Revenue growth was the gospel of 2018. Then survival was the gospel of 2022. Now everyone's betting on AI productivity that hasn't fully shown up in the numbers yet — and the S&P PE ratio suggests the market believes the story completely.
(This is personal opinion and not financial advice. Do your own research before making any investment decisions.)
Why This Matters: The Playbook Keeps Changing
I started investing seriously around 2018. That era had a clear religion: revenue growth above all else. Positive EBITDA was almost beside the point. If you were growing fast enough, the market believed profitability would follow. SaaS multiples were extraordinary — companies burning cash were celebrated like they'd invented fire. Marketplaces were appearing everywhere. Every week there was a new app promising to be "the Airbnb of dog grooming" or something equally ambitious.
Then the pandemic distorted everything. Streaming, delivery, e-commerce, stay-at-home anything — all exploded. The numbers looked incredible, but a lot of that demand was borrowed from the future. Peloton thought it had conquered fitness. It had just borrowed everyone's lockdown boredom.
2022 corrected hard. There's a well-documented historical pattern of midterm election years in the US being difficult ones for markets — and 2022 delivered on that pattern with particular conviction, with the S&P 500 falling roughly 19% over the year. Rate hikes, multiple compression, and a lot of portfolios humbled. My own included. Looking ahead, 2026 is the next US midterm year, and while history doesn't repeat on a schedule, it does tend to rhyme — worth keeping somewhere in the back of your mind as you think about positioning.
And then, almost immediately after, the AI rocket launched.
Nvidia, Microsoft, Google, Meta — infrastructure and application plays across the board. Even Intel found a reason to rally (briefly, painfully, and then less so — but still). The AI wave pulled capital into a new narrative with enormous velocity. We've been riding that wave since, and the question now is whether the fundamentals are catching up to the story, or whether the story is still running ahead of the fundamentals.
For context: GDP growth expectations in Europe have been revised downward repeatedly, with the eurozone growing below 1% in recent years and forecasts remaining modest. The US has fared better, but the picture is uneven, and productivity gains from AI investment have not yet translated clearly into broader economic output. That gap is worth sitting with.
Which means the usual metrics are worth questioning.
What Financial Metrics Actually Matter Right Now for AI-Era Companies
Revenue growth still matters, but it's no longer sufficient on its own. For the large, mature companies leading their sectors, user growth is largely saturated. The question has shifted from "how fast are you growing" to "how productively are you deploying capital."
Here's what I'm actually watching — with real companies as reference points, though none of this is a recommendation to buy or sell anything:
ROIC (Return on Invested Capital). This is becoming more interesting to me than almost anything else right now. If AI investment is genuinely making companies more efficient and more profitable, it should start showing up here over time. A company like Microsoft is worth tracking here — it has historically maintained ROIC above 20%, and with massive Azure AI capex now flowing through, the question is whether that ratio holds or drifts. The signal is still early and noisy, but I watch it. I'm not concluding anything yet.
Gross Margin Trend. This is a cleaner signal than net margin because it strips out the noise of one-time charges, restructuring costs, and accounting choices. Nvidia is a fascinating case — its gross margins expanded dramatically through the AI infrastructure wave, reaching above 70% in recent quarters. That's a real signal of pricing power and structural leverage. For software companies this is easier to trace. For hardware it's messier, but still worth tracking over consecutive quarters.
Revenue Per Employee. This is the one I keep coming back to. If AI is genuinely transforming productivity inside organisations, we should see revenue per employee climb — not just because of layoffs, but because fewer people are generating more output. Meta is often cited here: after its "year of efficiency" in 2023, revenue per employee climbed significantly. But it's worth asking honestly how much of that was AI-driven output and how much was just aggressive headcount reduction. That's a meaningful distinction, and the earnings call transcripts usually give you enough to tell the difference.
Operating Leverage (for asset-heavy businesses). For companies with significant fixed cost bases, the question is whether incremental revenue is dropping to the bottom line at an improving rate. Think of a company like Amazon Web Services within Amazon's broader P&L — as revenue scales, do infrastructure costs grow slower? Right now I see volatile results here across the market, which makes me cautious about drawing strong conclusions. But directionally, it's worth checking whether a company's cost base is becoming more or less fixed relative to revenue growth.
Is the AI Rally a Bubble Like the Dot-Com Crash of 2000?
