$220 Billion and Counting: Amazon, Google, Microsoft, Meta All-In on AI — The Hyperscaler Capex Verdict

By Yogurt · 2026-08-02 · Earnings Analysis

All four hyperscalers reported Q2 2026 earnings this week — and the verdict is unanimous: AWS growing 37%, Google Cloud up 82%, Microsoft Azure accelerating, and a combined CapEx commitment approaching $300 billion for the year. Demand still outstrips supply. Here's what that means for the AI trade.

The Four Giants Have Spoken — and They're All Saying the Same Thing

With Nvidia the sole exception among the hyperscale AI infrastructure players, Q2 2026 earnings season delivered the reports investors have been waiting for all summer. Amazon, Alphabet (Google), Microsoft, and Meta — the four companies Micha Stocks calls the "monster" hyperscalers — all delivered their quarterly numbers, and the headline story isn't just the beats. It's the size of the bets they're placing on AI.

These four companies alone will collectively spend approximately $73 billion on infrastructure in Q3 2026 — data centers, chips, energy, networking. For the full year, Amazon alone has committed to spending $220 billion in capital expenditure. To put that in context: Amazon's entire annual revenue was below $220 billion just two years ago. Now, that's what they're spending in a single year to build the future.

And here's the critical detail that separates this capex cycle from prior tech spending booms: the CEOs aren't just spending — they're doing it because demand for AI cloud services is running ahead of their ability to build capacity fast enough.

Andy Jassy's Stunning Admission: Supply Can't Keep Up With Demand

Amazon CEO Andy Jassy made a remarkable statement during the earnings call that encapsulates the entire AI infrastructure moment. Despite committing $220 billion to capital expenditure in 2026, Jassy said:

"Even at this level of investment, we still won't have enough capacity to meet all the demand we have in 2026."

He went further: he believes this demand-supply imbalance will persist into 2027 as well. The backlog of enterprises waiting to migrate workloads to AWS AI infrastructure — what Google Cloud calls "the pipeline" — stands at $514 billion in identified potential revenue yet to be activated. These are real customers with signed letters of intent, waiting for compute capacity to become available.

This isn't hype. This is a constraint problem. Amazon can't build fast enough. Neither can Google. Neither can Microsoft. When the world's largest and best-managed companies say they cannot fulfill their own customers' orders, that's a signal worth taking seriously.

The Numbers: Revenue, Growth, and Margins

Here's the scoreboard for Q2 2026 across all four hyperscalers:

Amazon ($AMZN): $200 billion in total revenue — the first $200B quarter in history — up 20% year-over-year. AWS grew at 37% YoY, its fastest pace in 18 quarters. Operating margin expanded to 39%.
Alphabet/Google ($GOOGL): $120 billion in revenue, up 24% YoY. Google Cloud grew at 82% year-over-year — an extraordinary acceleration that partly reflects the integration of Waze and other acquired assets into the cloud revenue line. Operating margin expanded from 20% to 35%.
Microsoft ($MSFT): $90 billion in revenue, up 18% YoY. Azure AI services continued accelerating. Microsoft broke above key technical resistance levels on the report.
Meta ($META): $60 billion in revenue. Despite strong underlying business growth — Meta now has 3 billion daily active users across its family of apps — the stock initially fell because investors couldn't fully understand the rationale for Meta's aggressive infrastructure spending, which Zuckerberg described as investing because "the potential is huge."

Notice the pattern: not one of these companies is shrinking margins while investing. Amazon is at 39% operating margin. Google went from 20% to 35%. These are not companies burning cash in desperation — they're investing from a position of strength, with expanding profitability funding the next leg of growth.

Going Cash Flow Negative: A Feature, Not a Bug

All four hyperscalers are in the process of becoming — or have already become — free cash flow negative as their capital expenditure exceeds operating cash generation in this investment phase. Historically, that would be a red flag. In this cycle, the market is treating it as a sign of confidence.

