A $75 billion capital-spending plan gets cheered when investors can see the cash register. A bigger one gets punished when they are still being asked to imagine it.
That was the real split in Big Tech after earnings. GOOGL rose after reporting first-quarter revenue of $90.2 billion, up 12% year over year, while Google Cloud grew to $12.3 billion, up 28%, and operating income in the segment reached $2.2 billion from $900 million a year earlier, as shown in the company’s earnings release. At the same time, Alphabet said it still expects 2026 capital expenditures of approximately $75 billion, a large number but one now tied to a business line that is already showing meaningful operating leverage.
META delivered the opposite message. In its first-quarter release, the company said it now expects 2026 capital expenditures of $72 billion to $80 billion, up from the prior $60 billion to $65 billion range, while management also lifted the low end of 2026 total expense guidance to $119 billion from $114 billion, with the high end kept at $124 billion, per the company’s earnings materials and filing. Investors know META can print cash. What they are debating is whether the next round of AI infrastructure spending will monetize on a timetable that justifies another step-up in the bill.
The tape reflected that distinction. The Nasdaq Composite slipped about 0.3% while the Dow gained about 0.9%, a useful reminder that “tech up” and “AI up” are no longer synonyms. This is becoming a sorting machine. The market is paying up for visible conversion of AI infrastructure into revenue, margins, or both. It is discounting companies that are still in the phase of writing bigger checks and promising future payoff.
That sounds obvious, but for the last year investors often treated AI capex as a virtue in itself. That era looks shorter than many hoped. As Reuters reported, the industry’s cumulative AI spending plans are swelling toward $700 billion by 2026. Once the numbers get that large, the old “land grab” framing stops doing enough work. At that scale, investors start asking a very dull and very important question: who is earning an acceptable return on the concrete, chips, networking gear, and power contracts?
Alphabet had an answer this quarter. Google Cloud’s margin improvement matters because it suggests the company is not simply renting expensive compute to customers at thin spreads in order to win the AI beauty pageant. The cloud business is growing quickly, and the profit pool is widening. That does not make the spending cheap. It does make it legible.
Meta’s problem is not weak quality; it is sequencing. The company’s core advertising engine remains formidable. But a higher capex range and a higher expense floor tell investors that AI ambition is moving faster than near-term financial payoff. That is tolerable when ad growth is reaccelerating or when management can point to a fresh monetization wedge. It is less tolerable when expectations were already rich and the ask just got larger. Even good businesses get repriced when the duration of the payoff extends.
There is a second-order implication here for the rest of the AI stack. The winners from this split are not just the obvious cloud names. The more durable beneficiaries may be the suppliers tied to deployed, revenue-bearing workloads rather than to speculative capacity grabs. If hyperscalers keep spending, semis, networking, power, and data-center construction still have support. But the market will increasingly care which buyer is spending from operating strength and which buyer is spending on faith. Same shovel, different gold field.
That matters for valuation. A dollar of capex attached to a cloud segment growing 28% with sharply better operating income is not the same as a dollar of capex attached to a still-forming monetization story. The market is finally acting as if those dollars are different. Good. It should have been doing that all along.
What to watch: over the next two quarters, do more AI spenders start showing Alphabet-style margin and revenue conversion, or does Meta’s message become the template — bigger infrastructure budgets, higher expense bases, and a longer wait for proof?