
This quarter's growth looks strong. Total revenue reached $127.6B, up 25% year over year. Operating profit was $40.6B, compared with $31.2B a year earlier. Google Cloud, the business that rents computing capacity and AI services to outside customers, generated $30.7B in revenue, up 103% year over year. The stock moved in the opposite direction. It closed at $324.89 on October 28 and at $283.82 the next day.
The market sold on three numbers that missed expectations.
The first was operating profit. Cantor Fitzgerald said it was about 7% below expectations. Earnings per share of $2.12 missed the market's roughly $2.94 forecast by $0.82. Of that shortfall, $0.55 came from declines in the book value of shares Google holds in other companies. Those losses totaled $8.7B this quarter and had no connection to its core operations. Adding them back gives about $2.67, still below expectations. The remaining gap came from the core business.
The second was search. Search revenue reached $63.3B, up 12% year over year, against Wall Street consensus of 15% growth.
The third was free cash flow: cash generated by operations minus capital spending, the money spent building data centers and buying servers. Third-quarter operating cash flow was $42.3B. Capital spending was $51.4B, leaving negative free cash flow of $9.1B. The second quarter had already recorded negative $5.9B. Analysts had expected negative $2.2B this quarter, according to FactSet data cited by Axios. Of the $9.1B shortfall, $5.2B was a cash payment this quarter for a 2018 EU fine. Section 6 examines that detail.
On the same evening, Microsoft reported positive quarterly free cash flow of $18.6B. Reuters compared the two that day. The market rewarded Microsoft for funding itself and rejected Google's cash burn.
When I wrote my previous Google report on July 23, the stock was $319.62. I treated cloud margins as the deciding factor and search as the safe part. Three months later, cloud margins have passed this quarter's test, though they need another check in the fourth quarter. Search has instead become the reason the stock is under pressure. I retain my bullish conclusion. What needs proving has shifted from cloud to search.
First, what exactly is this quarter's $51.4B in capital spending chasing?
Backlog is the value of contracts customers have signed that Google has not yet recognized as revenue. Google Cloud's backlog was $155B at the end of the third quarter of 2025. It reached $240B in the fourth quarter, $462B in the first quarter of 2026, $514B in the second quarter, and $561B in the third.
The largest jump came in the first quarter, and two factors need separating. Several very large TPU hardware sales contracts entered the total. The company also widened its reporting scope to include contracts lasting less than a year. Those added about $7.3B across the whole company, not just cloud. Chief Financial Officer Anat Ashkenazi said on the earnings call that most of the backlog still consisted of ordinary cloud platform contracts.
She also gave a delivery schedule. Slightly more than 55% will become revenue over the next 24 months. That percentage applies to company-wide backlog, nearly all of which is cloud. Using $561B as a rough base gives about $309B over two years, or roughly $154B a year. Multiplying Google Cloud's quarterly revenue by four gives about $123B a year.
Think of a restaurant whose reservation book stretches two years ahead. More than half of those guests will arrive within two years, but the kitchen lacks enough stoves.
With too few stoves, Google is borrowing the kitchen next door. Ashkenazi said Google began using third-party computing capacity at the end of the third quarter and would use more in the fourth. Google Cloud CEO Thomas Kurian said on the call that Google could clearly serve much more demand if it had more capacity. A company willing to rent other people's machines and accept margin pressure to deliver capacity to customers has demand running ahead of supply.
Not all of the $51.4B goes toward cloud contracts. On the fourth-quarter 2025 call, Ashkenazi said slightly more than half of 2026 machine-learning computing capacity was expected to support cloud. That is the capacity used to run AI models. The rest would support Google's other businesses.
Third-quarter capital spending was $51.4B, against $24.0B in the same quarter a year earlier. Full-year 2026 guidance remains $195–205B. Ashkenazi said 2027 capital spending would increase substantially, with a specific figure to come on the fourth-quarter call.
My previous report argued that Google could not avoid this spending for three reasons: demand constrained by supply, competition among AI assistants, and search. Search was the most consequential. This quarter's orders and rented capacity put numbers behind the first reason. On the companies' reported figures, Google Cloud grew 103%, Microsoft's Azure grew 48% in its latest quarter, and Amazon's AWS grew 37%. The 103% includes TPU hardware sales, discussed next. I think customers want computing capacity their provider can actually deliver. If Google cannot supply it, those orders will go to another cloud provider.
Operating margin is the operating profit left from every $100 of revenue. Google Cloud's margin was 23.7% a year ago, 35.6% in the second quarter, and 36.2% in the third. Revenue doubled while the margin increased. Cloud operating profit grew by $7.5B in a year.
