Five numbers to watch when Microsoft, Alphabet, Meta, and Amazon report Q1 earnings on April 29, and what each one actually means for the next 12 months of this story. On April 29, 2026, Microsoft, Alphabet, Meta, and Amazon report Q1 earnings in a single 24-hour window. Together, they’ve committed more capital to AI infrastructure this year than the entire GDP of Sweden. Today is the day the market stops taking that on faith. That’s not framing. That’s the arithmetic. Now for the atmosphere, because it matters too. There’s a particular silence that falls over a trading floor the night before a verdict. Not fear, exactly. Something closer to the suspension of disbelief. The last moment before the story either holds together or comes apart at its seams. I’ve watched enough of these inflection points over the years following AI infrastructure spending from the first GPU cluster announcements through to the current era of hundred-billion-dollar annual commitments to know that the silence is never quite as calm as it looks. The anxiety just migrates from the trading floor to the analyst note, from the analyst note to the earnings call prep meeting, from the prep meeting to the exact phrasing a CFO chooses when someone asks about return on invested capital. Tomorrow, four CFOs will be asked that question simultaneously. The phrasing will matter. How We Got Here and Why This Week Is Different The AI spending era unfolded in recognizable phases, even if nobody labeled them clearly while they were happening. Phase One was conviction as currency. From late 2022 onward, every major tech executive made the same pilgrimage to the same earnings call microphone and delivered some version of the same sermon: AI is transformative, we are investing aggressively, the returns will be enormous. The market listened. Valuations soared. Nobody demanded receipts because nobody needed them. The size of the bet was treated as evidence of the quality of the intelligence behind it. Phase Two was the capital itself. The capex budgets that followed weren’t just aggressive they were historically unprecedented. The four hyperscalers alone committed roughly $645 billion in infrastructure investment for 2026, a 56% jump from the prior year. Data centers multiplied across three continents. GPU clusters expanded to scales requiring their own dedicated power grids. NVIDIA’s market capitalization eclipsed every benchmark anyone had previously set for it. The spending was the validation. The bigger the bet, the more serious the believer. What starts tomorrow is something else entirely. Phase Three has no inspirational framing. It’s the reckoning. Investors are no longer asking whether these companies believe in AI. They’re asking whether AI believes in them specifically, whether AI-related revenue is scaling at a pace that can plausibly justify the size of the infrastructure being constructed to support it. Whether the enterprise customers who signed contracts are genuinely deploying the products or running pilots that never convert. Whether margins are holding as depreciation schedules on $200 billion in annual spending begin to compound. Vision statements don’t answer those questions. Numbers do. The Scorecard Let me be direct about what each company is walking into tomorrow. Microsoft: The Company with the Most to Prove Microsoft has committed close to $146 billion in AI and cloud infrastructure for fiscal 2026, with projections for 2027 already drifting toward $170 billion. Azure grew 40% year-overyear last quarter, a number that would define another company’s entire legacy. None of that is the issue. The issue is 17 cents. Microsoft has targeted roughly $25 billion in AI-related revenue for the fiscal year. On $146 billion in capital deployed, that’s approximately 17 cents of revenue for every dollar spent, and that’s before depreciation, before operating costs, before the upgrade cycle is already scheduled. The math is not comfortable. The market has been doing it quietly for months, which is the most dangerous kind of arithmetic: when it finally surfaces loudly, the move is violent. What tomorrow makes unavoidable is the Copilot question. Microsoft 365 Copilot, priced at $ 30 peruser per month, is the company’s primary near-term AI monetization vehicle. Not the most ambitious one. The primary one. If enterprise customers are paying for it, deploying it broadly, and using it with the frequency that justifies the price point, the spending story survives another quarter. If adoption is tepid, if the purchasing data shows companies buying seats they’re not filling, no slide explains that away. The company has prepared many impressive slides. Azure growth guidance is the second number to track. Last quarter pointed toward 37–38% growth in the coming period. A beat tells you AI enterprise demand is real and pulling customers toward the cloud. A miss is more complicated: it suggests the infrastructure build is running meaningfully ahead of actual customer adoption, which is exactly the pattern that precedes every overcapacity narrative in technology history. Alphabet — Defending Two Businesses at Once Of the four companies reporting tomorrow, Alphabet occupies the most structurally precarious position. It is simultaneously one of the largest beneficiaries of AI infrastructure spending and one of the businesses most existentially exposed to what that spending eventually produces. The same technology driving Google Cloud’s backlog to $243 billion may be quietly dismantling the ad-click revenue model that built the company. The good news is real. Google Cloud surged 47.8% in Q4 2025, hitting a $70 billion annual run rate. The backlog figure alone represents a kind of financial certainty most companies never experience. Analysts broadly expect Q1 cloud growth to accelerate, which would be the clearest available signal that AI workloads are generating genuine enterprise demand rather than experimental budgets that look like revenue until the CFO reviews them. The question Sundar Pichai cannot avoid tomorrow is AI Overviews. Google’s decision to answer search queries directly, serving