AI Capex ROI: Morgan Stanley Sees Payoff, But Peak Spending Looms
TREE NEWS reports: Morgan Stanley’s analysis of the mega-cap tech companies’ massive AI infrastructure spending suggests that returns on investment are more promising than feared, with potential ROIC ranging from 25% to 46% depending on the business model. However, the bank also projects that capital expenditure growth will sharply decelerate by 2028, signaling a shift from building capacity to monetizing it.
What Happened
Morgan Stanley released a detailed report examining the capital efficiency of the trillion-dollar AI buildout. The report calculates that model companies offering API services using their own computing infrastructure can achieve a return on invested capital (ROIC) as high as 46%. Hyperscale cloud providers renting out GPUs see an ROIC of approximately 31%, while model companies relying on third-party infrastructure still manage around 25%.
Furthermore, the report highlights that AI investments are increasingly generating their own cash flow. The combined operating cash flow of Amazon, Google, Microsoft, and Meta is projected to rise from $739 billion in 2026 to $1.23 trillion by 2028. Meanwhile, their additional debt financing needs are expected to fall from $238 billion to $90 billion over the same period, meaning new debt will equal only about 7% of operating cash flow by 2028.
Capex growth, however, is nearing a cyclical peak. Data center capital expenditures for the four hyperscalers are forecast to jump from $917 billion in 2026 to $1.47 trillion in 2027 (up ~60% YoY), but then slow to just ~12% growth in 2028.
Market Impact Analysis
- Stocks: The findings suggest that the market’s focus will shift from hardware and semiconductor makers to companies that can convert existing compute into revenue and profit. This could favor cloud platform providers, model developers, and software application layers over pure-play chip suppliers as capex growth decelerates.
- Bonds: With mega-cap tech companies needing less debt to fund AI expansion, their credit profiles may improve, potentially tightening spreads on their corporate bonds.
- Crypto & Commodities: The direct impact is limited, though reduced growth in data center construction could temper demand for commodities like electricity and specialized equipment. For crypto, AI-related narratives may weaken as the focus turns to commercialization.
- Currencies: The shift in capital flows could influence tech-heavy currencies, but the overall macro effect is likely indirect.
Context and Why It Matters
The report underscores a crucial inflection point. Investors have been concerned that AI spending is an unprofitable ‘arms race.’ Morgan Stanley’s analysis counters that by quantifying returns and showing that the largest players can self-fund their investments. The deceleration in capex growth by 2028 suggests that the ‘build-out’ phase is maturing, and the next stage will be judged on AI adoption and revenue generation. As of Q2 2026, ~25% of S&P 500 companies could already quantify GenAI benefits, up from 14% a year earlier, indicating real commercial traction.
For investors, the key takeaway is to watch for evidence of AI monetization in earnings reports, as this will likely drive valuation differentiation. The winners will be those who not only own infrastructure but also have the models and distribution to generate high returns.
