Big Tech earnings are quietly setting the stage for Nvidia

Over the next couple of weeks, executives in hoodies and tailored suits will sit under flattering lighting and explain why they’re spending like the future has a delivery deadline. 

On Wednesday, Microsoft $MSFT will talk about Azure and enterprise demand; Meta $META will talk about ads and ambition. Next week, Amazon $AMZN and Alphabet $GOOGL will do the same from their own corners of the cloud. The AI boom has a paper trail now: capex ranges, supply constraints, and a growing pile of analyst notes that treat every hyperscaler sentence as a demand signal. And everyone wants to see whether these companies are still talking like they’re racing ahead — committed to building out AI infrastructure at whatever scale the supply chain can tolerate — or whether the words start to soften into “discipline,” “optimization,” and “measured investment.”

Each call stands on its own — these companies move markets because they largely are the market — but together they build a live stress test for the AI supply chain. The AI trade has become a chain of inference, one where hyperscaler capex guides the mood, “AI revenue” determines patience, and Nvidia $NVDA — reporting last of the Magnificent 7 on Feb. 25 — inherits whatever conviction or doubt Microsoft, Meta, Amazon, and Alphabet leave behind.

That’s the weird little parlor trick of this earnings stretch: A chipmaker that isn’t up yet still ends up in the room, because the AI trade has turned into a high-speed game of posture-reading. Raised projections mean the AI build keeps pulling forward; careful phrasing plants AI — and AI bubble — worries weeks before Nvidia CEO Jensen Huang even says a word. Investors price the chip trade off the customers’ appetite, the customers’ confidence, and the customers’ willingness to keep buying the expensive option while they pitch cheaper alternatives.

Wall Street doesn’t need anyone to say “Nvidia” for the implication to land. If the language is about accelerators and delivery cadence, the read-through turns immediate. If the language is about power, permitting, and build schedules, the demand stays real, but the timing gets fuzzier.

Wall Street wants an answer now on the only question that has reliably moved the AI complex: Is the buildout still expanding — and at what price? Microsoft, Meta, Amazon, and Alphabet are about to provide the closest thing investors get to a preview, and their guidance will become the public scoreboard for AI infrastructure appetite. The quarter’s beats and misses will matter. The phrasing around capex — raised projections or raised doubts — will matter more.

The capex referendum

AI has turned into a corporate investment binge with a social component. One company spends because it has to; the next spends because it can’t afford to look like it isn’t. Bridgewater’s co-CIOs warned this week of a spending frenzy that’s reshaping markets and pulling the whole supply chain — semis, data centers, power — along with it. 

On the coming earnings calls, nobody wants to say “we’re slowing,” because that sounds like “we’re losing.” Nobody wants to say “we’re accelerating,” because investors now hear “accelerating” and start doing involuntary free-cash-flow math. So you get the corporate thesaurus: “capacity,” “lead times,” “procurement,” “efficiency,” “lease versus build,” “time to power,” “time to revenue.”

Microsoft is the cleanest example of the split-screen problem. The Street wants Azure strength and AI revenue lift, while simultaneously demanding reassurance that the bill isn’t compounding faster than the benefits. A Wedbush note reads like a fan letter written in spreadsheet form — “another quarter of robust results,” with Azure growth “beatable,” plus the reminder that “AI capex … is expected to continue accelerating throughout FY26” and that Microsoft is “fully committed to the AI buildout.” 

TD Cowen lands on the other side of the emotional spectrum — not bearish, just unimpressed by the near-term physics. It “suspect[s] lack of growth acceleration due to capacity constraints will keep shares range-bound,” even while it stays “constructive” on Microsoft’s position for AI workloads. Same story, different temperature. And a target-cut drumbeat also shows the market mood; UBS kept its Buy on Microsoft but cut its price target to $600 from $650 on Jan. 23. “Good” results can still land in a “prove it” tape.

Meta is the cleaner capex referendum because the business model is simpler: Ads pay the bills, and Mark Zuckerberg decides how big the bills get. Evercore ISI expects Meta to deliver a “Modest Beat & Bracket Q4” but flags “potential risk to Wall Street’s 2026 total expenditure and capital expenditure assumptions.” That’s analyst-speak for the thing everyone is actually trying to gauge on the call: whether Meta can talk about spending without sounding like it’s reenacting a yacht scene from a movie about hedge funds.

Investors largely believe in Meta’s ad machine, but they’re still deciding whether Meta’s AI spending plan deserves the same automatic trust. One bullish case from Rothschild & Co Redburn floating into the print leans hard into Meta’s ability to turn AI into better advertising, framing the system as a “demand machine,” with the idea that the market is underestimating how much runway that creates. The flip side is that the same upgrade logic quietly admits the pressure point: The more Meta sells itself as an AI company, the more it inherits AI-company scrutiny — returns, timelines, and whether “we’re building” is code for “we’re still — yes, still — searching for the payoff.” (And Meta is already getting a lot of scrutiny.)

Now fold in Amazon and Alphabet, and the capex referendum gets more interesting for Nvidia because both companies are trying to pull off the same stunt: Keep buying the best shovels while also insisting they’re building their own. Because Nvidia’s quarter gets shadow-priced first, then reported.

