The Numbers Are Good. The Mood Is Bad. Inside Q3’s AI Earnings Reckoning.

The Numbers Arrive, and Nobody Claps

The first peak week of Q3 earnings season began on a Monday. 1,885 companies were scheduled to report. FactSet had the S&P 500 EPS growth number at 29.5%. The profit beat rate was 83%. The revenue beat rate was 66%. These are strong numbers. They arrived in a room with no applause.

I read the releases at a desk with bad lighting. The press releases came in waves. Western Digital. Bloom Energy. Comfort Systems. IBM. Meta. Each one carried a story. The stories were not about the numbers. The stories were about what the numbers might mean later. The market has developed a habit of looking past the quarter. It looks at the next quarter and the quarter after that. It looks at the capex line. It looks at free cash flow. It looks at the gap between spending and return.

The season's dominant theme arrived early and stayed late: strong growth is not enough. You need confidence in its sustainability. That sentence sounds reasonable. It is also a warning. Sustainability is a question about the future. The future is a place where numbers have not yet arrived. The market is asking companies to prove something that cannot be proved until it happens. This is the trap of the season.

The two forces in the headline — strong beats and sky-high AI scrutiny — sit at the same table. They do not argue. They do not agree. They simply occupy the same room, and the room is tense.

The Beats That Should Have Mattered

Western Digital reported fiscal first-quarter adjusted earnings of $4.64 per share. The consensus was $4.10. Revenue rose 55% to $4.38 billion. The consensus was $4.14 billion. Adjusted gross margin reached 58.1%. Free cash flow was $1.38 billion. The guidance for the next quarter called for revenue of $5.1 billion and adjusted EPS of $5.80. The consensus was approximately 11% and 20% lower, respectively.

By any traditional measure, this was a blowout. The stock reversed its initial gains and fell with the broader AI complex. The beat was real. The reaction was not about the beat.

Bloom Energy reported adjusted EPS of $0.94. The expectation was approximately $0.68. Revenue jumped 138% to $1.236 billion. The estimate was near $1.08 billion. Adjusted operating margin expanded to 23.1% from 8.9% a year earlier. Management raised full-year revenue guidance to $4.4 billion–$4.6 billion and adjusted EPS guidance to $3.10–$3.40. Negative free cash flow of approximately $181 million sat in the same report. The stock moved violently. The operating strength was overshadowed by the cash line.

Comfort Systems USA reported earnings of $13.91 per share. The expectation was $12.72. Revenue reached $3.77 billion. The consensus was $3.38 billion. Backlog expanded to $17.81 billion from $14.06 billion in the previous quarter. Management described demand as relentless. Gross margin fell approximately 100 basis points below expectations. The stock did not celebrate.

These are the beats. They are real. They are large. They did not produce joy. The market is not in a mood to reward the present. It is in a mood to interrogate the future.

Meta Spends $34.9 Billion and Forgets to Smile

Meta reported third-quarter revenue of $64.6 billion. The number was up 26% year over year. The company beat expectations. Capital expenditures reached $34.91 billion for the quarter. The figure was near $35 billion. It exceeded operating cash flow. Free cash flow went negative at -$2.1 billion. Meta raised the lower end of its 2026 capital spending guidance. The range became $135 billion to $145 billion. The company declined to give a 2027 figure. It also announced plans to create a new class of nonvoting shares to facilitate future fundraising.

The stock fell 7%. The revenue growth was strong. The capex was stronger. The market read the combination and reached for the sell button.

Mark Zuckerberg used the call to talk about Muse, Meta's AI model. 10 million people in the U.S. and Canada had set it up in six weeks. Around a quarter of that group used it daily. The pitch was clear: Muse can become the business that makes the investments pay off. Whether it can generate enough revenue to justify the spending remains unproven.

Meta's report set a new bar for the season. Investors are no longer giving tech giants a blank check for AI infrastructure. Runaway capex, even when paired with double-digit revenue growth, triggers immediate margin anxiety when free cash flow shrinks. The market wants the spending to show up in the cash flow statement. It wants the money to come back. It wants the story to close.

UBS Writes Three Tests on a Napkin

UBS laid out three tests for the AI trade before the mega-cap tech reports began. The bank said rising interest rates and higher financing costs have narrowed the tolerance for open-ended AI spending. Capex guidance now sits at the center of market reaction. Any sign that spending is outpacing revenue can hit hyperscaler shares even on headline beats.

The first test is the health of the core business. The largest tech groups enter the season with strong positions in advertising, software, devices, and digital commerce. Continued demand in those areas helps fund AI investment. The question is whether that growth can hold up as year-over-year comparisons get tougher and economic risks persist.

The second test is monetization. Rising use of AI products is encouraging. Investors increasingly want evidence that usage translates into revenue and profit. Subscriptions. Consumption-based pricing. Advertising. Transactions. Greater cloud utilization. Clarity on pricing and customer willingness to pay would strengthen the case that AI is becoming a revenue driver rather than mainly an investment program.

The third test is returns on spending. Even if AI revenue grows, the scale of infrastructure investment has implications for margins and cash flow. Investors will focus on capacity utilization, pricing power, capital requirements, and returns on AI investment. Management commentary on the future pace of capex will shape expectations for 2027 and beyond.

UBS expects a strong quarter from a fundamental perspective. It also warned that elevated expectations could produce company-specific volatility. The framework is simple. The application is not.

The Man From IBM Says the Quiet Part

IBM reported third-quarter revenue of $17.1 billion, up 5% year over year. The company beat Wall Street's estimates for revenue and adjusted earnings. The stock slid anyway. The full-year revenue outlook was unchanged. CEO Arvind Krishna offered a pointed assessment of enterprise AI spending on the earnings call.

