Investors in the “Magnificent Seven” stocks are growing more restless about whether heady investments in AI capital expenditures will yield fruitful results.
The analysis: The AI capital expenditures boom is likely to weigh on returns on equity (ROE) for the megacap tech stocks in the coming years, Goldman Sachs strategist Ben Snider warned in a new note.
While ROE for megacap tech stocks is extremely elevated after years of strong cash flow growth, the forward outlook is facing increasing challenges. Consensus estimates suggest each of the megacap tech stocks will experience declining ROEs next year by an average of seven percentage points, Snider said (see visuals below).
The largest ROE declines in 2027 are slated to come from Nvidia (NVDA) and Apple (AAPL).
ROE is a vital metric for investors because it measures exactly how efficiently a company uses shareholders’ capital to generate net income. A high and rising ROE demonstrates that a company can successfully reinvest its earnings to drive highly profitable growth without constantly relying on outside debt or diluting stock. More often than not, this favorable combination supports a higher stock price for a company.
Tracking this percentage allows investors to compare the quality of management and financial health of different companies within the same industry.
This metric is arguably as important as ever in assessing a Magnificent Seven member since all are aggressively investing in AI capital expenditures.
By the numbers: Amazon (AMZN) is now guiding to roughly $200 billion in capital expenditures for 2026, Microsoft (MSFT) has pushed its number up to around $190 billion, Alphabet (GOOG, GOOGL) is sitting in the $175 billion-$185 billion range, and Meta (META) has raised its guidance to somewhere between $125 billion and $145 billion to keep pace with rivals on data centers. Add it all up, and you’re looking at roughly $700 billion -$725 billion in combined spending for 2026 from these hyperscaler bames alone — a jaw-dropping 77% jump from last year’s already-record $410 billion.
Make no mistake, this isn’t a one-year blip. Executives across the tech board are signaling that this AI infrastructure arms race will extend well into 2027 and beyond as these companies race to build out the compute capacity.
Goldman Sachs analysts recently said the Wall Street consensus estimates for 2027 hyperscaler capital expenditures — around $920 billion — are far too conservative.
Instead, Goldman projects that unrelenting demand for AI computing infrastructure and data centers could skyrocket 2027 hyperscaler capital expenditures to at least $1.1 trillion.