Then the cash flow statement. Operating cash flow came in at $39.1 billion, up 41% year over year. Capital expenditures came in at $44.9 billion, roughly double the prior-year quarter. The difference — negative $5.9 billion — is the first negative free cash flow print in twenty-two years as a public company.
The market did not treat it as a rounding error. Alphabet shares fell roughly 7% after the print. The company repurchased zero shares for the second consecutive quarter, against $13.2 billion of buybacks in the year-ago quarter.
One Note on the Headline Profit
Net income available to common shareholders rose 298% to $112.1 billion, and diluted earnings per share rose 294% to $9.11. Those numbers deserve an asterisk that most coverage skipped.
Other income for the quarter reflected a net gain of $98.0 billion, driven almost entirely by unrealized gains on Alphabet's equity securities. The company disclosed that the $99.0 billion equity-securities gain by itself added $77.1 billion to net income and $6.26 to diluted EPS. Strip it out and EPS lands near $2.85 rather than $9.11.
The record that matters operationally is the $40.8 billion of income from operations, up 30%. That one is real, and it is the number to hold against the capital spending. The headline EPS figure is a mark-to-market artifact that could reverse in any future quarter, as Alphabet itself warns in the filing.
The Quarter in One Table
| Line | Q2 2026 | Direction |
|---|
| Revenue | $119.8B | +24% YoY |
| Income from operations | $40.8B | +30% YoY, margin 34% |
| Other income, net | $98.0B | Mostly unrealized equity gains |
| Google Cloud | $24.8B | +82% YoY; segment income tripled |
| Search & other | $63.3B | Core advertising |
| Operating cash flow | $39.1B | +41% YoY, down from $45.8B in Q1 |
| Capital expenditure | $44.9B | ~2x YoY |
| Free cash flow | –$5.9B | First negative quarter since 2004 |
| Trailing-12-month FCF | +$53.3B | Down ~20% YoY |
| FY2026 capex guidance | $195–205B | Raised from $180–190B |
| Buybacks | $0 | Second straight quarter |
| Cash & marketable securities | $242.5B | Up from $126.8B at year-end |
The Capex Line Is the Story
Five quarters of capital expenditure tell it faster than any analysis:
- Q2 2025 — $22.4 billion
- Q3 2025 — $24.0 billion
- Q4 2025 — $27.9 billion
- Q1 2026 — $35.7 billion
- Q2 2026 — $44.9 billion
That is a double, year over year. Trailing-twelve-month capex now stands at $132.4 billion, up 98%. Capital expenditure now absorbs 37.5 cents of every dollar Google collects in revenue — the highest share in the company's public life. For scale, Alphabet's full-year 2022 capex was roughly $31 billion. The company now spends more than that in a single quarter.
Management then raised full-year 2026 guidance to a range of $195 billion to $205 billion, up from $180 billion to $190 billion. That is the second raise in three months and the third distinct FY2026 capex number in three quarters, against analyst models sitting near $187 billion. Roughly 60% of the quarter's infrastructure spend went to servers, with the remaining 40% to data centers and networking equipment.
Chief Financial Officer Anat Ashkenazi attributed the increase to "an acceleration in the delivery of capacity to meet growing demand" — not new demand, faster fulfillment of demand already booked. She told analysts free cash flow would remain under pressure and confirmed that 2027 capital spending would rise significantly. She declined, for the second consecutive quarter, to attach a number to 2027.
How Alphabet Is Paying For It
This is the part the headlines skipped, and it is where the numbers have been most widely misreported.
The figure in circulation is $85 billion. It comes from Alphabet's own June 2 pricing release, which headlined a total equity raise of $84.75 billion. That number is a program size, not cash collected. It combines the gross June offerings, a $10 billion private placement to an affiliate of Berkshire Hathaway, and the full $40 billion capacity of a newly established at-the-market equity program.
What Alphabet actually raised, per the cash flow statement:
- $30.5 billion — net proceeds from the June issuance of Class A and Class C stock, including the Berkshire placement
- $19.1 billion — net proceeds from 6.25% mandatory convertible preferred stock, issued June 5
- $20.3 billion — net proceeds from senior unsecured notes issued during the quarter
Equity and preferred together came to $49.6 billion. Add the notes and the real figure is roughly $70 billion of new capital in a single quarter. The $40 billion at-the-market program had sold zero shares as of June 30, and Alphabet says its proceeds are earmarked primarily for employee equity tax obligations rather than capital expenditure.
The balance-sheet consequences are larger than the headline number implies either way. Long-term debt more than doubled year over year, from $46.5 billion to $98.2 billion. Interest expense nearly quintupled, from $261 million in Q2 2025 to $1.28 billion. Net cash provided by financing activities swung to a positive $61.2 billion, against negative $5.8 billion a year earlier. The company that spent two decades as the cleanest balance sheet in technology now carries a preferred instrument senior to its own common stock, converting to common around May 2029.
