TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
Morgan Stanley Research Report Analysis: Four Major Cloud Providers Further Increase 2027 Capital Expenditure by 20%

Morgan Stanley Research Report Analysis: Four Major Cloud Providers Further Increase 2027 Capital Expenditure by 20%

2026.08.10
Share
Morgan Stanley clearly stated in the report that hyperscalers' capital expenditure forecasts will continue to face upward pressure.

By: Rita

The four major hyperscalers plus SpaceX will have a combined capital expenditure of $1.15 trillion in 2027.

This figure is 20% higher than market expectations in July. A year-over-year growth rate of 47% means that even in the most optimistic AI narrative, the arms race among cloud giants is still accelerating. However, Morgan Stanley believes that what the market has truly underestimated is 2028. The 2027 consensus of $1.15 trillion has already caught up; the 11% growth expectation for 2028 is the weakest link in the entire forecast curve.

In the weekly report released this week, Morgan Stanley's European Software & Services team used a set of data to redefine the boundaries of compute power investment. Why is capital expenditure still rising? The judgment framework given by Morgan Stanley is: the diversification of the model ecosystem is strengthening compute power demand. Closed-source models, open-weight models, hybrid deployment—more choices mean more demand for computing and storage. Morgan Stanley explicitly stated in the report that capital expenditure forecasts for hyperscalers will continue to face upward pressure.

The following breaks down Morgan Stanley's judgment logic from four dimensions and what this means for the compute power industry chain.

$1.15 Trillion Is Not the Ceiling

Microsoft, Google, Amazon, and Meta, plus SpaceX, will have a total capital expenditure standing at $1.15 trillion in 2027.

Year-over-year growth of 47%, approximately half of the growth rate in 2026. The base for 2026 is already high enough, but 2027 is still expanding at a rate close to 50%. Since July, the 2027 forecasts for just the four cloud vendors (excluding SpaceX) have been raised by 20%. Morgan Stanley Internet Research Head Brian Nowak pointed out in the report that as the number of tokens processed monthly grows exponentially, cloud revenue accelerates, data center commitments expand, and supply chain vendors emphasize accelerating demand and extended visibility, capital expenditure forecasts for hyperscalers will continue to face upward pressure.

This forecast comes from Visible Alpha consensus data as of August 7, 2026. Morgan Stanley has not adjusted its own model, but the consensus numbers themselves are already catching up to Morgan Stanley's judgment. The market has fully priced in the 2027 growth; how long the 11% growth expectation for 2028 can hold is the real question.

Open-Source Models Are Feeding Back into Compute Power Demand

Will open-source models dilute the compute power demand for closed-source models?

Morgan Stanley's answer is no. Global Thematics & Sustainability Research Head Stephen Byrd released a dedicated report this week, dividing the AI model competitive landscape into three scenarios: closed-source wins, hybrid coexistence, open-source reaches frontier performance. Morgan Stanley's judgment leans between scenario two and scenario three. Open-weight models help competition and can accelerate the diffusion of AI; the Jevons paradox is coming true, where efficiency improvements instead stimulate greater demand.

A key data point: 60% of enterprises are already using open-weight models in their tech stack, mainly concentrated in specific scenarios requiring speed, security, or high-frequency calls. What do these models run on? Still compute power. Lower-cost inference means more application scenarios are unlocked, more scenarios mean more tokens, more tokens mean more computing. Morgan Stanley used the cases of Wix and Thomson Reuters to corroborate this logic. Wix's self-developed model significantly reduced AI inference costs, with non-GAAP gross margin rising from near zero at the beginning of the year to about 60% in the second half. Thomson Reuters' proprietary model trained with less than 10% of legal content achieved effects comparable to frontier models, while costing significantly less than third-party models.

Cost declines have not suppressed compute power demand; instead, they make more business segments willing to integrate AI.

Momentum Is Being Transmitted Level by Level from Cloud Vendors

The upward revision of capital expenditure forecasts is not an isolated event.

Morgan Stanley's research report also covers software and service companies such as SAP, IONOS, and Legora. The underlying thread throughout is: investment in AI infrastructure is transforming from financial report numbers of cloud vendors into performance guidance for every company on the supply chain. Data center commitments of cloud vendors are expanding, and demand visibility for component suppliers is extending; this is the micro foundation for Morgan Stanley's judgment that capital expenditure forecasts will continue to move upward.

European web hosting company IONOS was upgraded to an Overweight rating by Morgan Stanley this week, due to customer growth momentum combined with pricing effects; the group's 2027 revenue growth rate is expected to accelerate to about 8.6%. AI product innovation (such as AI phone receptionists) is seen as incremental growth not yet fully reflected in current forecasts. The AI products of a European small and medium-sized cloud service provider are already pulling the growth curve; there is even less reason for the capital expenditure of leading cloud vendors to slow down.

SAP's distributor AI survey provides another corroboration. 33% of distributors reported customers have adopted paid AI, and 53% of customers are still in the pilot phase. 87% of distributors expect AI to increase customers' SAP spending within the next 12 months. 80% of distributors believe SAP's AI migration tools will increase customers' willingness to migrate to S/4HANA. The pull of AI on cloud vendors is not a long-term story; it is already reflected in distributors' order expectations.

2028 Is the Real Weak Link

Back to the figure of $1.15 trillion.

Market consensus has been raised by 20% compared to July, but the core signal conveyed by Morgan Stanley is: this number may still not be enough. The cloud revenue acceleration, expansion of data center land and power commitments, and extended order visibility for supply chain vendors cited in the report are all verified facts, not forecasts. Based on these facts, Morgan Stanley's judgment is that capital expenditure forecasts will continue to face upward pressure.

$1.15 trillion is the current bottom; the future top is far from arriving. Market consensus has left nearly half of the growth slowdown for 2028, but Morgan Stanley believes the support for this assumption is not solid. Data center commitments of cloud vendors have already covered beyond 2028, and supply chain order visibility is extending synchronously; the 11% growth expectation for 2028 is more likely underestimated.

The Morgan Stanley report also touched on a longer-term variable: three world scenarios for open-source models. If open-source models reach frontier performance (Scenario 3), compute power costs will further decline, and the speed of AI adoption will exceed all current linear forecasts. At that time, the 2027 figure of $1.15 trillion may look like just a midway station.

Disclaimer

This article is a compilation and interpretation by TechFlow Research of a third-party brokerage research report (Morgan Stanley, August 10, 2026), combined with a compilation of public market information. The ratings, target prices, earnings forecasts, and related judgments cited in the text are the views of the brokerage analysts, represent only the position of their affiliated institution, do not represent the views of TechFlow Research, and do not constitute any investment advice.

Market involves risks, investment decisions should be made independently. This article should not serve as the basis for buying or selling any securities.