3Q25 earnings exceeding expectations 

The 3Q25 earnings season is once again exceeding expectations, tracking for a fourth consecutive quarter of double-digit growth. Analysts are projecting S&P 500 earnings per share (EPS) growth of 10.3% y/y vs. expectations of 7.3% at the end of the quarter. Sales have driven around two-thirds of EPS growth, with margins supplying one-third. About 82% of companies have beaten on earnings (vs. 73% on average), coming in 6.6% above expectations (vs. 6.2% on average). At the sector level, most of the index’s earnings growth is coming from technology and financials, while health care and consumer companies look softer. 

• Financials: All five subsectors are reporting double-digit y/y profit growth. 88% of financials companies have beaten, and earnings have come in 8.6% above estimates, thanks mainly to capital markets activity. Equity volumes are up 55%, hitting their highest level since 2021, and high yield, leveraged loan syndications and M&A also saw strong growth of 40-59% y/y. Goldman Sach’s CEO stated, “Though near-term policy considerations are still relevant, many CEOs have shifted their focus back to long-term and strategic decision-making, particularly amid a more supportive regulatory environment.” Commercial & industrial loan growth is still modest, but balances are ticking up in consumer and auto. Thus far, banks are not seeing material cracks in the low-end consumer either. According to Citizens’ CEO, “When I look at the health of the actual U.S. consumer, it's also very, very stable. You have to really de-average it to see stress and it's on the lower end of the market…I don't see anything right now that would suggest even really a blip in terms of consumer credit right now for us.”
• Health care: Health care is weak this quarter as margin pressure overwhelms sales growth. While the pharmaceuticals industry is poised for strong earnings gains, health care providers are on track for an EPS contraction of 7%. Medicaid loss ratios are elevated, as patients have been sicker than predicted, and treatment costs have continued to rise. Medical supplies companies are also struggling as questions about funding mute demand from U.S. government & academic institutions and trade tension mute demand from China.
• Consumer: Both discretionary and staples companies are struggling through weak demand and increased costs from tariffs. Despite management commentary about mitigation measures, the 58 reporting consumer companies are seeing an aggregate decline in both gross and operating margins, but there is significant dispersion between the best and worst performing names.¹ Companies have started passing through these costs, but the critical question is how much this will discourage already cautious consumers. According to O’Reilly’s CEO, “We began to encounter modest pressure to DIY transaction counts midway through the third quarter, which we believe reflects some degree of initial short-term reaction…to rising price levels…we saw a significant ramp in tariff-driven acquisition cost increases and made appropriate adjustments to selling prices.”

Mag 7 monolith no more

Although earnings season stretches roughly seven weeks, these days, investors only care about one. Five of the Magnificent 7 reported last week and the biggest names in tech are driving 38% of the index’s profit growth, sporting margins of 22%. Markets are rewarding these names more selectively, even after strong results. Admittedly the value created for shareholders has been impressive -- the Mag 7 are up 191% since the launch of Chat GPT on 11/30/2022.

• 60% of the Mag 7 return has been driven by earnings growth. Last year, the Mag 7 grew profits a whopping 40%, and there wasn’t much to complain about in this quarter’s results either. Hyperscaler cloud revenues are up an average of 27% y/y and customers are shelling out for AI tools that can code and advertise. Internally, AI is also driving revenue growth and cost savings. Amazon’s AI-powered shopping agent Rufus is “on track to deliver over $10bn in incremental annualized sales,” and according to Alphabet’s CEO, “…now nearly half of all code [is] generated by AI.”
• The other 40% of the Mag 7’s post Chat GPT return has come from multiple expansion. But, at 32x forward earnings, valuations are not much higher than where they started the year, and much lower than the 70x peak of the info tech sector in 2000. On a trailing 3-year basis, the Mag 7 are trading at a PEG ratio of 1.2x vs. 2.2x for the broader market.
• Investors are also still being paid to wait and see. The Mag 7 are expected to grow free cash flow by 5.9% in 2025. However, seismic AI investments are starting to take precedence over shareholder returns. The Mag 7’s shareholder yield is currently tracking around 1.7%, down from a peak of 4.2% in 2022.

However, investors should also be attuned to the volatility and risks of this cohort. Its defensive properties have been inconsistent: the Mag 7 had a less pronounced drawdown than the index as a whole in 2020, but a deeper correction in 2022 and in spring of 2025. In 2022, the Mag 7 experienced a peak drawdown of 43%, highlighting that significant pullbacks can still occur within a longer bull run.

Though perhaps the bigger risk is already underway: is the Mag 7 a fading concept? Four of the Mag 7 are underperforming the index this year, and only two are in the top 50 performing S&P 500 stocks this year. With the technology sectors expected to contribute 64% of 2025 earnings growth, all eyes are on the broader AI complex. As prices hit all-time highs, and announcements about more capex, billion-dollar investment deals and debt issuance fly around, investors are growing concerned we’re in a bubble. 

Is AI a bubble?

