TL;DR — Alphabet spent $13.2 billion on stock buybacks a year ago. This quarter it spent zero, despite $69.5 billion in unused buyback authorization, because AI infrastructure spending now sits ahead of shareholder returns in the capital allocation queue.
A year ago, Alphabet spent $13.2 billion buying back its own stock in a single quarter. This quarter, it spent zero. That’s not a rounding error or a pause between board authorizations — the company still has $69.5 billion of buyback capacity sitting unused. Alphabet chose not to touch it, and instead raised $49.6 billion in fresh equity in June to help fund the AI buildout. The switch got a little coverage as a capital-allocation story. It’s also something else: the moment the mechanism that was quietly absorbing Alphabet’s stock compensation bill got switched off.
The Trick Buybacks Were Playing
Stock-based compensation shows up in the income statement as an expense and in the cash flow statement as a non-cash add-back, which is technically correct and practically misleading. The company isn’t writing a check, but it is handing out new shares, and new shares dilute everyone who already owns one. For a decade, Big Tech’s answer to that math problem was simple: buy back roughly as much stock as compensation was creating, and the share count stays flat, EPS stays clean, and the cost never quite shows up where anyone’s looking.
That only works as long as the buyback budget keeps pace with the compensation bill. This earnings season, for at least one of the three largest spenders on stock comp, it stopped.
Three Companies, Same Quarter, Different Bills
Alphabet’s stock-based compensation ran $8.0 billion in the second quarter of 2026, up from $6.0 billion a year earlier — a 33% increase. Layered against that, buybacks went from $13.2 billion to zero. Capital expenditures hit $44.92 billion for the quarter, ahead of the $39.07 billion the business generated in operating cash flow, which pushed free cash flow to negative $5.86 billion. The buyback wasn’t cut because Alphabet ran out of cash to spend on it in isolation. It was cut because AI infrastructure now sits ahead of shareholder returns in the queue, and the equity raise is the clearest evidence of how far ahead.
Meta’s version of the same story has a sharper edge. Stock-based compensation jumped 52% year over year to $13.69 billion, the fastest growth of the three. Total expenses rose 55% to $42 billion, part of that a $2.4 billion legal charge and $1.18 billion in severance from a May headcount reduction, but the comp line was already running hot before either of those hit. Operating income fell 8% to $18.78 billion even as revenue grew 28% to $60.8 billion, and operating margin compressed from 43% to 31%. Diluted EPS came in at $6.18, down 13% from $7.14 a year ago. Revenue up 28%, earnings per share down 13% — that gap is the dilution and cost growth showing up exactly where buybacks used to keep it hidden.
Amazon is the outlier, and the reason why is instructive. Its stock-based compensation ran $10.07 billion in the quarter, down 1.5% year over year — essentially flat. Amazon has never run a buyback program on the scale of Meta’s or Alphabet’s, so its share count has absorbed compensation-driven dilution directly for years. Investors pricing Amazon have already built that into the model. What’s changing at Meta and Alphabet isn’t the existence of dilution. It’s the disappearance of the tool that used to keep it off the page.
Why the Cover Is Coming Off Now
Capital budgets are finite even at companies that generate tens of billions in quarterly operating cash flow. When AI infrastructure spending climbs into the tens of billions per quarter, something in the capital allocation stack gets squeezed, and buybacks are the easiest line to cut — there’s no contractual obligation, no customer relationship, no data center half-built. Alphabet’s board didn’t reduce the buyback authorization, it just stopped using it, which is a decision that can reverse as easily as it was made. But for as long as it holds, the compensation bill that used to get offset dollar-for-dollar is instead landing directly on the share count.
This is the same pattern we flagged when Alphabet leaned on buyback discipline to defend its capital architecture and dividend sustainability — the buyback was never just a shareholder-friendly gesture, it was doing structural work on the balance sheet. Pull it out, and the structure it was supporting has to hold up on its own.
The Math Worth Running Yourself
The headline EPS number already nets out share count changes, which is exactly why it’s the wrong place to spot this early. Two numbers do the job better: stock-based compensation as a percentage of revenue, and the trailing four-quarter trend in diluted shares outstanding. When the first is rising and the second stops falling — or starts rising — the offset has broken down, whether or not that quarter’s EPS print has caught up yet.
Meta’s 13% EPS decline this quarter is what it looks like once the gap catches up. Alphabet’s hasn’t shown up yet, partly because years of prior buybacks already shrank the share count and a single quarter at zero doesn’t undo that overnight. A full year at zero would be a different conversation. Worth tracking either way, because the buyback authorization sitting unused at $69.5 billion means Alphabet could flip the switch back on with a single board decision — and whether it does is arguably a better signal of management’s read on the AI capex cycle than anything said on an earnings call.
We’ve written about what’s hiding on the other side of this earnings season too — Microsoft’s backlog number that didn’t make the headline, and Amazon’s profit jump that came from a markup, not the business. Different line items, same lesson: the number in the press release headline is rarely the one that tells you what actually happened this quarter.
Sources: Alphabet Inc. Q2 2026 earnings release and Form 10-Q; Meta Platforms Q2 2026 earnings release and Form 10-Q; Amazon.com Q2 2026 earnings release, quarter ended June 30, 2026. Third Pole Markets holds no position in Alphabet, Meta, or Amazon as of publication. This is not investment advice — see our About page for our full disclosure policy.
The buyback figures come directly from Alphabet’s investor relations page and its 10-K and 10-Q filings on SEC EDGAR.