In July, I worried about three pressures on margins: rented computing capacity, dilution from the acquisition of cloud security company Wiz, and rising depreciation. Depreciation spreads machine costs across annual income statements. The third-quarter figures have not yet shown that downward pressure.
This needs another test. The rented capacity came online only at the end of the third quarter, limiting its effect on this quarter's margin. Ashkenazi has already warned that using it will pressure margins in the near term. Quarter over quarter, 36.2% was only 0.6 percentage points above the second quarter. UBS described cloud margins as stalled. TPU system sales already made a “significant contribution” to cloud growth this quarter. These sales involve complete TPU servers installed in customers' data centers. Google does not disclose their margin, so we cannot tell whether hardware helped or hurt the reported 36.2%.
Three of them asked about the financials. None of the three received a number.
Barclays' Ross Sandler asked how much revenue a roughly $35B, one-gigawatt complete TPU system would generate, and at what margin. One gigawatt is 1 billion watts, a measure of data-center scale based on electricity use. Ashkenazi replied that the company does not disclose financial data for individual products. Bank of America's Justin Post asked why this was a good business if Google would not disclose the margin. Kurian replied that Google does not break out individual product margins and would keep working to improve profitability.
Bernstein's Mark Shmulik asked the most direct question: how fast was Google Cloud growing without TPU system sales? Kurian repeated the 103% figure. He said GCP, Google Cloud Platform, accounted for the vast majority of revenue. GCP is the main business renting out computing capacity and software. Growth excluding TPU system sales was “very, very strong.” There was no number.
With management silent, Wall Street is debating two things. The first is whether TPU margins approach those of the core cloud business. After earnings, on October 29, Mizuho inferred from the continued rise in cloud margins that they were approaching that level. The second is what earnings multiple those profits deserve even if the margins converge. That means how much the market will pay for each dollar the business earns. In a pre-earnings note on October 6, Wells Fargo wrote that market thinking was moving toward roughly 15 times, similar to Nvidia. It did not specify the earnings period. After earnings, Morgan Stanley described third-party TPU sales as “lower-multiple earnings.”
First are repurchase commitments. Some TPUs that Google sells could come back. Certain contracts allow customers to return hardware under specified conditions. Google would have to buy it back at a discount, up to $6.6B. Google considers this unlikely. The discounted repurchase price makes returns unattractive for customers, and the quarterly filing says the current value of these obligations is small.
Second are related-party contracts. Backlog includes $16.8B from a multiyear agreement with Crux AI. Crux AI is an unconsolidated joint venture in which Google holds a stake. It is a related party, meaning a counterparty connected to Google through ownership. Google has also committed to provide a backstop of up to $10.0B for its future data centers and energy facilities under certain conditions. Of this quarter's roughly $47B increase in backlog, more than one-third came from this single agreement.
Why does that affect valuation? Think of two businesses earning the same dollar of profit. One collects rent each month. The other sells a piece of equipment once. The market will pay different prices for those profits. Rent returns every year. After an equipment sale, the seller needs another order.
For the bullish case to hold, ordinary cloud contracts must therefore remain most of the backlog: contracts signed but not yet recognized as revenue. TPU hardware should not receive the valuation of a business collecting rent year after year.
My previous report treated search growth below roughly 12% as a sign that AI was starting to erode the business. This quarter came in at exactly 12%, right on that line. In July, I said AI was expanding search rather than eroding it. I based that on the second quarter's 17% revenue growth, without examining clicks separately. CEO Sundar Pichai said year-over-year query growth this quarter was the fastest in five years. Users have not left, so I see slower monetization. But I have another layer of concern compared with July.
Chief Business Officer Philipp Schindler listed the reasons for slower search growth. Currency was the largest. Industry factors also mattered. The EU's removal of duty exemptions for low-value parcels affected retail advertisers in Asia-Pacific. Media and entertainment advertising eased after the World Cup boosted the second quarter. Last came AI Mode: queries are growing quickly, but monetization is lower. Ashkenazi added that the third quarter began comparing against last year's second-half acceleration in search. Faster growth a year ago makes this year's comparison harder. That reduced growth by about 3 percentage points. Both pressures continue into the fourth quarter: slightly stronger currency headwinds and another one or two points from the difficult comparison.
Of these factors, management quantified only the comparison effect.
The quarterly filing breaks search revenue into two components. Paid clicks, those users make and advertisers must pay for, grew 8% year over year in the third quarter, against 13% in the second. The price per click grew 4%, compared with 3% in the second quarter.