the answer inside the results page rather than routing users to a publisher’s website, is either a masterclass in retention or a slow-motion dismantling of the mechanism that generates revenue. It might be both. Q1 will be the first time Pichai has to address that tradeoff with a full quarter of data sitting behind him, and the data will be interpreted uncharitably if Search revenue shows even modest softness. Watch Search revenue growth carefully. If it decelerates while Cloud accelerates, what you’re watching is AI cannibalizing Google’s highest-margin product to grow its second-highest-margin one. That’s a structural transition the market hasn’t fully priced. It may not panic about it tomorrow. But it will remember it. Meta The One That’s Actually Working Meta is the anomaly in this earnings cycle and understanding why matters more than the headline numbers will. Analysts are projecting $6.65 EPS on roughly $55 billion in revenue, 32% year-over-year growth. There are zero sell ratings across 42 covering analysts. That kind of consensus is unusual enough to require explanation rather than simple acceptance: Meta’s AI story isn’t aspirational. It’s already in the income statement. AI-powered ad targeting has made Meta’s advertising engine materially more efficient, and the revenue improvement shows up in quarterly numbers rather than future guidance language. It’s the difference between “we believe this will generate returns” and “here are the returns, counted.” This is why Meta matters so much to everyone else reporting tomorrow. It’s the proof-of-concept case. The concrete, auditable evidence that the broader AI spending thesis isn’t just a story someone is telling. If Meta beats expectations again, it extends the runway for Microsoft, Alphabet, and Amazon to keep building without delivering equivalent returns on their own timelines. If Meta misses the strongest argument for AI monetization that cracks the narrative, it doesn’t just weaken. It loses its load-bearing wall. One thread worth watching beyond the headline: Meta has committed to deploying 1 gigawatt of custom MTIA chips built with Broadcom on a 2-nanometer process. This is a direct architectural bet against Nvidia dependency. If that chip strategy is scaling ahead of schedule, the $115–135 billion capex plan starts looking considerably more defensible than it does as a raw number on a spreadsheet. Amazon The One Spending the Most, Saying the Least Amazon’s 2026 capital expenditure plan tops $200 billion. Nearly 60% above 2025 levels. A number that makes Microsoft’s $146 billion commitment look, in relative terms, almost restrained. The vast majority goes toward AI infrastructure data centers, proprietary silicon, and the high-speed networking fabric that makes all of it function at scale. AWS grew 24% year-over-year last quarter. The threshold that matters tomorrow is whether that growth clears 20% again with AI workloads as the identifiable driver, and whether the margin picture is holding as the capex cycle intensifies. Amazon’s Q4 2025 net income already missed expectations partially because of the spending load, and the market absorbed it with reasonable equanimity. Forgiveness has conditions. The implicit bargain in excusing an earnings miss for infrastructure reasons is that the infrastructure eventually produces the returns that justified the forgiveness. That bargain has an expiration date, and it gets closer each quarter. One more item: any update on the Anthropic relationship? Amazon has invested heavily in the company, and the strategic question of whether that investment translates into durable differentiation in AWS AI services has quietly become a piece of the investment thesis. An update, in either direction, changes the picture more than most analysts are currently pricing. The Uncomfortable Math Nobody Is Saying Out Loud Step back from the individual company stories. The aggregate picture deserves more attention than it’s getting. $645 billion in 2026 AI infrastructure spending. Across four companies. In a single year. That number has to generate returns not eventually, not in the medium term, but on a timeline that compounds against the spending itself, because each year’s capital outlay builds on the last. Microsoft’s $25 billion AI revenue target against $146 billion in spending yields 17 cents on the dollar before depreciation. Alphabet’s $175–185 billion capex commitment requires Google Cloud to sustain 40%+ annual growth just to maintain the internal logic of the investment thesis. Amazon’s $200 billion is the most aggressive single-company capital allocation decision in the history of the technology industry. Here’s what makes AI infrastructure spending uniquely unforgiving compared to previous cycles: it doesn’t pause. The GPU clusters commissioned today require the next generation of GPU clusters to remain competitive. The models that justify the infrastructure today will be functionally obsolete in 18 months. Depreciation schedules are long; competitive upgrade pressure is relentless. The economics of the next cycle begin before the current one has paid out. The math works if AI revenue compounds faster than AI spending. The spending is clearly compounding. Whether the revenue is moving fast enough at real enterprise scale rather than in press releases and pilot programs is precisely what tomorrow’s calls are designed to reveal. I’ve followed this industry long enough to know that the optimists usually win over a long enough horizon. I’ve also been around long enough to know that “a long enough horizon” has quietly buried a great many investors who were entirely correct about the technology and entirely wrong about the timing. Both things can be true simultaneously. They often are. This pattern companies are restructuring their workforce while redirecting capital toward AI infrastructure, has been building quietly for years. I wrote about the mechanics of it last week in “People Lost Their Jobs. Oracle and DeepSeek Did It In the Same Week.” And the layoffs that funded some of this spending weren’t always announced. See: “Nobody Got Fired. The Desk Just Never Got Refilled.” What to Actually Watch When the Numbers Drop Ignore the headline beats and misses on first read. Five data points will tell you more about the next 12 months than any single EPS figure. 