摩根士丹利：AI基建豪掷1.4万亿美元，META的“算力账”能回本吗？

TL;DR

• Morgan Stanley estimates that the total capital expenditure of the five major hyperscale cloud providers could reach $1.4 trillion by 2028.
• Costs per GW are being pushed higher by memory, power, and construction; computing capacity may expand from 30GW to 120GW.
• META is listed as the top pick in AI internet; the $775 target price depends on the monetization of APIs, advertising, and subscriptions.

Morgan Stanley has raised its capital expenditure estimates for major hyperscale cloud providers in a sell-side research report, projecting total capex for the five major platforms to reach $1.2 trillion in 2027 and $1.4 trillion in 2028. It continues to name META as the top pick in AI internet, with a target price maintained at $775.

These figures are based on the research report's model estimates and do not equate to official company guidance. Publicly available Morgan Stanley materials have already mentioned that global AI-related infrastructure investment could approach $3 trillion by 2028, with data center capex around $2.9 trillion. The $1.4 trillion figure for the five major platforms is primarily derived from the sell-side's breakdown analysis of the major cloud and internet platforms.

The most newsworthy shift in this report is the continued upward revision of AI infrastructure spending. By 2028, the available computing capacity of major platforms is modeled to approach 120GW, approximately four times the 30GW in 2025. The cost of building a single GW has also been raised. Newer generation platforms like GB200, GB300, and Vera Rubin require more memory, power, racks, and engineering investment.

For investors, the question has shifted from "Will AI giants spend money?" to "How quickly will this spending turn into revenue?" META is positioned at the forefront because it faces higher AI capex pressure while having more direct monetization channels like advertising, consumer applications, model APIs, and subscription tools.

$1.4 Trillion Spending Betting on 120GW of Computing Power

The report raises the capital expenditure expectations for the five major hyperscale cloud providers for 2027 and 2028 by 9% and 10% respectively, to $1.2 trillion and $1.4 trillion. This scope covers AI infrastructure spending by Amazon, Google, Microsoft, META, and related SPCX entities.

Capacity expansion is a primary driver of the upward spending revision. In this model, the available computing capacity of major platforms increases from approximately 30GW in 2025 to nearly 120GW by 2028. Amazon's total capacity is expected to be around 35GW by 2028. Google is projected to add the most capacity in 2027 and 2028. META's capacity is expected to rise from around 3.5GW at the end of 2025 to 14GW in 2027 and 21GW in 2028.

Capital expenditure forecasts for the five major hyperscale cloud providers, totaling $1.4 trillion in 2028, with upward revisions of 9% and 10% for 2027 and 2028 respectively compared to previous estimates.

Available computing capacity increases from ~30GW in 2025 to nearly 120GW in 2028. META reaches 21GW, while Amazon's total is approximately 35GW.

It is important to note definitional differences for META's capital expenditure estimates. In the report's model, META's capex for 2027 and 2028 is raised to $225 billion and $250 billion respectively. Some secondary public reports citing Morgan Stanley's estimates suggest META's total 2027-2028 capex is around $380 billion, which may involve different scopes such as total capex, AI infrastructure, total expenditure, or including off-balance-sheet financing.

These differences do not change the main narrative: AI data center spending continues to pressure free cash flow, depreciation, and short-term EPS, and it will also determine whether future revenue from cloud, advertising, search, APIs, and enterprise tools can materialize. The companies that can convert more computing power into chargeable products will find it easier to justify today's capital expenditure.

Cost Per GW is Increasing; Memory and Power Infrastructure Raise the Stakes

The upward revision in spending is not solely due to "building more data centers," but also because "each GW costs more."

In the report's bottom-up cost model, the construction cost per GW for GB200 is approximately $35 billion, a 16% increase from the previous assumption. For GB300, it is about $39 billion (up 19%). For Vera Rubin, it is roughly $49 billion (up 20%). Google TPU v7 is estimated at around $27 billion, and Amazon Trainium3 at about $21 billion.

Updated deployment costs for GW-scale data centers using GPUs and ASICs. GB200 is ~$35 billion, GB300 is ~$39 billion, and Vera Rubin is ~$49 billion.

Cost pressures come mainly from two areas. The memory share in high-end AI systems continues to rise. Additionally, shell costs for data centers, including power, land, cooling, power distribution, and construction, are also increasing. The report assumes these related costs rise from approximately $10 million/MW to between $11 million and $19 million/MW.