It is also renting. Google has said it will lean on third-party cloud providers including CoreWeave and Nebius through Q3 as a bridge while its own data centers come online. Reporting on the quarter put Alphabet's Nvidia-chip rental arrangement with SpaceX at roughly $920 million a month.
Companies holding $242.5 billion in cash and marketable securities do not issue mandatory convertible preferred stock and suspend buybacks for a rounding error. They do it when the buildout cannot wait for operating cash to arrive.
What the Coverage Missed
Within seventy-two hours the number had been covered by Semafor, Benzinga, Search Engine Journal, Seeking Alpha, CNBC, and effectively every finance newsletter in circulation. The consensus frame: an infrastructure spending story. Capex is exploding. Margins are at risk. Return timing is unclear.
That frame is accurate and incomplete.
Alphabet did not break twenty-two years of positive cash flow to buy warehouses. It broke them to buy the answer layer.
The compute Alphabet is installing serves Gemini, Google AI Overviews, and AI Mode. Those products are the mechanism by which Google intends to remain the place a buyer asks the question, as the question migrates from a search box returning ten links to an engine returning one answer. Every dollar in that $205 billion guidance is a dollar spent defending the position of default answer.
Read the disclosure that way and the cash flow statement stops being a capex story. It becomes the most expensive competitive signal yet filed about where discovery is going.
What $205 Billion Actually Buys
Alphabet's own investor materials tie the spend to product, not to abstraction. Sundar Pichai's quarter-opening remarks put hard usage figures behind it: the Gemini app now has 950 million monthly active users, Gemini models process 22 billion API tokens per minute, and nearly 90% of the Fortune 100 use Gemini Enterprise.
Since Gemini 3 shipped in November 2025, the company has pushed out generative media models, features across Chrome and the Gemini app, the Antigravity agent-development platform, and the Gemini 3.5 series. AI Mode has expanded toward 200 countries and 98 languages. Search agents that run in the background and agentic booking for local services are rolling out in the United States.
Each of those is an inference load. Training is a capital event; inference is a recurring one, and it scales with every additional user who asks rather than searches. Google is not building to win a model benchmark. It is building to serve an answer to two billion people a month without losing money on each one.
For brands, the practical consequence is that Gemini and AI Overviews are now the same distribution problem wearing two interfaces. The Google AI Overviews Citation Source Index 2026 found the top 1% of domains capturing 47% of all AI Overview citations, with schema markup lifting citation likelihood 2.3x. The Gemini Citation Source Index 2026 found 43% of Gemini citations pointing to Google-owned properties, YouTube alone at 29.5%. Concentration is the design, not a bug in it.
The Demand Signal on the Other Side
Capex is the supply side of this trade. The demand side is measurable, and it moved first.
- 68.01% of Google searches in the first four months of 2026 ended without a click, per Similarweb clickstream data analyzed by SparkToro. US zero-click stood at 60.45% in 2024.
- A randomized field experiment published on SSRN in 2026 found that showing an AI Overview caused outbound organic clicks to fall 39.8% and zero-click searches to rise 34.5%, while paid clicks stayed roughly flat. That is causal evidence, not correlation.
- An Ahrefs study across 300,000 keywords put the click-through reduction for top-ranking pages at 58%, nearly double the 34.5% the same researchers measured in April 2025.
- Chartbeat data cited by Press Gazette put the global decline in search referrals to publishers at roughly 33% since AI Overviews rolled out broadly.
- Google Network revenue — the third-party publisher line running through AdSense, AdMob and Ad Manager — came in at $7.30 billion, down slightly against $7.35 billion a year earlier. It is the only Alphabet revenue line not growing, and it fell 4% year over year in Q1 2026.
AI Overviews now reach roughly two billion monthly users across more than 200 countries. ChatGPT reported 900 million weekly active users in February 2026. The behavior did not fragment across a dozen new destinations. It consolidated into a smaller number of answer surfaces, and Google is spending $205 billion to make sure one of them is still its own. Everything-PR has tracked that consolidation in the Five Answer Engines reference and in the Paywall Penalty Index, which found hard-paywalled publishers capturing 0% of AI-retrieval citations across every category tested.
The Number Behind the Number
Alphabet is not the whole of it. Meta raised 2026 capital expenditure guidance to a range of $125 billion to $145 billion and watched its shares fall more than 9% on the news. Microsoft is tracking toward roughly $190 billion for the calendar year. Amazon is guiding near $200 billion.
Combined 2026 capital expenditure across the four largest hyperscalers is expected to reach approximately $725 billion, up about 77% from roughly $410 billion in 2025. Goldman Sachs now models $5.3 trillion of hyperscaler capex from fiscal 2025 through fiscal 2030, revised up from $4.5 trillion before first-quarter earnings. Analysts already project the annual figure crossing $1 trillion in 2027.
Three quarters of a trillion dollars, in a single year, building the infrastructure that answers questions.
For an industry that has spent two decades optimizing for a search engine's ranking algorithm, the relevant figure is not the deficit. It is the commitment. Companies do not suspend buybacks, issue convertible preferred stock, and post the first negative cash flow quarter in twenty-two years to defend a business they expect to decline.
