Анализ Morgan Stanley Research: 1,6 трлн долларов капитальных затрат на ИИ — на ИИ приходится четверть роста S&P 500

Coincamps · 2026-10-21T08:13:02.000Z

Капитальные затраты в размере 1,6 трлн долларов соответствуют примерно 5% выручки компаний S&P 500 и 25% роста выручки. В отчёте от 20 октября 2026 года Morgan Stanley суммировал денежные расходы и затраты на аренду гиперскейлеров и новых облачных провайдеров, получив 1,58 трлн долларов за 2027 год и 1,81 трлн долларов за 2028 год. Согласно прогнозу отчёта, рост капитальных затрат на ИИ замедлится примерно со 100% в 2026 году до около 60% в 2027 году и примерно 14% в 2028 году. Рынок обеспокоен замедлением роста в 2028 году. Расчёты Morgan Stanley показывают, что рост выручки компаний S&P 500 достигнет 8,6% в 2027 году, а без учёта прямой выручки от ИИ составит 6,6%. Согласно отчёту, около 65% капитальных затрат 2027 года придётся на компании S&P 500, крупнейшим бенефициаром станет NVIDIA.

$1.6 trillion in capital expenditure corresponds to approximately 5% of S&P 500 revenue and 25% of revenue growth. In its October 20, 2026 report, Morgan Stanley summed up the cash and lease expenditures of hyperscalers and emerging cloud vendors, arriving at $1.58 trillion for 2027 and $1.81 trillion for 2028. The report projects that AI capex growth will decelerate from approximately 100% in 2026 to around 60% in 2027 and roughly 14% in 2028.
The market is concerned about the slowdown in 2028. Morgan Stanley's calculations show that S&P 500 revenue growth will reach 8.6% in 2027, remaining at 6.6% after excluding direct AI revenue. The report tracks approximately 65% of 2027 capex flowing to S&P 500 companies, with NVIDIA as the largest beneficiary.
Lease expenditure is a new variable added to this report. Hyperscalers have already disclosed over $1.1 trillion in lease commitments that have not yet commenced; these commitments do not fall under the capex line item on the cash flow statement. Morgan Stanley incorporates lease capex into cash capex, estimating the lease portion at $195 billion for 2027 and $246 billion for 2028.
Lease Expenditure is a New Variable
2027 cash capex stands at approximately $1.4 trillion, reaching $1.58 trillion when leases are included. 2028 cash capex is around $1.57 trillion, rising to $1.81 trillion with leases added. Lease expenditures amount to $195 billion in 2027 and $246 billion in 2028, primarily driven by Oracle, Microsoft, Meta, Amazon, and Google.
Uncommenced lease commitments reached $1.118 trillion in Q2 2026. These break down as follows: Oracle $288 billion, Microsoft $329 billion, Meta $279 billion, Amazon $137 billion, and Google $85 billion. Economically, these leases resemble debt-financed capex, bypassing GAAP cash flow statements. Morgan Stanley estimates them using three metrics: signed but uncommenced leases, annual incremental capacity, and the lease spend ratio per GW.
Capex growth is slowing. From approximately 100% in 2026, it drops to around 60% in 2027 and roughly 14% in 2028. However, the absolute scale continues to rise. Approximately $700 billion in incremental AI spend is added between 2026 and 2027. Morgan Stanley tracks that about 65% flows into the S&P 500.
AI Contributes One-Quarter of Revenue Growth
In 2027, customer revenue from hyperscalers and emerging cloud vendors accounts for 5% of total S&P 500 revenue, driving approximately 25% of year-over-year growth. S&P 500 revenue growth stands at 8.6%, remaining at 6.6% after stripping out direct AI revenue. Morgan Stanley believes S&P 500 broadening is underway, with growth remaining robust even without AI investment.
Of the $1.6 trillion in capex, approximately $1.0 trillion flows into S&P 500 revenue, with NVIDIA capturing about $690 billion. NVIDIA accounts for two-thirds of the S&P 500 absorption share, representing roughly 1.6% of the index's 2027 sales. The remaining $339 billion flows to other S&P 500 companies.
AI names in the S&P 500 see sales growth of 28% in 2027, compared to 5% for non-AI companies. Profit growth stands at 25% for AI firms and 4% for non-AI firms. AI names contribute 78% of the S&P 500's 2027 profit growth, accounting for 37% of the index's net income. The Magnificent 7 comprise 70% of AI cohort sales and 75% of their profits.
NVIDIA Captures Two-Thirds
NVIDIA generates approximately $690 billion in revenue from hyperscalers and emerging cloud vendors in 2027. Morgan Stanley estimates NVIDIA's share of hyperscaler capex: SPCX at roughly 79%, and AWS at about 29%. SPCX and AWS combined account for 54% of NVIDIA's hyperscaler revenue, while the remaining 46% comes from Microsoft, Oracle, Meta, and Google.
NVIDIA's data center revenue reaches $630.9 billion in 2027, with the hyperscaler segment contributing $337.5 billion, or 53.5%. Morgan Stanley believes NVIDIA trades at 15x FY28 EPS, making it reasonably valued; stock price appreciation can be driven by earnings growth even without multiple expansion.
Other beneficiaries include Broadcom, AMD, Micron, Samsung, and SK Hynix. AI chips, memory, networking, and ODMs capture the bulk of capex. Exposure to data centers for industrial goods and power suppliers is increasing.
A 2028 Slowdown Does Not Signal a Peak
With capex growth at 14% in 2028, the market fears AI investment may be peaking. Morgan Stanley views 2027 as a pivotal year for adopters, as AI productivity opportunities expand for both enterprises and individuals. Forward margin expectations improve most noticeably for companies with strong AI adoption and high pricing power.
The risk lies in power bottlenecks. Morgan Stanley's $1.6 trillion 2027 capex assumption assumes every ordered chip can be powered and run. Their Powering AI model indicates that after accounting for time-to-power solutions, the U.S. still faces a net shortfall of 32GW by 2028 and approximately 72GW by 2029. If commissioned capacity fails to keep pace with chip deliveries, the capex risk becomes a timing mismatch.
Memory, networking, optical modules, and ODM servers are most sensitive to the power wall. Power solution providers are relative beneficiaries. Capex may shift overseas, as power constraints in the U.S. drive hyperscalers to build facilities abroad.
Leadership among AI names shifts from pure enablers to mixed enabler/adopters and adopters. S&P 500 sales growth hits 9% in 2027, or approximately 7% excluding AI investment, which still aligns with nominal GDP and historical sales growth regression. AI spending incrementally drives growth, and index broadening has fully unfolded.
Whether the 2028 slowdown will be amplified by power bottlenecks is the key variable to watch next.
Disclaimer
This article is a compilation and interpretation by TechFlow Research of a third-party broker research report (Morgan Stanley, October 20, 2026), combined with publicly available market information. The ratings, target prices, earnings forecasts, and related judgments cited herein reflect the views of that broker's analysts, representing only the stance of their respective institutions, and do not represent the views of TechFlow Research nor constitute any investment advice.
Markets carry risks, and decisions must be independent. This article should not serve as the basis for trading any securities.
Join the official Coincamps community:
X: https://x.com/coincamps
Telegram: https://t.me/coin_camps