Alphabet's Q2 results are out. Revenue came in at $119.8 billion, up 24% year over year, beating the $117.1 billion consensus.The numbers behind the headline tell a story about AI capital intensity that every investor in every AI infrastructure stock needed to hear tonight.
Google Cloud grew 82% year over year to $24.8 billion, against a consensus of approximately $22.4 billion and last quarter's already-impressive 63% growth rate. That is not deceleration. That is acceleration. Cloud operating income reached $8.8 billion and operating margin expanded to 35.6%, up from 20.7% a year ago and up from 32.9% last quarter. The fastest-growing major cloud segment in the market is also expanding its margins as it scales. That is the opposite of the overbuild scenario the bears were pricing in.
The Capex Number That Defines the Rest of Earnings Season
The chart says everything. Q2 capex came in at $44.9 billion, up 100% year over year and up 26% sequentially from Q1's $35.7 billion. The trajectory is clear from the quarterly progression: $22.4 billion, $24.0 billion, $27.9 billion, $35.7 billion, $44.9 billion. Each quarter larger than the last. The first half of 2026 alone consumed $80 billion in capital expenditure.
That $44.9 billion spread across 90 days is approximately $500 million per day, every day, including weekends. For context, that daily spend approaches what many S&P 500 companies invest in an entire year. Free cash flow turned negative at minus $5.9 billion for the quarter, the first negative FCF quarter in the company's history. Alphabet raised $49.6 billion of equity during Q2 to partly fund this buildout and authorized a separate $40 billion at-the-market program for employee equity tax obligations.
The question every semiconductor and AI infrastructure investor had coming into this report was simple: is the AI infrastructure buildout slowing, holding, or accelerating? Alphabet's answer is unambiguous. It is accelerating. On the earnings call, management raised its full-year 2026 capex guidance to $195 to $205 billion, up from the prior range of $180 to $190 billion. That means $80 billion in the first half leaves $115 to $125 billion to deploy in the second half. Critically, management cited "accelerated capacity deliveries to meet stronger demand" as the reason for the raise, and disclosed that due to supply constraints it will expand the use of third-party vendors in Q3 as a temporary bridge, which is expected to create modest near-term margin pressure.
Cloud at 82%: The Return on Investment Beginning to Show
The critical debate about AI infrastructure spending has been whether the capex is generating proportional revenue. Google Cloud's Q2 result begins to answer that question affirmatively. The segment went from $13.6 billion in Q2 2025 to $24.8 billion in Q2 2026. That $11.2 billion in incremental quarterly revenue, annualized, is $44.8 billion of new run-rate revenue from a single segment in a single year. The capex required to generate that revenue is enormous. So is the revenue itself.
Cloud operating margin at 35.6% matters as much as the revenue number. Depreciation from newly placed data centers hits immediately while utilization ramps over 12 to 18 months. The fact that margins are expanding, not compressing, while capex is doubling suggests that the assets placed in service in 2025 and early 2026 are already reaching utilization levels that justify the spending. The buildout is not ahead of demand. It is, by the margin data, just barely keeping pace with it.
Gemini reached 950 million monthly active users this quarter, which if accurate would make it the most widely used AI assistant globally. Gemini Enterprise is now deployed across nearly 90% of the Fortune 100. API tokens processed hit 22 billion per minute. These are not research metrics. They are utilization metrics that translate directly into cloud revenue, and they explain why the $462 billion cloud backlog is converting to recognized revenue at an accelerating pace.
Gemini 4, Bigger Models, and Why the Capex Raise Was Inevitable
On the earnings call, CEO Sundar Pichai disclosed that Gemini 4 is being trained as a much larger base model than its predecessors, because Google believes the next generation of frontier AI will require significantly more scale to remain competitive. That single statement explains the $15 billion capex guidance raise more clearly than any financial metric. Larger models require more compute to train, more memory to serve at inference, and more data center capacity to host at scale. The guidance raise from $180-$190 billion to $195-$205 billion is not a discretionary investment decision. It is the downstream consequence of a competitive dynamic in which the leading AI labs are all simultaneously concluding that the next capability frontier requires more capital, not less.
Pichai also said Alphabet is "so constrained on AI infrastructure that we will be using third-party vendors to help bridge the gap until we can build our own out more." This is the scarcity signal that every memory and semiconductor investor has been waiting to hear from a hyperscaler earnings call. Google has a $462 billion cloud backlog it physically cannot build fast enough to serve, and it is turning to third-party capacity as a bridge. That third-party capacity includes the GPU clusters and memory-intensive compute that companies like CoreWeave, Oracle Cloud Infrastructure, and others provide, all of which flow through Nvidia, Micron, and SK Hynix at the hardware layer.
What This Means for MU, SKHY, NVDA, and the AI Infrastructure Complex
Alphabet spending $80 billion in the first half of 2026 on AI infrastructure is the most direct available confirmation of the demand thesis that every AI infrastructure stock in the market is priced against. Micron's $100 billion in contracted take-or-pay revenue from hyperscalers is not a forecast. It is a contractual obligation from companies whose actual spending behavior tonight confirmed that AI infrastructure demand is running at or above contracted levels. SK Hynix, which reports its own Q2 results on July 29, now has an unambiguous tailwind going into its call: the company with the largest single cloud backlog in the world just demonstrated that it is deploying at an accelerating pace and the returns are materializing.
Nvidia captures GPU revenue from Alphabet's infrastructure build. TSMC manufactures the TPUs and GPUs that fill those data centers. ANET provides the networking. VRT provides the power and cooling. Each of these names benefited from Alphabet's Q1 capex of $35.7 billion. Q2's $44.9 billion is a 26% sequential increase in that demand signal. The read-through is not that AI infrastructure demand is slowing. It is the opposite.
The one complication for the broader market is the negative free cash flow and the equity dilution that accompanied the spending. Alphabet raised $49.6 billion in equity during the quarter. That is a signal that even a company generating Alphabet's level of operating income cannot fund this buildout entirely from internal cash flows. The implication is that AI infrastructure at this scale requires capital markets participation, which means higher interest rates or tighter credit conditions are not just a valuation headwind. They are a direct funding constraint on the pace of buildout.
Berkshire Just Bought Alphabet: The Other Signal of the Day
Separate from the earnings, the Motley Fool reported today that Warren Buffett personally selected Alphabet as Berkshire Hathaway's newest mega-stock investment, with Greg Abel loading up on $10 billion in Alphabet stock. This follows the pattern of Berkshire's $10 billion Alphabet private placement announced earlier in July. Two major institutional validators, BofA adding Micron to the US 1 List and Berkshire personally buying Alphabet, arriving in the same week, and both investments directionally confirming the AI infrastructure thesis.
The One Number to Watch in Q3
The negative free cash flow of minus $5.9 billion is the market's primary concern heading into Q3. Cloud at 82% growth with expanding margins is the answer to the demand question. The FCF turning negative is the question about capital intensity sustainability. Alphabet management did not provide formal Q3 guidance. What the market will watch is whether Q3 FCF returns above zero, which requires either capex to stabilize or cloud revenue to continue its acceleration. Given that the second half now implies $115 to $125 billion more in capex under the raised $195-$205 billion guidance against a cloud segment now running at nearly $100 billion annualized, the math on FCF recovery depends on whether utilization continues ramping on the assets already placed in service.
The AI infrastructure buildout is not slowing. It is accelerating. The returns are beginning to materialize in cloud margins. And the world's most closely watched smart money, from Berkshire to BofA, is putting capital behind the thesis in the same week the data confirmed it.
Explore more exclusive insights at nextfin.ai.
