Morgan Stanley: Amazon, Microsoft, Google's AI spending yields 25–50% returns

Morgan Stanley released a report arguing that Amazon, Alphabet, and Microsoft's elevated artificial intelligence infrastructure spending will generate attractive long-term returns despite rising investor concerns over capital expenditures. The three cloud providers are on track to invest a combined $1.4 trillion（約220兆円） in capital expenditures, and Morgan Stanley projects those investments could produce returns on invested capital of roughly 25% to 50%. The firm attributed these returns to strong demand for AI computing, software models, and cloud services over time.

The report outlined three specific revenue drivers. First, hyperscaler GPU rental services—where the companies lease access to high-performance graphics processing units—could deliver returns on invested capital of 25% to 40%, supported by strong incremental operating margins. Second, AI application programming interface (API) services, which allow developers and businesses to access AI models without building infrastructure themselves, could generate returns exceeding 40%. Third, AI models operating on third-party infrastructure may produce returns of about 25%, though profitability would depend on controlling computing costs and pricing.

Morgan Stanley emphasized that continued investment in chips, software, and AI products remains strategically important as Amazon, Alphabet, and Microsoft compete to expand their positions in the rapidly growing AI infrastructure market. The forecast serves as a counterweight to concerns that the companies' capital intensity will erode shareholder returns, suggesting instead that the scale and durability of AI demand will justify the spending over the long term.

Morgan Stanley's projection addresses a key investor anxiety: whether the massive capital commitments by Amazon, Alphabet, and Microsoft to AI infrastructure will generate sufficient returns to justify the spending. The firm's analysis breaks down the path to profitability into three distinct revenue streams, each with its own return profile. GPU rental services, which allow customers to access high-performance computing resources on demand, carry the tightest margin expectations (25–40%) but benefit from strong incremental operating margins—meaning each additional customer adds profit at a steep rate. API services, which abstract away the infrastructure layer and let developers and businesses integrate pre-built AI models without owning hardware, command the highest return potential (exceeding 40%), reflecting their software-like economics and reduced marginal costs. The third stream—selling or licensing AI models trained on third-party infrastructure—carries more risk because profitability hinges on controlling computing costs and maintaining pricing power; Morgan Stanley estimates roughly 25% returns here. Together, these three levers suggest the companies' AI infrastructure play is not a speculative bet but a reasonably defensible capital deployment strategy, contingent on continued demand for AI services and the firms' ability to compete effectively.

No comments yet. Be the first to share your thoughts!