Krishna said enterprise customers are becoming "a lot more cognizant and a lot more aware of what is the true ROI of all of their AI spend." He noted growing concerns among clients about "some of the AI spending completely out of control with no real return." He framed the scrutiny as an opportunity for IBM. Companies need help redesigning workflows and reining in AI token spending. "That is the tailwind for us," he said. He urged businesses to "pick a few" AI initiatives and scale them meaningfully. He cautioned against spreading resources across "100 little experiments" that never receive sufficient investment to deliver tangible results.

The remarks unsettled investors. The stock declined despite the beat. The reaction is a signal. The enterprise customer is not the same as the retail investor. The enterprise customer pays for the software and the consulting. The enterprise customer is looking at the invoice. Krishna described what he sees from the front row. The front row is worried about the return.

Apollo Finds the Zero

Torsten Slok, chief economist at Apollo Global Management, wrote a note on a Saturday. The note said artificial intelligence has delivered "zero financial payoff to corporate America outside of tech." Revenue per worker is rising for the Magnificent 7. It is flat for the S&P 493. Mainstream corporate America has begun "pulling back and starting to view token spending not as a source of growth but as a cost center that weighs on margins."

The note landed during the first peak week of earnings season. It gave a name to the anxiety. The anxiety is not that AI is failing inside tech. The anxiety is that AI is not spreading the way the spending suggests it should. The hyperscalers build the data centers. The enterprises buy the tokens. The productivity shows up in the tech sector's revenue per worker. It does not show up in the broader economy's revenue per worker. The gap is the problem.

Slok's note is not a prediction. It is an observation. The observation is that the spending has run ahead of the evidence. The market is now asking for the evidence. The evidence is not arriving fast enough for the market's taste.

The Hyperscalers Line Up

Microsoft, Alphabet, and Amazon reported this week. Apple follows next Monday. Meta's report provided the playbook. The critical test is whether cloud revenue growth accelerates enough to justify the surging capital budgets.

CFRA's preview put the numbers in context. Google Cloud was poised for 85%–90% growth. Microsoft Azure and Amazon Web Services were expected to accelerate from the second quarter. Microsoft's new reporting structure would give investors their first look at Azure revenue in dollars. Advertising growth at Meta and Alphabet was expected to slow on tough comparisons. Early metrics on Meta's Muse assistant were a key watch item. NVIDIA's Vera Rubin platform was set to be the fastest product ramp in the company's history. CFRA already modeled another 50% jump in hyperscaler capex for 2027. The question is which companies can show a return on that spend.

The hyperscaler calls are not about the quarter. They are about the guide. The guide is about the next year. The next year is about the return. The return is about the cash. The cash is about free cash flow. Free cash flow is the number that matters now. Meta proved it. The market will apply the lesson to every name on the calendar.

A Spotty Start and a High Bar

The season did not begin with a unified message. Banks delivered mixed results. Financials logged an 80% EPS beat rate against a revenue beat rate of 53.3%. Eight companies beat revenue estimates. Seven fell short.

Blackstone reported EPS of $1.32, beating estimates by $0.06. Revenue fell 1% year over year to $3.06 billion, missing by $150 million. AUM reached $1.37 trillion. Nasdaq delivered EPS of $1.04, topping estimates. Capital One reported EPS of $6.69 against $5.43 expected. SLB reported adjusted earnings of $0.64 per share against a consensus of $0.62. First Citizens BancShares reported adjusted earnings of $55.37 per share against a consensus of $46.61.

Deckers Outdoor delivered a mixed report. Adjusted EPS reached $2.31 against expectations of $1.79. The result included a $0.28-per-share tariff refund benefit. Revenue of $1.48 billion missed the $1.506 billion consensus. Hoka and Ugg underperformed. Management reduced its fiscal 2027 revenue outlook to $5.7 billion–$5.725 billion from $5.86 billion–$5.91 billion. Shares declined approximately 8%. Volvo Cars reported retail sales declining 11%.

State Street's research arm noted that first-half 2026 earnings surprised by 16.8% in Q1 and 26.0% in Q2. The median company beat expectations by 5.8% in Q2. The aggregate surprise was 26.0%. The disparity suggests that the largest index constituents accounted for a disproportionate share of the outperformance. Technology, Communication Services, and Consumer Discretionary generated roughly 75% of total earnings growth in each quarter.

The season is strong at the index level. It is uneven underneath. The market knows the difference.

The Narrative Eats the Beat

CFRA's analysts wrote that Q3 "won't be won on EPS beats alone." The stocks set to keep running are the ones telling the most convincing AI monetization story. Cloud is the bright spot. The question is which companies can show a return on the spend.

This is the sentence that explains the season. The beat is the entry ticket. The narrative is the ride. A company can report a strong quarter and still lose the room. Meta did. IBM did. Western Digital did. The market is not grading the quarter. It is grading the story about the next four quarters.

The story has three chapters. The first chapter is the core business. The second chapter is monetization. The third chapter is returns on spending. UBS wrote the outline. The companies have to fill in the pages. The market is reading ahead.

The Season Grinds On

The press releases keep coming. The numbers stay strong. The mood stays tight. The first peak week of Q3 earnings season is a machine that produces records and anxiety at the same time.

The records are real. 29.5% EPS growth. 83% profit beat rate. 1,885 companies reporting. The anxiety is real too. The capex is large. The free cash flow is negative in places. The return is unproven outside tech. The market wants proof. The proof is not here yet.

I closed the laptop. The desk lamp hummed. The numbers were good. The room was quiet. The season had done what it always does. It delivered the beats. It withheld the comfort.