Meta and Microsoft report on July 29 with the same problem in different clothes. One has to explain as much as $145 billion in planned capital spending, the other has to prove a $37 billion AI run rate can carry a much larger buildout.
Mark Zuckerberg was asked last quarter how Meta would know whether its AI investments were producing a healthy return. His answer was short: "That's a very technical question." That line has now become the problem. Meta reports second-quarter results after the market closes on Wednesday, July 29, and investors don't need a grand theory from Zuckerberg. They need a clearer answer than that.
Meta lifted its 2026 capital expenditure guidance after its April 29 first-quarter results to $125 billion to $145 billion, up from $115 billion to $135 billion. The company said in its SEC filing that the spending would support its AI efforts and core business. That's the accurate phrasing. Calling the whole number pure AI infrastructure overstates it, but nobody should pretend the market is confused about the main driver. The money is following servers, data centers, component costs and capacity for the AI cycle.
The first-quarter numbers were strong. Meta reported $56.31 billion in revenue, up 33% from a year earlier, and guided second-quarter revenue to $58 billion to $61 billion, with foreign exchange expected to add about two percentage points. The stock still sold off after the April report because the spending line moved faster than the comfort level. Investors can forgive heavy spending when the return is visible. Meta's return is still being described through advertising efficiency and engagement.
That's harder to underwrite.
Microsoft reports the same evening, and its position is cleaner. In its fiscal third quarter, Microsoft said its AI business had passed a $37 billion annual revenue run rate, up 123% from a year earlier. Azure and other cloud services grew 40% in constant currency, and Microsoft said demand continued to exceed available capacity. That is a good problem, but it isn't free. On the same call, CFO Amy Hood said fourth-quarter capital expenditures would rise to more than $40 billion, including roughly $5 billion from higher component pricing.
The spending keeps moving up
Meta and Microsoft aren't isolated cases. According to analysis published by Value Add VC, Microsoft, Amazon, Alphabet and Meta were already on pace for roughly $725 billion in combined 2026 capital spending before Alphabet's latest increase. Alphabet has now raised its own 2026 capex guidance to $195 billion to $205 billion after reporting second-quarter results on July 22, according to MarketWatch and other outlets. The market reaction was blunt. Alphabet's stock fell because investors saw strong revenue and still focused on the cash leaving the building.
Goldman Sachs Research has put the larger number at $5.3 trillion in capital spending from 2025 through 2030 by the largest technology companies leading the AI and data center buildout. That isn't a normal investment cycle. It's a corporate spending race that touches chip supply, power infrastructure, data center leases and debt markets. If you're building anywhere downstream of those companies, you are living inside their price system whether you like it or not.
Moody's has also warned about what sits outside the usual headline capex figures. As Fortune reported on Moody's analysis, Amazon, Meta, Alphabet, Microsoft and Oracle had $969 billion in total undiscounted future lease commitments tied largely to data centers, with $662 billion not yet commenced and therefore not yet recorded as current balance sheet liabilities. Moody's calculated that the unrecorded piece equaled 113% of the group's adjusted debt. Dry accounting detail, yes. Also the kind of detail that tells you the AI buildout is not only an income statement story.
Meta has the harder explanation
For Microsoft, the question is whether the AI revenue line keeps growing fast enough to justify the capacity being added. Azure OpenAI Service and GitHub Copilot give investors something direct to measure - Copilot alone is showing up in renewal conversations across enterprise. You can argue over margins and payback periods, but at least the revenue has a label on it. Microsoft's own transcript says AI contributed to its cloud results, and the company has put a real run-rate number in public.
For Meta, the answer runs through a blurrier route. The company can use AI to improve ad ranking, targeting and creative tools, and to lift time spent across Facebook, Instagram, WhatsApp and Threads. Those are real levers. They are also harder for shareholders to separate from the rest of the advertising machine. A stronger ad quarter doesn't automatically prove a $125 billion to $145 billion capital budget is earning its keep.
Frankly, Meta needs a plainer answer. Zuckerberg doesn't have to give a product-by-product spreadsheet, but investors deserve the markers he is watching: better ad pricing, higher conversion rates, more commercial messaging revenue - or paid AI products gaining real traction. If the answer is still mostly faith in future products, the market will hear that.
Startups should pay attention because these numbers don't stay on Big Tech balance sheets. A founder trying to hire machine learning engineers or rent GPU capacity is competing in a market shaped by Microsoft, Meta, Amazon and Alphabet. When the largest buyers push up component prices and reserve capacity years in advance, smaller companies pay in slower launches and worse negotiating power.
That is why July 29 matters. Microsoft has to show that its AI business is not being drowned by the infrastructure bill. Meta has to show that its advertising engine can turn massive capital spending into something investors can see. The question Zuckerberg called technical is really quite plain: what exactly are you buying, and when does it pay?
Also read: Prentis AI lab co-founded by Reid Hoffman and Marc Pincus is in talks to raise $100 million • Verizon signs a billion-dollar dark fiber deal with Google and says several billion more are coming • Kevin Warsh wanted markets to read the data and now the data is reading him