This is also why the spending curve for AI giants is unlikely to decline in the short term. While improved chip supply can alleviate some pressure, factors like power access, racks, construction, skilled labor, and local approvals will continue to extend construction timelines. Some project timelines may be stretched to around three years. The larger the capital expenditure, the faster the revenue side needs to demonstrate a return.

META's Focus Shifts to How AI is Monetized

META is designated as the top pick, primarily because its AI revenue options are more concentrated than most internet companies.

The report breaks down META's potential upside into areas like Meta AI search, new cloud services, API revenue, subscription tools, and advertising upgrades. Collectively, these could contribute approximately $10 to EPS in 2028. Under the base case, META's 2028 EPS is $33.41. If some of these options are realized, there is further upside potential for EPS.

Cumulative contribution of META's five AI upside options to 2028 EPS. Base EPS is $33.41, with total upside from options approximately $10.

This estimate is not entirely consistent with the "four products or catalysts" or "2028 EPS upside of $1 to $3" mentioned in some secondary public reports, and is better viewed as a scenario analysis within this specific report. The portion that ultimately translates into financial statements depends on product adoption rates, pricing power, and utilization rates.

APIs are the most direct channel. On July 9, Meta announced the public preview of the Meta Model API. Third-party information from price tracking agencies like Artificial Analysis shows that the input and output prices for the Muse Spark 1.1 API are $1.25 and $4.25 per million tokens, respectively, which is lower than some leading competitors.

The report's model further hypothesizes that utilizing 100MW of GB300 capacity for APIs, corresponding to approximately 53,300 GPUs with a 75% utilization rate, could generate approximately $8.59 billion in revenue and $640 million in incremental EBIT, contributing about $1.91 to 2028 EPS. This estimate relies on high utilization and sustained demand. Low pricing alone can help acquire customers but cannot guarantee profitability.

Subscription tools are also a potential channel. The model assumes that 25% of META's 15 million advertisers pay approximately $200 per month for tools like business agents and coding assistants. This could contribute about $8 billion in revenue and roughly $2 to 2028 EPS. Ultimately, whether advertisers are willing to pay continuously depends on whether these tools can deliver higher conversion rates, lower production costs, or stronger automation capabilities.

Amazon and Google Benefit, but Revenue Validation Must Follow

Amazon and Google are also major players in this round of capex increases, though they serve more as contextual references in this main narrative.

For Amazon, the report raises its AWS revenue growth outlook, projecting 40% and 36% growth for 2027 and 2028 respectively. It also estimates that the AWS backlog increased by approximately $110 billion quarter-over-quarter in Q2 to roughly $475 billion. As Amazon has not yet released its corresponding official Q2 financial report, this backlog figure should be considered a sell-side estimate. Official documents have confirmed that AWS sales grew 28% year-over-year in Q1 2026, OpenAI has made an additional $100 billion multi-year commitment, and cash capital expenditure continues to rise.

Google's strength lies in its full-stack capabilities combining the Gemini model, TPUs, and cloud business. The report's model shows that Google is expected to add the most capacity among the major platforms in 2027 and 2028. A short-term challenge is that computing resources may still constrain product scaling, especially when search, cloud services, and model APIs simultaneously compete for compute power.

These threads point to the same real-world issue: AI spending has entered the trillion-dollar level, and the market will increasingly ask directly, "How much revenue does each dollar of capital expenditure generate?" Cloud services, AI search, APIs, advertising tools, and enterprise subscriptions will all serve as channels to validate the return on this spending.

Massive Spending Must Navigate Power, Approvals, and Real Demand

This round of upward capex revisions has clear boundaries.

The first constraint is supply. Chips, HBM memory, racks, power access, and skilled labor will all impact the pace of construction. The journey from planning to production for an AI data center must also go through local approvals, grid upgrades, and construction cycles, and cannot follow model assumptions in a linear fashion.

The second constraint is politics and regulation. The significant demand for power, water, and land by large data centers can generate local opposition. Energy policies and local approval timelines may also shift around the 2026 US midterm elections and the November 2028 presidential election.

The third constraint is demand. META's APIs, subscriptions, and advertising upgrades remain upside scenarios. Revenue realization requires real customer payment and sustained usage. Lower prices than competitors can help acquire customers, but long-term profitability will depend on usage volume, gross margins, and the ROI of the tools.

The $1.4 trillion capital expenditure paints a picture of a high-cost growth trajectory. Giants are pre-emptively securing AI computing power, and the market will continue to ask when this computing power will translate into revenue and profit. META's $775 target price is predicated on the gradual realization of AI monetization. The most difficult step is turning the model's EPS upside into cash flow on the financial statements.