Morgan Stanley Report: Top 5 AI Giants to Spend $1.4 Trillion on Capital Expenditure by 2028

Written by: Rita

Tide Guide

Morgan Stanley significantly raised its capital expenditure forecast for mega-cap companies, projecting that the combined CapEx of the five largest firms—Microsoft, Google, Amazon, Meta, and SpaceX—will reach approximately $1.2 trillion and $1.4 trillion in 2027 and 2028, respectively. By 2028, available computing capacity is expected to grow from around 30 GW in 2025 to nearly 120 GW, a fourfold increase. In this report, Morgan Stanley explicitly named Meta as its top pick, citing five underpriced market options, with the most notable being its API business; the Muse Spark 1.1 model is priced 30%–86% lower than competitors, generating approximately $8 billion in revenue per 100 MW of computing power. The ultimate winner in the computing arms race won’t be the one who builds fastest, but the one who sells best.

Morgan Stanley employs a bottom-up cost model to estimate the construction cost per GW of data centers, considering two dimensions: in-rack IT hardware (chips, memory, CPUs, networking) and out-of-rack costs (building materials, electrical and mechanical systems). Due to rising memory prices and inflation in out-of-rack costs, the cost per GW has been increased by approximately 20%. Coupled with GPU/ASIC supply allocation constraints, Morgan Stanley has raised its total capital expenditure forecast for the five major hyperscalers to approximately $1.2 trillion and $1.4 trillion for 2027 and 2028, respectively. Meta’s capital expenditure for 2027 and 2028 has been raised by 29% and 22% to $225 billion and $250 billion, respectively; Amazon’s total company capital expenditure has been raised by 15% and 29% to $308 billion and $318 billion, respectively; Google had previously made substantial upward adjustments.

Morgan Stanley expects the combined available computing capacity of the five major hyperscalers to reach nearly 120 GW by 2028, a fourfold increase from approximately 30 GW in 2025. AWS will have the largest capacity in 2028 at 35 GW, followed by Google at 31 GW. Meta’s capacity is projected to grow from approximately 3.5 GW at the end of 2025 to 14 GW in 2027 and 21 GW in 2028. Of Meta’s new capacity additions in 2026 and 2027, 55% and 90%, respectively, will come from self-built infrastructure, with the remainder sourced from third-party leases.

Morgan Stanley conducted a detailed analysis of Meta’s API monetization opportunities. Meta’s latest AI model, Muse Spark 1.1, is now open to third-party API access, priced at $1.25 per million input tokens and $4.25 per million output tokens—30% to 86% lower than comparable market offerings. Morgan Stanley built an API revenue model: allocating 100 MW of GB300 computing power to API services could generate approximately $8 billion in revenue and an EPS increment of about $1.90 per share, equivalent to roughly 6% upside to EPS by 2028. On the paid user side, Morgan Stanley assumes that 25% of Meta’s 15 million advertisers—approximately 4 million—pay around $200 per month, also generating about $8 billion in annual revenue. Morgan Stanley noted that Meta only needs to allocate 100 MW out of its total capacity of approximately 21 GW by end-2028 to support this business, leaving substantial remaining capacity available for further monetization.

Amazon AWS is expected to achieve revenue growth rates of 40% and 36% in 2027 and 2028, respectively. Morgan Stanley forecasts EPS of $11.53 and $15.05 for 2027 and 2028, with a target price of $330. Google is projected to add the largest increase in computing capacity, with new additions of 9 GW and 11 GW in 2027 and 2028, respectively. Morgan Stanley maintains an overweight rating on both companies.

The deciding factor in the AI compute arms race is shifting from "how much you build" to "how much you sell." For the past two years, the market focused on who hoarded the most GPUs; now, attention turns to who can convert compute into monetizable revenue streams. Meta received the highest score in this report—not because it built the fastest, but because it has the most diverse monetization pathways among the five: core advertising, subscription revenue, NeoCloud compute leasing, API model calls, diffusion models, and business agents. Muse Spark’s aggressive pricing strategy is itself a strategy to maximize compute utilization. Idle compute is worthless; low pricing at scale at least covers marginal costs and builds an ecosystem. If previously you selected investments based solely on capital expenditure and capacity planning, you now need to add a new dimension: who has already set up sales channels while building.

Disclaimer

This article is a compilation and interpretation by Chaoxiang Research of a third-party brokerage research report (Morgan Stanley, July 12, 2026). The ratings, price targets, earnings forecasts, and related judgments cited herein are the views of the brokerage’s analysts and represent only the position of their respective institution; they do not reflect the views of Chaoxiang Research nor constitute any investment advice.

The market carries risks; make decisions independently. This article should not be used as a basis for buying or selling any securities.