I spend a lot of time asking one simple question: Can earnings keep doing the heavy lifting? Valuations are stretched at the top end of the market, and the AI investment cycle is still in full flight. Yet, for all the worry, the income statement remains surprisingly resilient.
So far, the reporting season has two messages:
Weak earnings: Number one risk
When forward profits rise, markets tend to increase. And earnings look pretty good:
2025 numbers are holding up. 2026, after an earlier wobble, is ticking higher again. Q3 estimates are being revised up, and Q4 is holding firm.
Earlier this month, we saw pockets of pain in auto insurance and Jamie Dimon warning where there is one cockroach, there are often more. Despite that, financials were the largest positive contributor to growth. Booming trading, capital markets, and decent wealth and insurance results.
Detractors include energy, communication services ex the mega caps, and healthcare. None of that is shocking. The takeaway is that the average company still looks okay, not just the headline tech names.
Profit margins continue to hit new highs. Will they revert to longer-term averages? I say no.
My take is that Warren Buffett(rightly) told us that high margins are good, and so management now try to get high margins. Some of this comes from accounting trickery, but I do think most of it is real.
First, the long-term switch from manufacturing to services mostly means higher margins.
Second, companies are much more comfortable outsourcing low-margin business than they used to be: Boeing, for example, went from making everything themselves to assembling planes from other people's components. Companies stopped hiring cleaners directly and started hiring cleaning companies. Basically, low-margin parts of businesses got outsourced to unlisted smaller companies or to China.
One of Trump's windmills illustrates both parts: shipbuilding. Yes, China, Korea, and Japan now do all the shipbuilding globally. But these are low-margin businesses. The high-margin part is the engine and propulsion systems. Shipbuilders of yesteryear worked out that they were better off letting someone else do the low-margin part of the work so they could focus on the high-value, high-margin parts.
Even if margins are mean-reverting in the longer term, Trump is busy cutting regulation and running big government deficits. Both are good for corporate margins:
Unemployment is a potential land mine. Layoff headlines are noisy. And government shutdowns are limiting data.
Despite the headlines, recent layoff announcement readings sit near the five‑year average. Not benign, not alarming. The real worry would be layoffs rising while companies protect margins. That would weigh on consumption into sticky inflation—a stagflation setup that is visible at the edges in both Australia and the US.
I am watching this closely, but the hard data is not flashing red yet.
Capital expenditure on Artificial Intelligence is the pillar holding up both the growth narrative and, increasingly, profits. If AI capex slows, the share price boom is likely over.
Management teams across the biggest platforms continue to guide to heavy outlays. Over the past 12 months, the top five players alone have flagged more than $300 billion in spend, and the trajectory is up.
As a share of US GDP, AI capex from just five companies is approaching 1%. Importantly, excluding Oracle, they are funding this with free cash flow, not debt. This is a reinvestment cycle powered by cash, not leverage.
In some ways, the big tech companies sell ads, software, and retail products globally and then plough the cash into building data centres. Mostly in the US.
That makes it more durable than debt-driven booms, and arguably more durable than the equity‑financed dot‑com era as well. Or simply a sign that the bubble still has scope to grow into a debt-driven boom.
Broader tech investment as a share of GDP is back near prior peaks over 4%. That scares people because of the dot‑com echo. But the economy is far more digital now.
Spend that once lived in “mechanical” budgets has migrated to “digital.” Agriculture is a neat example. A tractor 25 years ago was not IT spend. Precision software, drones, and sensor networks shift costs into the tech bucket.
The 1800s railroad boom reached about 6% of GDP and held above 4% for years. I am not using that as a prediction, more as an indicator that this can run further.
What does all this mean for portfolios? The market has split into AI and non‑AI. The non‑AI universe screens as fair to slightly rich versus long‑term history. Nothing extreme. The AI leaders, by contrast, are expensive. In my valuation buckets across roughly 1,500 global stocks, the largest 2%—where most AI leaders sit—trade above 30x on 12‑month forward earnings.
On a 24‑month basis, that falls to roughly 25x. If they deliver another year of 20%+ earnings growth, you can pencil that down below 20x. It is still not cheap. The path to “reasonable” hinges on how many years of outsized growth they can deliver.
That’s why the AI capex cycle is the fulcrum. If it cracks, I move from constructive to cautious. So far, I can't see the cracks. Platform spending plans are intact. Data‑center demand is robust. Power and chip constraints are real but manageable. Margins are high. And recent reports from the Magnificent 7 show strong earnings and upbeat guidance tied to AI infrastructure.
My stance is practical. I want to keep exposure to the AI boom, but recognise the risks. I hold a small sleeve in the pointy end—names with less current profit but high optionality. Most of my exposure sits in large, profitable platforms with real revenues and clear visibility on monetization. I pair that with non‑AI holdings that can weather a stagflationary patch: resilient balance sheets, pricing power, and less earnings cyclicality. The rest of the market is still available at roughly normal multiples. You do not have to own only the crown jewels to participate.
Three signals I am watching:
So, can earnings keep carrying this market? My answer today is yes, with guardrails. The data says profits are still rising, especially where AI is thickest. Valuations at the top are rich but not irrational if growth persists. The capex boom is funded by cash, not debt, and constrained by physics more than finance. Fiscal policy remains a tailwind to margins, for now. The risks—labor softness, fiscal tightening, or a genuine slowdown in AI returns — are (at worst) on the horizon, not at the door.