Microsoft and Alphabet reported earnings on the same Wednesday afternoon. By the time the after-hours session closed, one stock was up 4.3% and the other was down 7.5% โ and the gap had almost nothing to do with who is winning the cloud race. Both are. The difference was cash.
Microsoft generated roughly $19 billion in free cash flow for the quarter. Alphabet burned $9.1 billion, worse than the prior quarter's shortfall of just under $6 billion. That single line item, more than any growth metric, is what investors are now using to sort the biggest spenders in technology, according to Anurag Rana of Bloomberg Intelligence and Ed Ludlow of Bloomberg Tech, speaking on Bloomberg Businessweek Daily.
"The best performers are those throwing off cash," Ludlow said. "The worst performers in the stock market are those with negative free cash flow."
That framework explains a lot about this earnings season. It also explains why Alphabet's genuinely spectacular cloud numbers โ Google Cloud grew 103% โ did almost nothing to rescue the stock.
Microsoft's quarter had almost nothing to dislike
Microsoft reported adjusted earnings of $4.89 a share on revenue of $92.28 billion, both ahead of estimates. Azure revenue rose 48% excluding currency effects against a 44.8% estimate. Cloud revenue came in at $63.9 billion, also a beat. The one line that came in below expectations was capital expenditure โ and Rana treated that as a gift rather than a warning.
"When you have lower capex and strong cloud growth, there's not much not to like in Microsoft results," he said. He also flagged Microsoft 365 revenue as "really good," reinforcing that the strength wasn't confined to the cloud unit.
Two figures from the release answered the questions Rana said he would have asked on the call. Microsoft's remaining performance obligation โ its backlog of contracted cloud work โ hit $677 billion. Strip out OpenAI, and that backlog is still growing at 25%. The company also booked a $207 million net gain from its OpenAI investment, a three-cent-per-share boost to net income.
Rana said the two things he wants from Microsoft's call are the capital expenditure plan for the next 12 months and any hint about 2027 spending. He doubts management will give a 2027 number outright, but he'll be listening for clues. The second item is how much of the backlog addition comes from non-frontier customers โ ordinary enterprises buying cloud contracts that aren't tied to OpenAI or Anthropic. That metric matters because it shows whether demand extends beyond the model labs and into the broader economy.
On the OpenAI versus Anthropic question, Rana said Microsoft's relationship with OpenAI is far larger, and the most important piece is hosting ChatGPT usage rather than training models. He called inference "a much stickier business" than training, because model companies can cut training budgets in a downturn but ChatGPT usage is unlikely to fall. Microsoft's work with Anthropic, he said, gives customers a choice of models inside products like GitHub Copilot's coding agent, and he expects Microsoft 365 applications to eventually run on third-party models beyond OpenAI and Anthropic. Even then, Microsoft keeps the distribution โ and the customer relationship.
On Microsoft's offer of buyouts to 60,000 employees, Rana did the math that the headline didn't. The last such offer went to 8,750 people and drew a 30% take rate. Microsoft expects a lower take rate this time. At a 25% take rate, roughly 15,000 of 223,000 employees would leave โ about 6.7% of the workforce. "I know those 60,000 looks a lot," he said, but on a detailed read it's about 7.5% of total headcount.
Alphabet's real problem wasn't the cloud โ it was the search engine
Alphabet's cloud unit delivered $30.71 billion in revenue against a $29.1 billion estimate, with growth of 103%. Ludlow called the upside "pretty astonishing" and noted some on the Street had a whisper number above 100% growth. If cloud were the whole story, the stock would have rallied.
But search is still more than half of Alphabet's revenue and the engine of its profit, and it missed. Search and other revenue came in at $63.35 billion against a $64.85 billion consensus โ a shortfall of about $1.5 billion. Google Services revenue was $96.3 billion versus a $98.4 billion estimate. Total revenue excluding partner payouts was $111.21 billion against $111.29 billion expected. Management said in the release that AI is accelerating search query growth. The revenue didn't match that claim.
"Search grew 12%. But search revenues did not grow 12%. What's up with that," Ludlow said.
That's the question Ludlow expects analysts to press management on during the earnings call: whether Alphabet can increase search usage through AI Overviews and AI Mode without damaging the economics of search advertising. He called the issue "existential" and pointed back to a Businessweek cover story a year earlier, reported in part by Julia Love, on Alphabet's effort to adapt search for the next generation of users.
Co-host Tim Stenovec added an anecdote that makes the abstract problem concrete. A small business owner, Christina Stempel of Farmville Flowers, told the hosts the prior week that she had to rebuild her business around AI chatbots because Google search queries were no longer bringing her customers. That's what a monetization problem looks like at ground level โ not fewer searches, but searches that don't lead to the same advertising clicks.
On Gemini, Alphabet said the app has reached 1 billion users, but Ludlow noted the company had already disclosed that figure in August, so it isn't new. The more interesting AI metric was 32 billion tokens processed per minute through model APIs, up from about 22 billion the prior quarter. Ludlow said token usage is a metric you either track or you don't, and it measures AI activity rather than revenue or commercial market share.
Then came the spending. Capex for the quarter was $51.36 billion, below the $55.78 billion estimate, but Alphabet has spent about $132 billion through nine months and guided full-year capex to $195โ205 billion. A year earlier, Ludlow said, Alphabet was spending roughly $22โ25 billion a quarter. The cloud growth and the operating income from it are real, but they're being weighed against that commitment and its effect on cash flow. That's why the search miss can't be read in isolation from the spending โ the market is doing the subtraction in real time.