The Counter-Position
The bear read deserves a fair hearing, because it is not stupid.
One quarter is one quarter. Capex is lumpy by nature, and a single negative print says less than a trend. Trailing-twelve-month free cash flow remains solidly positive at $53.3 billion. Alphabet holds $242.5 billion in cash and marketable securities. Operating cash flow is still growing 41% year over year. Nothing in the quarter suggests distress.
The sequential decline in operating cash flow — $45.8 billion in Q1 down to $39.1 billion in Q2 — is worth isolating, since it happened while revenue accelerated. The filing attributes it largely to working capital rather than margin, with inventory alone consuming $6.7 billion of cash in the quarter.
The sharper version of the bear case is not about liquidity at all. It is about depreciation. Hyperscalers are buying property and equipment at close to three times the rate their income statements currently recognize as depreciation expense, because servers depreciate over five to six years and buildings over twenty-five to forty. Alphabet's own quarterly depreciation was $7.1 billion against $44.9 billion of capex. The expense wave arrives on a schedule regardless of what AI revenue does. That is a real risk to reported margins in 2027 and 2028, and Ashkenazi named it directly on the call.
Both things hold. The spending may prove badly timed for shareholders and still be an accurate signal about where buyer attention went. Those are separate questions, and communications teams only need the answer to the second one.
What It Means for Communications
Four implications, stated plainly.
One — the retrieval layer is capitalized, not experimental. For two years it has been possible to treat AI-mediated discovery as an emerging channel worth monitoring. A $205 billion annual commitment from the incumbent ends that argument. The answer layer is where the spend is going because it is where the attention went.
Two — the surface brands compete on is narrowing. Ten blue links gave ten brands a position. A generated answer names three sources, sometimes one. As the compute behind those answers scales, the answers improve and the citation set tightens. Citation Share becomes a scarcer asset every quarter this spending continues. The first GEO benchmark found the top Google result for a brand query appearing inside ChatGPT only 38% of the time. Ranking and citation are not the same asset.
Three — the measurement gap is a governance problem. Most brands cannot answer how often they appear when a buyer asks an AI engine a category-defining question. That was a tolerable gap in 2024. Against three quarters of a trillion dollars of infrastructure built specifically to produce those answers, it is a board-level blind spot. There is a documented methodology for measuring Citation Share, and no remaining excuse for not running it.
Four — the answer is becoming the transaction. Google's agentic booking, OpenAI's Instant Checkout, and the emerging commerce protocols collapse the distance between being named and being bought. Agentic commerce turns an answer-engine citation into a revenue line rather than a traffic line. Brands still measuring sessions are measuring the wrong end of the funnel.
What to Watch Next
- The Q3 2026 print. Whether negative free cash flow repeats, or whether Q2 was a delivery-timing artifact.
- The 2027 capex number. Deferred twice. When it lands, it sets the ceiling on the buildout.
- The ATM program. $40 billion authorized, zero drawn. Any drawdown signals the funding gap widened.
- Google Network revenue. The cleanest quarterly read on what AI Overviews are doing to the open-web publisher economy.
- Depreciation as a share of revenue. $7.1 billion this quarter against $44.9 billion of capex. The gap is the future margin pressure.
- Citation concentration. Whether better models cite more sources or fewer. Every index so far points to fewer.
The Reading
Alphabet's second quarter is not a warning about AI spending discipline. It is a disclosure about where discovery has moved, filed in a quarterly report by the company with the most to lose from being wrong about it.
The company that built the search era just spent more cash than it generated to make sure it also owns the answer. Brands still measuring rankings are optimizing for the surface Google itself is spending $205 billion to replace.
The answer engines
The Five Answer Engines in 2026 · Gemini: Google's Flagship AI Assistant — The EPR Pillar · Google AI Overviews and the Death of the 10 Blue Links
Citation research
The Google AI Overviews Citation Source Index 2026 · The Gemini Citation Source Index 2026 · Who AI Cites: 366,087 Citations Mapped · The Paywall Penalty Index 2026
Playbooks
How to Rank on Google AI Overviews and AI Mode in 2026 · How to Measure Citation Share · Inside Google AI Overviews' Brand Bias · Inside Gemini's Brand Bias
Context
Agentic Commerce: When ChatGPT and Gemini Become the Checkout · Technology Public Relations: The EPR Pillar · Google's PR Disaster Playbook · The 5W Reset: Why We Became the AI Communications Firm
Sources: Alphabet Q2 2026 earnings release, July 22, 2026 (primary — all revenue, operating income, cash flow, capex, financing and balance-sheet figures) · Alphabet Form 10-Q for the quarter ended June 30, 2026 (offering mechanics and Berkshire private placement) · Alphabet June 2, 2026 offering pricing release (the $84.75 billion figure) · Series A mandatory convertible preferred prospectus supplement · Q2 2026 earnings call transcript · Goldman Sachs hyperscaler capex estimates · Silicon Analysts on the depreciation gap · SparkToro / Similarweb zero-click data · AI Overviews CTR study roundup