Management ranked currency first and AI Mode last. I think that underplays AI Mode. The filing lists currency as a factor affecting the price per click. Yet price growth did not slow from the second quarter. Clicks did. Currency therefore cannot explain the entire slowdown. Nor can AI Mode take all the blame for slower clicks. The company has not provided that breakdown.
Think of a shopping mall where visitor growth is the fastest in five years and stall rents have not decreased. But its new floor has few ad spaces. At some branches, that floor has none installed yet. AI Mode is that new floor.
Across the last five earnings calls, this was the first time management directly said that AI Mode monetizes at a lower rate. On the same call, Schindler said most AI Mode queries were incremental. “Most” does not mean all. I think some queries are replacing traditional search. The company has not said how many.
Management's response is to expand ads. On this call, Google announced the first launch of AI Mode ads outside the United States. In the United States, it began placing ads in conversations with multiple exchanges. Pichai compared this with the move to mobile more than a decade ago. Then, too, Google developed the user experience first and monetized it later.
The comparison makes sense, but a number is missing. Analysts have asked how AI search makes money on five consecutive calls, starting in the third quarter of 2025. Management has never provided AI Mode's monetization rate, ad coverage, or share of queries. Without that number, the market can only wait for revenue to speak.
Year-over-year search revenue growth from the third quarter of 2025 through the third quarter of 2026 was 15%, 17%, 19%, 17%, and 12%. The next three quarters compare against 17%, 19%, and 17%. The first comparison against the lower 12% base comes in the third quarter of 2027. Google will release those results around late October 2027.
The bullish case also requires search monetization in AI interfaces to catch up. Revenue will make that clear in the second half of 2027 at the earliest. Before then, revenue growth will probably look weak. That does not itself mean monetization has failed.
Cloud is spending to expand. Services, meaning search, YouTube, and subscriptions, pays the bills. Third-quarter services revenue was $96.3B, up 11% year over year. Its margin was 40.7%. The reported year-earlier margin was 38.5%, but that included last year's $3.5B EU fine. That was not the 2018 fine paid in cash this quarter. Excluding it, the year-earlier margin was 42.5%.
The business paying the bills is still increasing profit, just more slowly than revenue. Excluding last year's fine, services operating profit grew about 6% year over year. The reported 17% growth reflects the lower base caused by that fine. Costs grew about 14%, faster than revenue growth of roughly 11%.
Query growth is still the fastest in five years, which means people are still searching. The next question is who will start those searches in the future.
Search advertising revenue comes from advertisers paying for human attention. A person searches, sees an ad, and clicks. The advertiser then pays. AI agents challenge that exact step. Ask an agent to book a flight, for example. It compares prices, chooses the flight, and places the order. You never see the results page, so nobody may see the ads. That is a major risk for Google. It must find new ways to charge.
This is only beginning. The financial figures do not yet show its effects, and Schindler said agents were still in their early stages. But they are already appearing. Pichai said the Gemini app, Google's own AI assistant, had added a Tasks tab. Users were already entering agent-style requests there. Later this year, consumer agents will start reaching the Gemini app's 1 billion users.
Some on Wall Street have already stated the issue plainly. In a pre-earnings note on October 6, Wells Fargo wrote that consumer agents replace commercial search rather than supplement it, requiring Google to respond. After earnings, Citizens downgraded the stock from a Buy-type rating to Neutral. One reason was greater agent competition and considerable execution risk in the transition of search monetization. Investing.com relayed Barclays' view that more aggressive agent workflows in search during 2027 could push traditional search growth further behind total query growth.
Bank of America's Justin Post asked directly whether the shift toward agents would be a headwind for search. Schindler replied that agents were clearly incremental for Google. His reasoning was that users dislike shopping chores and will delegate those to AI, but they enjoy choosing. Google will leave room for choice in agent workflows. An agent could offer one insurance policy it found, then another with a higher premium, a lower deductible, and a discount. Almost no decision has only one perfect option. He was therefore confident Google could build a good advertising model, alongside subscriptions and other payment methods. Pichai called this an opportunity for expansion.
Google also has assets to work with. It has its own agent, Gemini, with 1 billion users. It also has the Universal Commerce Protocol, which lets merchants complete purchases directly and already connects hundreds of thousands of brands.
I think this answer holds up where people still want to decide for themselves. It does not answer what happens when an agent handles a task from beginning to end. Nor does it include numbers. AI Mode ads already earn less than traditional search ads. Management has not said what agent ads could earn.