Railroads, electricity, landlines, cars, plastic, shale, internet. AI might be new, but innovations that spur capital investment cycles are not. Profits and productivity have always materialized eventually, but some of the time, though not all the time, markets have gotten ahead of themselves. It’s still early days for AI, but we’re not seeing too many signs of excessive exuberance.

In previous technological transformations, first movers haven’t always reaped the benefits. WorldCom, Global Crossing, and Level 3 built the internet we still use today, but two of the three went bankrupt. They grew investments faster than demand, prices collapsed, and a decade later, companies like Amazon and Google swooped in and built trillion-dollar businesses on that glut of free bandwidth. But today’s hyperscalers aren’t just building the compute AI needs to run. They also own and operate the network, architect the software layers on top and control the distribution. The AI transition is different in a few critical ways: 

• Profits are keeping pace with enthusiasm. Since the launch of ChatGPT, the info tech sector has risen 158%, supported by soaring profit growth. In contrast, returns became untethered from earnings growth during the dot com bubble. This doesn’t preclude a future AI bubble, but for now, price and earnings look reasonably aligned. 

• Demand precedes – not succeeds – investment. The hyperscalers have already invested close to $600bn in AI, and the bill keeps on getting bigger. After 3Q25 earnings reports, consensus estimates for 2026 and 2027 capex have increased by $47bn and $61bn, respectively. But it’s still not enough to meet existing demand. This earnings season, cloud providers reported an aggregate $1,202bn in backlog. Google has “signed more deals over $1billion through Q3 this year than…in the previous two years combined,” and Amazon’s CEO highlighted, “AWS is growing at a pace we haven’t seen since 2022.” While Meta’s stock sold off after it raised both 2025 and 2026 capex guidance, CEO Mark Zuckerberg doesn’t think it will go to waste: “I think that it's the right strategy to aggressively frontload building capacity…That way, if superintelligence arrives sooner, we will be ideally positioned for a generational paradigm shift...If it takes longer, then we'll use the extra compute to accelerate our core business which continues to be able to profitably use much more compute than we've been able to throw at it.”

• Financing is mostly cash – not debt. Hyperscalers are investing massive amounts into AI, but they generate massive amounts of cash each quarter too. Free cash flow is down from its 2024 peak of $238bn but still strong at $188bn, and growth is projected to inflect in 2026. Some of these companies have also started to tap both public and private debt markets to fund data centers. In and of itself, this isn’t a bad thing, as debt is part of an optimal capital structure, but it’s a risk worth watching, especially as less profitable companies lever up. Hyperscalers, however, have a lot of room to run, with an average net debt to EBITDA of 0.8x vs. 2.6x for the average investment grade issuer and 1.0x for the S&P 500 index. 

• Strategic circularity. The hyperscalers spend 18% of their cost of goods sold on NVIDIA and represent 42% of NVIDIA’s revenue. Add to that multi-billion-dollar deals announced between chip makers, hyperscalers and LLM developers, and investors are growing rightfully concerned about who’s footing the bill. Some have even compared it to telecom carriers’ buying up each other’s excess fiber cable capacity during the internet bubble. However, today’s AI deals are strategic investments tied to real demand, built on exponential revenue growth. Rather than circularity, these deals represent competition, as leading LLM developers secure compute from multiple cloud providers, who have deals with other LLMs. 

Still, investors shouldn’t get complacent. Power and GPU supply could constrain growth to below expectations, though this could help prevent excess supply. Adoption appears to be rapid, but sticky revenues from paid subscribers with enterprise-driven use cases are not guaranteed. Faster chip obsolescence could also change the economics. Even a 1-year haircut on the useful life of AI infrastructure could be a 3% hit to earnings.² And although circularity may be an essential foundation of investment today, one damaged link could have an outsized impact.

Investment implications

We might not be in another internet bubble, but today’s megacap tech giants aren’t destined for eternal domination either. They’re competing against each other, and disruptors will inevitably emerge. With valuations this high, and investments this astronomical, there isn’t much room for error. Luckily for investors, investment opportunities have broadened this year, throughout the AI value chain and outside of the realm of AI altogether:

• The tech sectors continue to post exemplary profits; however, greater dispersion is emerging among the Mag 7, emphasizing the need for selectivity.
• Within AI, beneficiaries should continue to broaden out from the innovators (tech) to the enablers (industrials, utilities) and eventually to the adopters (financials, health care).
• Private markets offer additional access to AI opportunities across the capital structure.
• Financials have boasted the largest increase in earnings since the beginning of the quarter, contributing 38% of overall 3Q25 profit growth. A resilient macro backdrop, healthy capital markets and transactions activity, and a steepening yield curve could provide a second wind to this recently sideways sector. 

¹ AMZN and TSLA were excluded from this analysis.

² Rochester Cahan, Tyler Albrecht, Yu Bai. Empirical Research Partners. “The Hyperscalers: Making the Jump to Hyperspace?” August 11, 2025.

dccd9987-ba5a-11f0-8cf0-29f8bd2a75d0

The firms highlighted above have been selected based on their significance and are shown for illustrative purposes only. They are not recommendations.