The reasoning from every CEO on every earnings call was essentially identical: we are investing now because the demand is real, the monetization timeline is clear, and the ROI on infrastructure built today will compound for decades. Andy Jassy noted that Amazon's data center equipment installed five years ago is still running profitably today — meaning each dollar of infrastructure spending generates revenue for years, not quarters.

The key distinction between hyperscalers who were rewarded by the market and Meta — which was initially punished — was communication. Amazon and Microsoft told a precise story: here's the demand, here's the capacity gap, here's when the investment pays off. Zuckerberg's message was more general: the opportunity is enormous, trust us. The market's differential reaction wasn't about the investment thesis — it was about the clarity of the explanation.

Google Cloud's 82% Growth — and What Waze Has to Do With It

Google Cloud's 82% year-over-year growth is the most striking single number in this earnings cycle. For context, AWS — the market leader — grew 37%. Azure grew in the mid-30s. Google Cloud was growing at more than twice that rate.

Part of the explanation may be the 2025 acquisition of Waze, the Israeli navigation app. Waze brought not just technology but a substantial enterprise customer base — particularly municipalities, transportation agencies, and logistics companies — that appears to be flowing through Google Cloud's revenue line. The combination of Waze's customer access and Google's AI platform (Gemini, TPUs, Vertex AI) created a multiplier effect on cloud adoption that competitors couldn't match in Q2.

Google Cloud also benefits from an infrastructure advantage that doesn't get enough attention: Google's custom Tensor Processing Units (TPUs) allow it to run its own Gemini models faster and cheaper than any third-party GPU can replicate. When customers run Gemini on Google Cloud, they're not just buying compute — they're buying the most optimized possible runtime for the model they're using. That's a structural cost advantage.

Microsoft Azure: The Satya Thesis

Microsoft CEO Satya Nadella delivered what analysts called one of the clearest forward-looking narratives of any earnings call this season. His framing: Microsoft is in the middle of a "reset for long-term growth," embedding AI across the entire commercial stack — Azure, Office 365, Dynamics, GitHub Copilot — so that every enterprise customer automatically consumes more AI the more they use Microsoft's products.

The market responded to that clarity. Microsoft stock broke through a series of declining highs on the back of the earnings report and extended momentum into $466–$470 resistance. The technical picture mirrors the fundamental one: a company that has been building a base for months and is now showing the first signs of a genuine breakout.

Microsoft's Azure AI infrastructure — which uses a combination of Nvidia GPUs and AMD alternatives — is benefiting from the same demand-supply dynamic that's driving Amazon and Google's capex surge. Enterprise workloads that were in AI experimentation mode 12 months ago are now moving into production. That transition generates a step-change in compute consumption: a production AI workload uses 10 to 100 times more GPU time than a pilot.

The Downstream Trade: New Cloud Companies and the Energy Bottleneck

The hyperscalers' capex commitments don't stay inside their balance sheets. They flow downstream into the entire AI supply chain — and understanding where they flow matters for investors in adjacent sectors.

The most direct beneficiaries are the companies that sell what the hyperscalers are buying:

Memory and semiconductors: Micron, Samsung, SK Hynix supply the high-bandwidth memory that goes into every AI chip. Each $10 billion of hyperscaler data center buildout requires billions in HBM purchases. Micron's stock went from $103 to $1,215 in roughly a year — a 12x move — as the market pre-priced this demand surge.
New Cloud / infrastructure plays: Companies like Coreweave, Nebius, Cipher, NBIS — the former crypto miners who pivoted to GPU cloud infrastructure — are direct beneficiaries of hyperscaler demand overflow. When AWS tells a startup it can't provision capacity for six months, that startup finds an alternative. The new cloud companies are those alternatives.
Energy: This is the bottleneck nobody is talking about loudly enough. Data centers require enormous amounts of electricity. Even if all the chips arrive and all the real estate is secured, energy is becoming the binding constraint on how fast this infrastructure can be built. Some U.S. states are already pushing back on new data center permits because their power grids can't absorb the load. That makes energy infrastructure — nuclear, gas turbines, grid capacity — increasingly investable as a derivative play on AI capex.

