Nvidia's guidance is already inside its customers' filings
The buyers have already told you what they will spend; the seller's Wednesday call is a formality with a stock price attached.
Four companies that buy most of Nvidia’s chips—Microsoft, Amazon, Meta, and Alphabet—completed their 2026 spending plans in the final weeks of July, and each has filed numbers that exceed Nvidia’s own projected guidance.
Nvidia will report its fiscal second-quarter results on August 26, with UBS previewing on August 15 that the market will focus on the company’s forecast—about $91 billion for the quarter. But filings show the news arrived weeks earlier. Microsoft closed its fiscal year on June 30 with Azure growing 43 percent and a contracted backlog of $678 billion.
Blockspace Media reported on July 29, citing Microsoft’s Q4 earnings, that the company spent a record $41 billion on capital expenditure in the quarter, putting the annual run-rate near $164 billion. Amazon raised its 2026 capital spending to about $220 billion during the July 30 earnings call, up from roughly $200 billion earlier this year, with Cloud Computing News reporting in August 2026 that chief executive Andy Jassy said AWS capacity remains constrained through 2027.
Meta lifted the floor of its 2026 spending range to between $130 billion and $145 billion after spending $31.1 billion in one quarter. Its Q2 2026 earnings release on July 29 showed free cash flow dropped to $784 million, the lowest in at least five years. Alphabet bumped its full-year range again to $195–$205 billion, against a Google Cloud backlog of $462 billion, per its Q2 2026 earnings release on July 27.
As a group, their intended 2026 capital spending exceeds $700 billion: Amazon near $220 billion, Alphabet up to $205 billion, Meta up to $145 billion, and Microsoft reaching approximately $164 billion annually according to each company’s Q2 2026 earnings releases between July 27 and 30.
That outlook was set before Jensen Huang speaks on Wednesday. The uncertainty lies in whether Nvidia will exceed its $91 billion guide, which UBS expects—it previewed August 15 that Rubin GPUs are ramping faster than planned—and how much upside was already priced in when the customers’ filings affected their share prices.
The trigger this week is Nvidia’s earnings print and the options market’s tendency to overreact. The underlying tension is longstanding: a handful of buyers spending faster than their revenues arrive. Meta reported $61 billion in quarterly revenue and spent more than half on data centers, according to its Q2 2026 earnings release on July 29.
Amazon attributed part of its increased spending to higher memory and component costs, Yahoo Finance UK reported in August 2026, meaning the bill for the same targets is rising.
The filings reveal diverging incentives. Nvidia wants its backlog converted before tough questions about returns arise, guiding high enough to sustain its valuation but low enough to preserve upside surprises. The hyperscalers want investors to see their spending as proof of unmet demand; each reveals a backlog—Microsoft at $678 billion, Amazon at $496 billion from its Q2 2026 earnings on July 30, Alphabet at $462 billion. Backlog figures are almost promotional tools for capital expenditure.
None, Meta most of all, can show the revenue these machines will produce. Meta’s Q2 2026 earnings release on July 29 disclosed no AI revenue line, despite spending as if it were an AI company.
Customer commitments and history
Between 1998 and 2000, Cisco’s guidance was seen as the cleanest signal for the internet economy, while telecom carriers kept signing purchase commitments till the debt backing them stopped. Filings looked like confirmation until the reversal. The lesson was not that demand was false; it was that commitments funded by borrowed confidence reverse faster than factories can adjust.
Microsoft’s cloud buildout in the 2010s featured big capital budgets against skeptical analysts, but enterprise demand and high software margins absorbed the spending. Today’s difference is who carries the risk. Microsoft then produced free cash flow to fund its data centers; Meta now does not, and Alphabet recorded its first negative free cash flow quarter in 22 years, according to its July 27 earnings release.
The constraint shifts
Nvidia’s August 26 print will land within a range the market already computed from filings, so the surprise trades smaller and after-hours moves swing both ways. The limiting factor is shifting to everything chips need: Amazon is raising spending partly due to memory cost increases, Cloud Computing News reported in August 2026. Jassy says power capacity will double by end-2027, and Amazon is already moving workloads regionally to chase electricity. Shortages extend to HBM memory, transformers, grid interconnects and construction crews, which is why Micron and utility companies follow the same story as the chipmaker.
Competition among buyers
The hyperscalers are now competing among themselves for limited input, driving costs higher and leading to another round of guidance increases justified by component inflation, not just demand. That cycle continues until someone's financing fails or their revenue catches up.
Who pays and who profits
Nvidia profits immediately, posting a record $75.2 billion in data center revenue last quarter and supporting the stock with an $80 billion buyback authorization, as shown in Nvidia Q1 FY27 investor materials from May 2026. The customers pay now and may collect later.
Tech workers absorb the squeeze on one side; Meta’s Q2 2026 earnings call on July 29 disclosed staff cuts even as machine budgets expanded. Debt holders funding data center construction absorb it twice, as repayment depends on future AI revenue lines not yet visible in income statements.
If this view is correct: Nvidia guides at or above $91 billion on Wednesday, hyperscaler backlogs continue growing faster than the revenue they are meant to generate, and at least one more buyer increases 2026 spending citing component costs before October. The break comes if any of the four trims their range or a backlog grows slower than in the prior quarter. One trim, not an Nvidia miss, is the sign—the filings move first, and the supplier’s guidance just confirms what customers have already committed to.
Nvidia’s guidance no longer reveals new information to the market that hasn’t already been disclosed by its biggest buyers.
Its forecast is a receipt for capital decisions made in Redmond, Seattle, Menlo Park, and Mountain View. When this buildout turns, consider the customers’ January filings, not Nvidia’s August call, for the clearest signal.