AI spending vs. investor patience
The industry-wide AI build-out is attracting huge capital, but investors are asking whether demand will arrive fast enough to justify the spending — Microsoft is cited as an example of heavy capex that’s testing investor patience. (aol.com) Analysts and research houses have raised their cloud and AI infrastructure spending estimates, yet warn that investment can race ahead of realised returns, which makes execution and contracted revenue crucial for standing up long-lived infrastructure bets. (prismnews.com)
Microsoft is making so much money from artificial intelligence that it says the business is already larger than some of its older franchises, and investors are still getting less patient. On January 28, 2026, the company reported $81.3 billion in quarterly revenue, but its stock fell after cloud growth moderated and capital spending hit $37.5 billion for the quarter. (microsoft.com) (cnbc.com) That $37.5 billion is the price of building the warehouses of the artificial intelligence era. Microsoft told investors that roughly two-thirds of that quarter’s spending went to short-lived assets like graphics processing units and central processing units, and that customer demand still exceeded supply. (microsoft.com) Wall Street usually forgives giant bills when revenue arrives right behind them. What is changing in 2026 is that analysts are no longer treating every artificial intelligence spender as interchangeable, and Goldman Sachs says investors are rewarding companies that can show a clear link between spending and revenue. (goldmansachs.com) That shift is happening while the bills keep rising. Goldman Sachs said on December 18, 2025 that consensus estimates for 2026 capital spending by the biggest artificial intelligence cloud companies had climbed to $527 billion from $465 billion at the start of the third-quarter earnings season. (goldmansachs.com) Bridgewater pushed the number even higher for the biggest four buyers. In a February 23, 2026 Reuters-reported note, the firm said Alphabet, Amazon, Meta, and Microsoft could spend about $650 billion on artificial intelligence infrastructure in 2026, up from $410 billion in 2025. (globalbankingandfinance.com) Microsoft is the cleanest example of why this debate feels tense. In the quarter ended December 31, 2024, it said its artificial intelligence business had passed a $13 billion annual revenue run rate; one year later, it said commercial remaining performance obligation, which is contracted future revenue not yet recognized, had jumped 110% to $625 billion. (microsoft.com 1) (microsoft.com 2) Those are real numbers, but the market is asking a different question: how fast can those contracts turn into cash while depreciation, power, chips, and data center leases are hitting now. CNBC reported on March 31, 2026 that Microsoft had just logged its worst quarter on Wall Street since 2008, with the stock down nearly a quarter for the year. (cnbc.com) This is why “demand” is no longer enough on its own. A cloud company can say customers want more computing capacity, but investors now want to know which part is backed by signed commitments, which part is profitable software, and which part is a very expensive guess about usage two or three years from now. (goldmansachs.com) (microsoft.com) The strange part is that both sides of the story can be true at once. Artificial intelligence demand can be strong enough that Microsoft says it is supply-constrained, and investor patience can still wear thin if each extra dollar of growth requires another wave of servers, chips, and power before margins catch up. (microsoft.com) (cnbc.com) That is the argument underneath the selloff. The market is no longer debating whether artificial intelligence is real; it is debating whether the return arrives in time for the size of the check being written now. (goldmansachs.com) (globalbankingandfinance.com)