The Valuation Risk: Has the Market Already Priced This?

Here's the honest tension that any thoughtful investor needs to hold: all of this growth is real — and the market has been pricing it in for two to three years. Micron's 12x run. NBIS up many multiples. The memory sector rallied long before the hyperscalers started announcing $200+ billion capex numbers.

When markets pre-price a cycle, they almost always overshoot in one of two directions. Either they over-price (the spend comes in but returns disappoint, and stocks correct 50–60%), or they under-price (the spend exceeds even bullish models, and there's a secondary squeeze higher). What almost never happens: precise pricing. The market doesn't do that.

The risk scenario: one quarter where a hyperscaler hints it might need "slightly fewer" chips than expected — a comment about efficiency improvements in Nvidia's Blackwell architecture, say — could cascade through valuations for the entire AI infrastructure stack. It happened in a small way earlier this year when news about alternative AI models triggered a chip sector selloff. It could happen again, and bigger.

The bull scenario: demand continues to outpace supply into 2027 (as Jassy explicitly projected), the enterprise AI production ramp accelerates, and the companies that built chip and cloud capacity early compound their advantage. In that world, the market still hasn't fully priced the duration of this cycle.

Stock Analysis: What the Charts Say After Earnings

Let's translate the earnings narrative into what actually happened to prices — and what to watch next.

Amazon ($AMZN): Surged +15% on earnings day and is now within roughly $7 of all-time highs. The stock had been essentially flat since January 2025 — building what technical analysts call "the base." The wider the base, the higher the space — and Amazon's 18-month consolidation was wide. A sustained move above the ATH opens a path toward $340, approximately 70 additional points from current levels. Watch for a potential pullback to fill the earnings gap before any continuation.

Alphabet ($GOOGL): Rose +6.7% on Friday and broke above the $349 level on high volume. The next meaningful resistance sits at approximately $362, where a series of declining highs converge. Breaking that level with conviction would confirm that Google is transitioning from "waiting mode" — where investors were uncertain about AI monetization — to genuine uptrend. Volume on Friday's move was strong, which is an encouraging sign.

Microsoft ($MSFT): Broke through declining highs and is now approaching the $466–$470 resistance zone. There's a gap below from the earnings day that may eventually fill, but momentum is clearly positive. Microsoft is the cleanest technical setup of the four right now — clear levels, improving trend, strong CEO narrative to support the move.

Meta ($META): The weakest performer of the group technically. The stock is trading around $556 with two overhead gaps acting as resistance. Key support is at approximately $520 — a breach of that level would be technically significant. The bull case: if you believe Zuckerberg has a plan and $520 holds as support, the next resistance zone above is around $660, representing substantial upside. The bear case: until Meta communicates its AI capex strategy more clearly, the market will keep discounting the story.

The Bottom Line

Four of the world's best-managed companies just told investors — with their own money — that the AI infrastructure buildout is not slowing down. They're going cash flow negative to build faster. They're telling customers they can't fulfill orders fast enough. And they're seeing demand that, by their own analysis, won't be fully met until 2027 at the earliest.

That's not a bubble narrative. That's a capacity constraint narrative — which historically has a different resolution. Bubbles pop when the demand doesn't materialize. Capacity constraints resolve when supply catches up — and in the interim, prices and growth rates stay elevated.

The risk isn't that AI demand is fake. The risk is that the companies downstream from the hyperscalers — the chip suppliers, the new cloud players, the infrastructure builders — are priced as if the demand will accelerate indefinitely. Some will be right. Some will get the timing wrong. And some will be Cisco circa 2000: real companies solving real problems, just not at the valuations the market assigned them at the peak.

Position accordingly. The hyperscalers themselves — Amazon, Alphabet, Microsoft, Meta — are generating the demand and monetizing it. That's a very different risk profile than the companies just selling picks and shovels into the gold rush. 🧘♂️📊