The moment is upon us.

The moment when tech capex overtakes buybacks and dividends. The AI arms race well and truly changed the game.

It’s been a while. More than two decades, in fact, since spending for growth outstripped shareholder returns across the US tech sector.

The figure above, from SocGen’s Manish Kabra, shows you three decades of history for tech capex, buybacks and dividends.

This is a scary thing. Heretical, even. This is shareholder capitalism. Or at least it used to be. Returning cash to stockowners should take precedence every, single time over investing for growth. The largest companies on Earth seem to have their priorities backwards all of a sudden.

We’re witnessing, across big-tech, a metamorphosis of historic import. Asset-light, debt-free cash machines are morphing into a giant, leveraged moonshot on a technology even proponents concede might become self-aware and kill us all. And I’m forced to fund this existential lark out of my buybacks?!

Some (most) of the above’s tongue-in-cheek, but underneath the snark there is indeed a fundamental shift afoot. The good news for those concerned about suddenly profligate hyper-scalers is that, as Kabra remarked, “residual cash flow after capex and distributions remains at cycle highs” at the sector level. You can see that on the right, below.

On the left is overall tech operating cash flow. At $1.2 trillion, it’s never been higher. “At the aggregate level [there’s] sustained surplus liquidity,” Kabra said.

Although this is intuitive — I’d call it “self-evident,” but Nvidia aside, how many of us put our life savings into semi names in 2024 and 2025 while calmly explaining that, “All that capex has to go somewhere!” — it’s worth pointing out that big-tech outlays are accruing directly to semi bottom lines.

The figure below, from Goldman, shows you the historic relationship between hyper-scaler free cash flow growth and semi profit growth. The two aren’t “going their separate ways,” exactly. Rather, the divergence is a function of the just-described “mirror image” dynamic.

As Goldman’s Ben Snider wrote, semiconductor earnings are “rising at the expense of hyper-scaler free cash flows.” You can take that figuratively and literally.

In his Thursday piece, SocGen’s Kabra wrote that “dispersion in tech is increasingly an AI financing story [with] hyper-scalers running a combined $180 billion cash flow deficit, as capex broadly exhaust[s] operating cash flow, pushing new financing needs higher.”

The table below’s another way to visualize the interplay between “spenders” (the hyper-scalers) and “beneficiaries” (semis).

All the major spenders “screen positive for incremental funding requirements [and] this deficit is effectively matched on the other side of the ecosystem,” Kabra said.

“In effect,” he went on, “AI capex is driving a transfer of cash within the value chain from hyper-scalers funding the build-out to semis monetizing it.”

If the self-evident nature of this makes you wonder why you’re not a billionaire in 2026 off the lottery ticket, deep OTM semi calls you didn’t buy last year, just remember: Everything‘s obvious in hindsight.

An important corollary says that most people who tell you they acted then on something that’s obvious now are lying. Because if they were telling the truth, they’d be somewhere spending their money, not bragging to you about making it.