion in capex to about $1.7 trillion in supplier revenues; the difference comes from interference such as spending split by private emerging cloud providers and bitcoin miners, timing differences, and vendor price increases not yet reflected in capex forecasts. Morgan Stanley interprets this as: the next round of capex upward revisions is already embedded in the supplier models.

Nvidia Takes Two-Thirds, Remaining 499 Companies Share the Leftovers

Of the approximately $1.03 trillion in revenue that S&P 500 absorbs from AI capital spending, 67% flows to Nvidia, totaling about $690 billion—equivalent to 44% of the entire $1.58 trillion capex budget hitting a single company's income statement. Excluding Nvidia, the contribution of the entire AI capex boom to S&P 500 sales is only 1.6%.

Rough breakdown: about $700 billion goes to Nvidia, around $340 billion shared by the remaining 499 S&P members, and some $550 billion flows outside the index, toward Korean memory manufacturers, Taiwan ODMs, and various private enterprises.

At the single stock level, Morgan Stanley examined what proportion of each company's 2027 revenue is tied to AI infrastructure construction. Nvidia leads at 91.7%, followed by Erock (EROC) at 91%, Vertiv (VRT) 90%, Applied Opto (AAOI) 83%, Broadcom (AVGO) 81%, and Bloom Energy (BE) at 80%. Even Caterpillar (CAT) has 21% of its revenue related 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 capital expenditure" and "revenue" are almost synonymous—rising when spending is up, and any cut to capex directly translating into profit adjustments, with no other business segments as a buffer.

$195 Billion in Hidden Liabilities

One of the report's most important methodological corrections is factoring leases into capex statistics. Morgan Stanley uses disclosures of signed but not yet activated leases, analyst forecasts for added capacity, and the capex-to-lease ratio per GW of new capacity to add $195 billion in incremental capex for 2027 and $246 billion for 2028.

This lifts 2027 AI capex from $1.39 trillion (cash capex) to $1.58 trillion, with a year-over-year growth of about 60%; and for 2028, from $1.57 trillion to $1.81 trillion.

On lease volume, Oracle leads with $57 billion, followed by Microsoft ($45 billion), Meta ($37 billion), Amazon ($28 billion), and Alphabet ($27 billion). As of Q2 2026, the pipeline of signed but not yet started leases has swelled to $1.118 trillion—doubling in just two quarters, with Microsoft ($329 billion), Oracle ($288 billion), and Meta ($279 billion) as the primary signatories.

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

Debt: The Most Underestimated Link

Morgan Stanley credit strategists Fernanda Lima and Vishwas Patkar forecast that hyperscale and emerging cloud providers' free cash flow will reach its most negative in 2027, and when lease payments are factored in, the funding gap will be even larger.

So far this year, hyperscale cloud providers have issued about $255 billion in unsecured bonds, around 30% of their estimated 2026 capex; Amazon ($98 billion) and Alphabet ($81 billion) are leading issuers. Comprehensive yields have climbed 100–180 basis points since the start of the year, with Oracle's 10-year bond spreads widening by the most, to 180 basis points.

Morgan Stanley's stress tests indicate that if AA-rated hyperscalers issued $400 billion in bonds at 7% rates, interest coverage ratios would fall from 30.7x to a still-comfortable 11.9x; but if Oracle and SpaceX raise $120 billion at 8.5%, their joint coverage ratio plunges from 4.0x to just 1.8x.

The report plainly states that emerging cloud providers rely most on debt financing, "almost entirely dependent on debt issuance to support their capex ambitions." Just days ahead of the report's release, CoreWeave's share price already plunged, and the company's capex in the Morgan Stanley map corresponds to $47 billion.

Growth Slowdown and the "Power Wall"

Morgan Stanley anticipates that AI capex growth will slow sharply from around 60% in 2027 to just 14% in 2028. Against this backdrop, the report screens for names with pricing power even as capex growth slows, with power equipment dominant, and storage, HDD, and cooling gear also included.

Power supply is another structural constraint. According to Morgan Stanley's proprietary "AI power supply" model, even with rapid supply solutions (covering 34% of the required 97GW), the U.S. will still face a net shortfall of 32GW before 2028, rising to about 72GW by 2029. The report characterizes this as a timing mismatch rather than demand collapse, but notes that storage, networking, optics, and ODM builds are most sensitive to power interruptions.

Regarding downside risks, Morgan Stanley's wording is unusually blunt: a current market mentality of "fear of missing out on procurement" prevails, with buyers paying high prices today due to expectations of even higher costs tomorrow. Once supply catches up with demand, this logic will reverse. They cite the 2022 NAND memory cycle as an example—shipments down only about 5%, but prices plummeted by 20–30%—and conclude: "We believe the market is underestimating the risk of both volume and price declining simultaneously."

Morgan Stanley's report is the first to provide the market with a dollar-by-dollar tracing map of AI capex flows, clearly showing that about 25% of S&P 500 sales growth and 78% of profit growth depend on a $1.6 trillion capex budget—of which, at the index level, two-thirds of the benefit concentrates in one company, and an ever-increasing proportion is sustained by leases uncounted as capex and debt requiring refinancing to show up.