Amazon has Trainium for training and Inferentia for inference — purpose-built accelerators pitched as a way to give customers more compute options and better economics inside AWS. Google, meanwhile, has TPUs — custom AI accelerators that it positions as the backbone for training and inference across its AI products. Both companies want pricing leverage and supply resilience, and both still benefit when demand is so large that every kind of compute gets pulled into the room. The earnings calls are where they’ll have to sell that story without sounding like they’re quietly trying to wean themselves off the very supplier — hello, Nvidia — that investors keep using as the scoreboard.

The receipts problem

The market used to treat capex like a bet. Now, AI has turned capex into a subscription — recurring, escalating, and now constantly audited. So the key tell across these four calls will be more than whether or not these companies beat consensus expectations; it will be whether they can credibly connect spend to monetization in a way that buys them another quarter of patience.

Microsoft has the most obvious monetization narrative because it sells software and has a captive enterprise base that will pay for add-ons if the add-ons feel mandatory. That’s why the Street keeps coming back to Azure, Copilot, and whether “AI services” are lifting revenue in a way that looks durable — not a temporary capacity sugar high. Meta’s receipt is different. Meta’s receipt is credibility. Wall Street likes an ad machine. Wall Street gets twitchy when the capex story starts sounding infinite. Meta is already an ad cash machine, which means the question becomes whether AI spend is improving the thing investors already trust (ad performance) and whether the company can explain “frontier” ambitions without spooking the part of Wall Street that prefers its innovation packaged as margin expansion.

Amazon’s receipts come through AWS and operating leverage. This quarter’s bullish framing is essentially: The AWS growth engine looks capable of re-accelerating, and efficiency work in retail (and tons of job cuts) can keep cash generation intact even while the company spends aggressively on AI and cloud. Barclays analyst Ross Sandler said, “AWS has secured significant AI capacity over the next several years.” The company is selling time — the ability to deliver compute when everyone else is still talking about construction schedules. Analysts at Bernstein argued in a January client note that AWS “regained the lead in net new dollars added” and called Amazon a top large-cap tech pick for 2026. That’s essentially the narrative Amazon needs in today’s market: Yes, it’s spending, but it’s spending into a demand curve that’s widening, and it can still act like a grown-up business while doing so.

Alphabet has the cleanest “stack” story, and it’s increasingly being sold as a reason its AI economics could look better over time: Search distribution, Cloud growth, models, and custom silicon all inside one corporate organism. Cantor Fitzgerald’s Deepak Mathivanan upgraded Alphabet and called it the “king of all AI trades,” explicitly pointing to its footprint “across the entire AI tech stack.” But Google’s receipts problem isn’t whether it can spend; it’s whether Cloud and AI-driven monetization show up fast enough to keep that level of spending feeling like strategy rather than inertia.

Ads still pay for a lot of Alphabet’s future. Cloud is the growth engine investors want to believe in. But Alphabet also carries the most obvious strategic vulnerability: If the AI internet changes how people search, the company has to spend to defend a cash machine — all while building the next one. That’s not a moral failing. It’s just expensive. Stifel, in a digital ads Q4 preview, said it prefers Alphabet into the print with “solid fundamentals that are likely ahead of Street,” while calling out a “wildcard” tied to incremental releases from OpenAI — then adding, almost with a shrug, “that’s something to worry about after the print.”

What matters (or changes) before Nvidia reports

The market treats Microsoft and Meta as direct pipelines to incremental chip demand — hyperscale training builds, enterprise AI rollouts, and the continuing “we need more — and more! — capacity” drumbeat. Amazon and Alphabet widen the set of “capex referees” and can confirm (or puncture) the idea that AI demand is broad-based.

Over the next month, investors don’t just want to know whether the bill is rising. They want to know what’s inside the bill, because different spend tells different Nvidia stories. By the time Nvidia reports on Feb. 25, the market will already have built a narrative out of these four sets of guidance and a stack of analyst notes. The question is: What kind of narrative will that be?

One version is raised projections.

That’s the world where Microsoft talks like capacity is easing into demand, Meta keeps the ambition but introduces sharper boundaries, Amazon frames AWS investment as the rational response to signed demand, and Alphabet shows strong fundamentals while keeping the AI spend story contained enough that investors don’t hear an endless construction project. In that world, the proxy print stays bullish for Nvidia’s setup. Customers are still hungry, still ordering, still talking in forward-looking sentences. This is also the world where the “alternatives to Nvidia” storyline stays exactly where Nvidia investors prefer it: as a long-term margin-management project, not a near-term demand re-route.

The other version is raised doubts.

That’s the world where language starts to tighten. “Pacing” becomes the star. “Discipline” stops being a reassuring adjective and starts sounding like a constraint. Capex ranges hold steady but tone changes. The notes follow: more emphasis on timelines, returns, competition, and what gets delayed. In that version, Nvidia doesn’t lose the quarter in advance, but it does lose the benefit of the doubt. The market starts acting like the best days of “everyone buys everything” are being replaced by “everyone buys, but everyone bargains.”

That’s how you end up with a chipmaker’s report being previewed in everyone else’s earnings call. Not because it’s the only story that matters, but because right now, it’s the one story the market is trying to price before the receipt arrives.

The next two weeks won’t settle the AI bubble debate. They’ll do something more practical. Microsoft, Meta, Amazon, and Alphabet will report as four separate empires with four different profit engines. But they’ll testify — sometimes willingly, sometimes through clenched teeth — about the same thing: whether the AI buildout still deserves to be treated as inevitable, and whether Wall Street is still willing to finance it on faith. And Nvidia will inherit that tone the moment its calendar slot becomes a trade again.