1. Capex today is the cost of control tomorrow.

 

Free cash flow (FCF) is being deliberately depressed by investment, not structurally destroyed. As Amazon's CEO Andy Jassy has said, "we're investing to be the meaningful leader, and our future business, operating income, and FCF will be much larger because of it." The FCF trough we're seeing now isn't a sign of a broken model — it's the price of owning the rails into the AI economy of the future. By most estimates, hyperscaler FCF recovers sharply after the trough, with aggregate free cash surpassing recent peaks toward the end of the decade. The window to buy that at a discount closes the moment the market starts pricing the 2028–2030 FCF explosion — and that repricing could happen faster than most expect. A bit like AI.

 

2. Cheaper industrials

 

One argument against owning technology stocks is the mix shift from capital-light to capital-intensive business models — the more capital-intensive, the more cyclical the earnings, and the lower the multiple applied to them. That logic shows up in the significant de-rating across many large tech stocks recently. Some now trade at multiples not far from more capital-intensive businesses like utilities — which, tellingly, share similar drivers and are often regulated too. That seems a little odd to me.

 

3. Equities are still scarce.

 

Despite the handwringing over new supply (the SpaceX IPO and subsequent lock-up expiry), fading buyback support from mega-cap tech, and Commodity Trading Advisor (CTA) strategies becoming marginal sellers, the reality is there's $1.2 trillion in authorised buybacks for 2026 — a record year if fully acted on. Buybacks should exceed equity issuance and continue absorbing supply. SpaceX has traded up since its first lock-up expired (clearly some positioned ahead of it), and vol-targeting strategies tend to buy more equities as volatility falls — and the CBOE Volatility Index is near its lowest level all year. P.S., for real asset scarcity, look at the UK FTSE, which is de-equitising via take-outs at an ever-increasing rate.

 

4. Future returns are found in the order book.

 

Granted, there's little evidence yet that incremental capex is generating higher returns — hence the overspending concerns. But it's simply too early; the rails haven't been fully laid. Still, there are signs returns are coming. Orders for many cloud providers are rising faster than capex, with book-to-bill ratios as high as 4x — extremely healthy. As Gavin Baker, managing partner and CIO at Atreides Management, discusses in this excellent podcast, graphics processing unit (GPU) shortages are now so acute that rental prices for old chips are rising faster than for new ones (spot prices are rising faster than contract) — meaning returns on those sunk costs are likely to compound quickly, good news for anyone with an installed base or excess capacity (see OpenAI and Meta's latest revenue lines). As an aside, if we already have a compute shortage while large language model and agentic AI penetration is still relatively low, what happens once we're all running a multitude of agents on a multitude of tasks? Which we will be.

 

5. Peak margins needn't mean peak earnings or a credit bust

 

One concern is that parts of the AI ecosystem may be approaching peak profitability. Memory margins may have peaked, leverage has risen, and some investors worry today's economics represent the high-water mark.

 

Perhaps. But earnings are a function of both margins and volumes.

 

Even if margins merely stabilise, volumes may continue expanding rapidly. Gavin Baker argues that higher memory demand is inevitable as agentic AI spreads, requiring more memory to make GPUs more efficient and generate more tokens. In other words, the market may be focused on the second derivative while the first derivative is still compounding.

 

Nor does the leverage concern appear quite as alarming through that lens. Debt becomes problematic when the underlying collateral is falling in value. Yet GPU scarcity remains acute, pricing continues to hold up better than many expected, and spot values for much of the underlying compute infrastructure remain firm. As contracts reset and cash flows grow, today's financing needs may prove more temporary than structural.

 

In short, the market may already be discounting the full downcycle while demand, volumes and earnings power continue to move in the opposite direction.

 

6. Circular/centralised finance is broadening and being distributed.

 

Until recently, there were only around ten players in the AI ecosystem, Nvidia central to all of them. Some degree of circular financing was unavoidable, and I can forgive the comparisons to cable roll-out in the internet bubble. But Nvidia's recent financing deal may ease some of that concern. What it's announced with the likes of BlackRock, Blackstone, Apollo and KKR should help spread funding risk (and returns) and start to decouple the circularity worries. The fact the underlying collateral value is stable-to-rising even after normal depreciation cycles doesn't hurt either.

 

7. It's not just AI.

 

I've spent a lot of time reading earnings transcripts over the years — they offer early signals on the economy and add context (and confidence) to growth forecasts. This earnings season was no exception, and beyond the extraordinary acceleration in earnings (unprecedented outside of recoveries), the standout was the breadth of companies talking about a brighter future. It's not just AI. Technology is still the driver, but where it once accounted for most earnings growth, it's now roughly half. The median stock in the Russell 3000 Index is expected to grow EPS +12% this quarter, a sharp jump from a year ago — and we see it in the commentary too. Deutsche Bank highlighted how frequently the phrase "broad-based" came up, not just in tech but across freight, steel, equipment rental, staffing, packaging and regional banks. Given our recent discussion of Alan Greenspan's Cardboard Box Index, I found the commentary from Packaging Corp. of America (PKG) instructive. Per its president, "demand was very strong throughout the quarter across our entire customer base. Shipments were up over 24% in total and per day versus last year, with the legacy business up 4.1%, achieving a record for total quarterly shipments." Capex comes with GDP multipliers — this kind of commentary, combined with small-business optimism, durable goods orders, and Purchasing Managers’ Indexes at multi-month highs, suggests growth could be about to accelerate from here.

 

8. China/open source will commoditise intelligence. And that's a good thing.

 

We expect the market for intelligence to commoditise — or bifurcate, similar to Android and Apple iOS — with China and/or open source serving mass-market AI needs, while U.S. firms hold the lead at the highest end of the innovation curve. Frontier labs may look to vertical commercialisation opportunities, but one thing is clear: They all run on the same cloud providers, so value still accrues to the compute and infrastructure layer, which is possibly why Jensen is such a fan of promoting open source. And in time, the new innovations and applications that will be built on top. Read more on this topic in the previous Weekender.

 

Of course, all of these counter-points could be just as biased as the ones they're answering. But they're worth holding up all the same, if only to see what's possible.

 

Back to the Balcony for Show 3

 

Bertrand Russell warned against dying for our beliefs because we might be wrong. Markets demand the same humility. The possibilist doesn't need to decide whether this is a bubble or a breakthrough. S/he simply weighs what's possible, updates as new evidence emerges and resists becoming captive to a single story.

 

Right now, the market's belief looks cautiously constructive: sober at the highs rather than euphoric, hedged rather than greedy, increasingly funded by order books rather than optimism alone. None of that proves anything. But it does suggest that the crowd reading about crashes still isn't the crowd causing them.

 

And from the balcony, that's enough to keep watching the third “show” before forming too strong a belief about how it ends.

 

Have a great week (and may New Zealand’s All Blacks be victorious in South Africa, in what’s termed the greatest rivalry on earth).

 

Gary