The Fed cuts meetings, not rates
The Federal Reserve left interest rates unchanged at its seventh meeting of 2026, the fourth under Chair Kevin Warsh, and reiterated its commitment to containing inflation without signaling a December move. The market read the decision as pointing toward a December hike.
Warsh used the press conference to announce a structural change: starting in fall 2027, the FOMC will hold eight meetings a year, one of them devoted exclusively to broad monetary policy strategy with no rate vote and no policy statement. By 2028, he said, two meetings would be strategy sessions and six would set policy.
"We believe eight meetings per year is the right number for the FOMC, but in 2027, one meeting will focus exclusively on broad monetary policy strategy. With neither a vote on the federal funds rate nor a policy statement anticipated," Warsh said.
The press conference itself was shorter than usual, with one question going to reporters who are normally called on. Stenovec said the change wasn't a complete surprise given Warsh's past comments that the Fed over-communicates. Carol Massar said she was surprised the reduction wasn't larger.
Omar Aguilar, CEO and chief investment officer of Schwab Asset Management, said the decision wasn't a surprise and pointed to two dissents as a signal of the internal debate. He expects a December hike unless inflation data change dramatically. On communication, he said Warsh has deliberately chosen to be vaguer and let others interpret the Fed's intent, but the underlying data-driven process hasn't changed. Asked why not raise rates now, Aguilar said much of current inflation is driven by energy, and it isn't yet clear how much is oil prices versus something more durable. This Fed, he said, is more patient than its predecessors โ "a double-edged sword," because patience can mean being late, but it also avoids overreacting to a single data set.
The AI risk nobody is pricing: companies borrowing to keep up
Aguilar offered the sharpest warning of the episode, and it wasn't about Microsoft or Alphabet. He said AI capital expenditure is decelerating โ not stopping, but becoming more selective. The bigger risk sits with the rest of the market: companies outside the hyperscalers that lack strong balance sheets and cash flow are using the debt market to fund AI investment. That, he said, is where credit-market challenges will show up.
"You have a company that doesn't necessarily have that cash flow or doesn't even have that strength of the balance sheet. And they're using the debt market to invest in AI. That's where we started to see challenges in the credit market," Aguilar said.
He said a bust isn't inevitable but volatility and dispersion are, with clear winners and losers, and that some credit events are a normal part of the cycle. His biggest risk to the investing environment is the profitability of AI investment: spending can continue, but returns eventually have to show up in earnings. He said earnings expectations going into next year are "way too high" and disappointment is latent.
Stenovec pressed the productivity argument โ that non-hyperscaler companies could become the biggest earnings beneficiaries of AI within a couple of years. Aguilar was unconvinced. He said it isn't clear there's evidence of cost savings yet, and the labor market shows neither firing nor hiring, which is consistent with companies waiting for productivity gains that haven't arrived. He allowed that productivity could accelerate other parts of a business without leading to layoffs.
Aguilar listed the risks his team is watching: the midterms, a U.S. war with Iran, higher energy prices, persistent inflation, a higher-rate environment, decelerating capex, and tougher earnings comparisons. He called it a "wall of worries" that is getting taller.
Schwab is trying to solve a talent shortage with a national holiday
Schwab used the event to announce the first National RIA Day, set for February 17, celebrating independent registered investment advisors. Lisa Salvi, head of business consulting and education at Schwab Advisor Services, said the idea started at Impact about two years earlier, when a professor chaperoning students wanted to bring the concept back to campus. Universities began hosting their own days organically, and Schwab turned it into a national event. About 30 universities are expected to participate, and Schwab will ring the closing bell in Washington, D.C.
Salvi said the industry needs 70,000 more people over the next five years to keep up with organic and inorganic growth โ a figure that excludes attrition, retirements and newly formed firms, and doesn't account for AI's effect. Schwab has helped found and grow certified financial planner programs at universities for over a decade and is now discussing with schools how AI will change the roles they train students for.
On hiring, Salvi said advisors are prioritizing soft skills. In Schwab's 2026 Independent Advisor Outlook study, 33% of advisors prioritized relationship and emotional intelligence skills, 30% prioritized judgment and critical thinking, and 26% wanted candidates fluent with AI. She said about half of advisory firms use AI in administrative operations and marketing strategy, but the client relationship and trust remain the human core. Her advice to students: become fluent in AI regardless of your job, because no one can do it for you โ "go build an app today."
What the split means for investors
The two earnings reports set up a clean test of the market's new discipline. Microsoft gets credit for cloud growth funded from cash flow. Alphabet gets punished for cloud growth funded with negative free cash flow. The analyst calls will matter more than usual โ Microsoft's capex guidance for 2027 and its non-OpenAI backlog, Alphabet's answer on whether AI search can grow usage without eroding monetization. Aguilar's warning about debt-funded AI spending outside the hyperscalers is the thread that connects the earnings discussion to the credit market, and it's the risk he says he'll be watching into next year. The Fed's December 9 meeting is the next scheduled policy event, with a hike priced in unless inflation data shift. For a reader holding AI exposure, the takeaway is simple: growth alone is no longer enough. The companies that can prove they're generating cash while they build will keep their premium. The ones that can't will keep answering the same question โ what exactly is all this spending buying?