1. Azure Growth vs. the 37–38% Guidance Corridor Microsoft guided to 37–38% Azure growth for the coming period. A beat means real AI enterprise demand is pulling customers toward the cloud on its own commercial merits. A miss particularly a meaningful one means the infrastructure build is running ahead of actual customer adoption. That phrase, precisely stated, is the definition of overcapacity in formation. Every major technology cycle that ended badly had a moment where that sentence was true before the market acknowledged it. 2. Copilot Paid Seat Engagement (Not Just Seat Count) Deployment is not adoption. Paid seats are not daily active users. The metric that matters isn’t how many companies bought licenses it’s how many employees are opening the product every morning because it has materially changed how they work. Watch for any engagement data Satya Nadella volunteers without being asked. A CEO who doesn’t mention it has already answered the question. 3. Google Search Revenue Trajectory A deceleration in Search revenue, even a modest one, alongside Cloud acceleration, is a structural story wearing an operational costume. The market may not price it aggressively in a single quarter. But if that’s what the numbers show, the implications extend well beyond Q1 2026 and the analysts who are currently treating it as a transition rather than a threat will eventually need to revise that view. 4. Meta’s Capex Guidance Update Zuckerberg has been the most aggressive AI spender relative to near-term revenue justification among the four. A guidance increase signals conviction backed by internal data. Wall Street hasn’t seen the kind of data that makes a CEO comfortable doubling down when everyone else is watching nervously. Any trim at all is the most important tell of the day, regardless of what the headline revenue numbers look like. 5. Any Company That Guides Down on AI Spending This is the scenario nobody has priced in. Not one analyst report, not one consensus model, has seriously incorporated the possibility that one of these four companies signals a slowdown in AI infrastructure investment. If it happens and the probability is low, but it is not zero , the narrative doesn’t modulate. It inverts. “AI is inevitable” becomes “AI has a ceiling,” and the repricing is immediate and severe. The correction wouldn’t be limited to the four companies reporting. It would cascade through every company whose valuation carries an embedded AI premium, which, at this point in the cycle, is most of the market. Two Scenarios for Tomorrow Afternoon The bull case: all four companies beat on revenue, cloud growth accelerates across the board, Copilot adoption surprises upward, and at least one company raises full-year AI revenue guidance. The spending narrative shifts decisively, in a single news cycle from faith-based capital expenditure to proven infrastructure investment. $645 billion looks prescient in retrospect. Phase Four begins. The bear case: numbers meet but don’t exceed estimates, guidance is cautious, Copilot engagement disappoints, and at least one company signals margin compression without commensurate revenue growth. The AI bubble narrative, which has been building quietly in the corners of financial commentary for months, goes mainstream overnight. The tech selloff of 2026 begins and history records, with some irony, that it started at exactly the moment the technology itself was most capable. The bull case is more probable. The bear case, if it materializes, is more consequential. Those two sentences are not contradictions. They’re the actual structure of the risk. What This Is Actually About Strip away the EPS estimates, the capex ratios, the guidance ranges, and the analyst commentary, and what remains is a simpler story. Four of the most powerful companies in human history made the largest collective corporate bet in history. They wagered that artificial intelligence would restructure computing, enterprise software, drug discovery, advertising, and eventually every industry that touches the digital economy. They committed hundreds of billions to this future. In the same breath, many of them laid off tens of thousands of workers, redirecting those savings toward infrastructure for a transformation that hasn’t fully arrived yet. Tomorrow, the market gets its first serious look at whether the bet is paying off. Not someday. Not in the medium term. Right now, in Q1 2026, the actual earned dollars are measured against actual spent dollars. One quarter is never the whole story. No verdict tomorrow will be final. But it establishes the tone for the next 12 months of the AI investment narrative and for the enormous ecosystem of workers, developers, investors, and executives whose professional futures are now bound to whether this technology delivers on its extraordinary, expensive, still-unproven promise. $645 billion is a lot of money to spend on a question. Tomorrow, we start getting the answer. Bull or bear going into tomorrow’s earnings? Drop your call in the comments. I’ll be back tomorrow evening with the full post-earnings breakdown. Follow me here so you don’t miss it. A message from our Founder Hey, Sunil here. I wanted to take a moment to thank you for reading until the end and for being a part of this community. Did you know that our team run these publications as a volunteer effort to over 3.5m monthly readers? We don’t receive any funding, we do this to support the community. If you want to show some love, please take a moment to follow me on LinkedIn , TikTok , Instagram . You can also subscribe to our weekly newsletter . And before you go, don’t forget to clap and follow the writer️! Big Tech Spent $645 Billion on AI. Tomorrow, We Find Out If Any of It Worked. was originally published in Artificial Intelligence in Plain English on Medium, where people are continuing the conversation by highlighting and responding to this story.