This is the question I can't fully answer yet, and I want to be honest about that.
The parallel is tempting. Enormous capital flowing into infrastructure for a transformative technology, valuations running ahead of demonstrated returns, a narrative that feels so compelling that scepticism seems almost anti-intellectual. My instinct says, that AI is so powerful and immediate that we will see a efficiency boost soon.
Because there are real differences. The companies at the centre of this cycle are, by and large, already profitable. Nvidia is not Pets.com. The infrastructure being built is not hypothetical. Compute demand is real. The question is not whether AI is useful, but whether the current level of investment will generate returns commensurate with the capital deployed — and on what timeline.
The PE picture is worth grounding in history. At the peak of the dot-com bubble in 2000, the S&P 500's trailing PE ratio reached approximately 30–33x. It then collapsed to around 15x by the mid-2000s. In the years following the 2008 financial crisis, the market bottomed around 10–13x. The long-run historical average for the S&P 500 sits somewhere around 15–17x trailing earnings, depending on the period you measure.
Today, the S&P 500 trailing PE ratio is sitting in the range of 24–27x — well above the historical average, though not quite at dot-com peak territory. Forward PE ratios (based on expected earnings) are somewhat lower but still elevated. The CAPE ratio (Shiller PE), which smooths earnings over a ten-year cycle, is currently above 35x — a level that has historically preceded lower long-run returns, though it's a poor short-term timing tool.
The PE compression we saw in the recent market pullback made valuations at least slightly more honest. Not cheap, but less obviously detached. Whether that compression continues — or whether earnings actually grow into current multiples — is the open question.
The soft landing versus hard crash question remains genuinely open. I don't trust anyone who tells you they know the answer.
What Most People Get Wrong When Reading Earnings Right Now
They're still looking at EPS and revenue beats versus consensus, and treating that as the whole story.
Consensus estimates are manufactured. Analysts lower estimates into earnings season so companies can beat them. This is not a conspiracy, it's just how the game is played. A beat against a sandbagged number tells you very little about whether the underlying business is actually healthier.
The more useful exercise is to read the capital allocation section of the earnings call carefully. Where are they putting money? What is the stated expected return on that investment, and over what horizon? How much of current AI capex is discretionary versus committed? That's where the real signal lives.
Also worth noting: companies are currently investing enormous sums into AI infrastructure with limited near-term return visibility. That is not inherently irrational, but it does mean the financial statements today are a lagging indicator of a bet being made on the future. Read them that way.
The market can stay irrational longer than you can stay patient, but it can also stay rational longer than your narrative can stay intact.
What to Actually Do
- Watch ROIC across 3-4 consecutive quarters, not just one. A single quarter of improvement means nothing. A trend means something. Set a simple tracker for the companies you hold and check it after each earnings cycle. Microsoft and Alphabet are useful benchmarks to compare against sector peers.
- Read gross margin trends before revenue headlines. Revenue can be bought with sales spend. Gross margin expansion is harder to fake and tells you more about the structural economics of the business. Nvidia's margin expansion is the reference case — hold other companies to a similar standard of scrutiny.
- Treat revenue-per-employee improvement with scepticism until proven otherwise. Ask whether it's driven by headcount reduction or genuine output growth. Meta's 2023 efficiency gains are the often-cited example — but dig into the transcript before you draw conclusions.
- For asset-heavy companies, calculate operating leverage manually. Take revenue growth percentage and compare it to operating cost growth percentage over the same period. If costs are growing faster than revenue, the story about efficiency is not in the numbers yet.
- Don't confuse infrastructure spend for immediate productivity. Massive capex into AI today is a future bet, not a present result. Be patient with the timeline, but be honest about the gap between investment and return.
- Keep the 2026 midterm in the background. History doesn't repeat, but midterm years have a tendency to test portfolios. Not a reason to panic — just a reason to know what you own and why.
- Check your own narrative. Every time I think I have the market figured out, it's worth asking whether I'm reading data or confirming a story I've already decided to believe. That question is uncomfortable. It's also the most useful one I know.
I'll keep watching and will write a follow-up when the picture gets clearer. Or messier. Probably messier first.
If you're working through the same questions — whether it's about valuation discipline, reading earnings, or just figuring out how to think about the AI cycle — I'd genuinely enjoy the exchange. Please get in touch and let's talk it through.