Google Cloud grew 48% in Q4 2025. Then 63% in Q1 2026. Then 82% in Q2 2026. Three sequential quarters of acceleration is not what a low-base bounce looks like. It is what a business finding sustained enterprise demand looks like. And the absolute dollar growth is meaningful too — Google Cloud added roughly $11 billion in YoY revenue in Q2 alone, up from about $9 billion in Q1.

The step function matters because it changes the argument about Google Cloud’s competitive position. For years the skeptics have said Google Cloud was structurally third in the hyperscaler race behind AWS and Azure, and would remain a distant third. Q2 challenges that framing directly. If Google Cloud can hold 80%-plus growth for another quarter, it stops being a #3 story and becomes a competitive-repricing story for the entire cloud market.

The pieces of the acceleration are clear from the earnings call. Enterprise AI workloads on Vertex AI and Gemini models continue to expand. Google’s TPU capacity — v6 Trillium and Ironwood — is a differentiated compute stack that OpenAI competitors and model labs are increasingly interested in. Large customers are running more apps on Google Cloud infrastructure, and Google’s search-adjacent AI expertise is becoming a real enterprise sales lever. This is the tape that Sundar Pichai has been arguing for on every call for two years, finally showing up as a number.

The margin story is arguably more important than the revenue story. Google Cloud operating income was $8.8 billion in Q2, up from $2.8 billion a year earlier — a 214% increase. Operating margin expanded to 35.6%, the first time in the segment’s history it has cleared the 35% line. AWS operates in the mid-to-high 30s. Azure blends into Microsoft’s overall Cloud segment but Intelligent Cloud typically prints in the 40s. Google Cloud is now within striking range of both.

The margin expansion undercuts one of the most common bear arguments about Google Cloud — that the business would scale revenue at low or negative operating margin because it was subsidizing customer acquisition to catch up. Q2 shows the opposite. Google Cloud is now scaling revenue faster than either competitor and expanding margins at the same time. That is unusually rare in cloud infrastructure and it implies either operational leverage from prior investments finally kicking in, or pricing power that Google Cloud did not previously have.

Management did flag modest margin pressure in Q3 from using third-party capacity as a bridge to meet demand. That is a fascinating disclosure — Google Cloud is capacity-constrained, and Alphabet’s own $44.9 billion capex quarter still isn’t enough to keep up with the enterprise pipeline. The Q3 margin will step down slightly, but the direction is unmistakable: this is a business that finally has the unit economics to justify the capex.

Google Cloud’s backlog — remaining performance obligations, in accounting terms — grew by more than $50 billion sequentially to $514 billion. That is the single most under-reported number from the print. At Q2’s roughly $25 billion revenue pace, the backlog represents about 20 quarters of forward revenue coverage. A hyperscaler with five years of visible commitments is a fundamentally different risk profile than one selling quarter-to-quarter.

The backlog build reflects multi-year enterprise AI infrastructure commitments. Anthropic, OpenAI, and several other frontier labs have publicly announced multi-billion-dollar Google Cloud commitments over the past 12 months. Meta reportedly is in talks with Anthropic to lease up to $10 billion in computing capacity over two years, potentially bringing new hyperscaler competition. Fortune 500 enterprises are locking in AI compute at scale to avoid capacity constraints in 2027 and 2028.

What’s changed is who is signing. Two years ago Google Cloud’s backlog was concentrated in a few very large deals. Q2 disclosures suggest it is broadening, with mid-market enterprise workloads and multi-year Gemini commitments showing up in the numbers. Sundar Pichai told the call that baseline capacity is allocated to Alphabet’s internal AI needs first, then to customers with committed backlog, then to spot demand — an ordering that implies the backlog is real commercial commitment and not renewals-at-a-discount.

Alphabet raised full-year 2026 capex guidance to $195-205 billion. Q2 alone burned $44.9 billion, up 100% YoY. Free cash flow turned negative for the quarter at -$5.9 billion — the first FCF-negative quarter for Alphabet since 2010. This is what set off the 3.65% after-hours drop, not the growth numbers. Investors know the growth is real. They are questioning whether the cost structure to support it is sustainable.

To fund the buildout, Alphabet raised $49.6 billion by issuing stock in June and brought in $20.3 billion from senior unsecured notes in Q2. This is a company that has historically been a net share-buyer and a debt-free balance sheet story pivoting to net share-issuance and material debt to finance AI infrastructure. That is a defensible strategy at 82% Cloud growth and a $514B backlog, but it is a different Alphabet than the one Wall Street was pricing 18 months ago.

Then came the punchline. Alphabet warned 2027 capex will “increase significantly” from the $205B guidance level. The market interpreted that as “not yet at peak.” Analysts are now modeling 2027 capex in the $240-270B range. For context, that is more than the entire annual GDP of countries like Portugal or New Zealand, spent by a single company in a single year. The AI infrastructure buildout is not slowing down — it is compounding.

Buried in the Q2 disclosures was a number that developer-tools investors will care about disproportionately: Google’s Antigravity AI coding tool now has 2.4 million weekly active users. Antigravity was launched with relatively little marketing fanfare and has been growing largely on word of mouth inside enterprise engineering teams. 2.4M WAU puts it firmly in the same conversation as Cursor and Windsurf as a serious AI coding assistant, and it is a native part of the Google Cloud ecosystem in a way that neither Cursor nor Anthropic-adjacent tools can claim.

The strategic implication is that Google Cloud is not just selling raw compute — it is building out the developer surface on top of it. Every Antigravity user is a potential Vertex AI customer, a potential Gemini API customer, and a potential enterprise decision-maker who pulls their team toward Google Cloud when the workload migration decision comes up. This is the same playbook Microsoft ran with GitHub Copilot and Azure, and it is arguably why Azure enterprise attach has held up as long as it has.

2.4M WAU is not the largest number in AI coding — Copilot enterprise likely runs bigger, Cursor’s user base is comparable — but the growth curve and the enterprise attach potential are the interesting parts. If Antigravity keeps compounding at this pace through 2027, Google Cloud has a second growth engine adjacent to raw infrastructure that changes the total addressable business.

Microsoft reports July 30. Meta reports July 29. Amazon reports July 30. Google Cloud’s 82% print resets expectations for all three. Investors will now hold Azure to a higher bar. Anything less than 40% Azure growth will be read as Google Cloud stealing share. Anything above 45% will be seen as strong enough to support Microsoft’s own capex commitments. The read-through window is narrower than it was 48 hours ago.

Meta is the more complicated situation. Meta is reportedly in talks to lease up to $10 billion in AI compute to Anthropic over the next two years — a move that would put Meta into the cloud infrastructure business it has historically avoided. If Meta confirms that pivot on the July 29 call, the hyperscaler map officially expands to four. Alphabet, Microsoft, Amazon, and Meta all fighting for enterprise AI capacity is a different competitive setup than the three-horse race investors have modeled for a decade.

Amazon and AWS are the most exposed to Google Cloud’s Q2 print. AWS Q1 growth was in the high 20s. If Q2 does not accelerate meaningfully, the market will read AWS as losing share to Google Cloud in AI-driven workloads specifically. Amazon’s stock has already lagged into earnings for exactly this reason. A weak AWS growth number on July 30 could extend the Software Selloff into August. A strong number could break the pattern and rebuild sentiment across cloud infrastructure.