Kurian explained on the call how agents become cloud revenue. Customers building agents buy computing capacity to execute tasks, storage for agent memory, and databases to provide context. TPU customers include Anthropic, both directly and through Broadcom. Apple also uses them. Anthropic makes an AI assistant that competes with Gemini. That point comes from my reading of public information, not from company disclosure. Other companies' agents may therefore also pay Google when they run.
This is another side of Google's second revenue engine, but I would not say it offsets search risk. The payers differ. Advertisers fund search, while companies building agents pay for computing capacity. A loss on one side will not automatically become a gain on the other. The certainty differs too. Signed cloud contracts have dollar amounts and customers. Neither is yet clear for agents' computing revenue. Cloud's 36.2% margin is not far from services' 40.7%. But selling computing capacity requires spending first and collecting the money gradually. The next section examines that cash flow. Other companies also sell capacity. Azure and AWS are both growing. Barclays' Sandler also noted on the call that companies such as Anthropic were taking more infrastructure into their own hands.
As for scale, I have not found anyone putting a number on it. Search generates $63.3B in quarterly revenue. The company has not disclosed how many queries agents bring, how much advertising they displace, or how much computing revenue they generate. None of the analyst reports I read calculated those figures either. We can see the direction, but we cannot calculate its size. I therefore base the bullish case on signed cloud contracts, search that is still growing, and today's price, rather than this computing revenue. Agents are the third issue to watch. They are not a premise of the bullish case.
Third-quarter operating cash flow of $42.3B covered about 82% of the $51.4B in capital spending. The $9.1B gap included two special items. One was the $5.2B paid after appeals ended over the 2018 EU fine. This was a one-time payment. Excluding it leaves a gap of about $3.9B. The other was inventory, which consumed $3.2B in cash this quarter, mainly to prepare for TPU system sales. Kurian said Google builds inventory only after signing a contract. Ashkenazi said cash returns only after systems reach customers and Google invoices them.
Across the first nine months, operating cash flow was $127.15B and capital spending was $131.95B, leaving a gap of about $4.8B. So far, the money Google earns has almost covered construction. The gap will widen. Subtracting the $131.95B already spent from full-year guidance of $195–205B implies fourth-quarter capital spending of about $63–73B, against $51.4B in the third quarter.
Financing in the first nine months far exceeded the shortfall, with almost half remaining on the balance sheet. Debt and new share issuance, after debt repayments, dividends, and taxes on employee shares, provided about $115.8B in net financing. The gap was only $4.8B. The two largest uses were $34.1B for acquisitions and other spending, and roughly $54B retained in cash and bonds.
The company's $265.1B in cash and marketable securities needs separating. It includes $90.4B in publicly listed equities, of which $75.2B is Google's investment in SpaceX shares that Google cannot sell in the near term. Actual cash and bonds total about $175B, against roughly $121B at the start of the year. I think Google is preparing this money for 2027, when capital spending will rise substantially again.
Google raised about $50B by selling shares in June. The third-quarter 10-Q, its quarterly filing with the U.S. Securities and Exchange Commission, breaks this into $20.5B of public common-stock issuance. Another $10.0B went to Berkshire Hathaway subsidiaries. Roughly $19B came from preferred shares that must convert into common shares at maturity. On the July call, Ashkenazi said Google had no immediate plan to return to the equity market apart from its ATM program. ATM means at-the-market, a program that sells new shares gradually at market prices.
Third-quarter ATM sales exceeded the stated use. Google says the program primarily funds taxes related to employee equity compensation. That expense was $5.4B this quarter, while ATM sales reached $12.4B, more than twice as much. Buybacks remained at zero, against $11.5B in the same quarter a year earlier. New share issuance has become part of Google's funding this year.
Third-quarter interest expense was $967M, against $143M in the same quarter a year earlier. That is about 2.4% of this quarter's $40.6B in operating profit. The weighted-average share count, averaging the shares outstanding across the quarter, was 12,256M, against 12,086M a year earlier. That is an increase of about 1.4%. It excludes the common shares that will result when roughly $19B of preferred shares convert at maturity. The regular quarterly dividend of $0.22 per share continued.
New shares reduce the profit attributable to each share. Debt requires interest payments. Both are costs shareholders pay for this construction cycle. For long-term holders, whether those costs are worthwhile depends on what this capacity earns in 2029 and 2030.
How much has today's price already accounted for?
From the price at writing, returning to the valuation seen during the last search scare means a decline of 13–18%. Reaching my fair-value midpoint of $316 means a gain of about 13%. Returning to 23–27 times means a gain of 18–38%. The difference is entirely the multiple the market will pay. The earnings figure is the same throughout.
This section uses one earnings figure: Wall Street consensus for Google's earnings per share over the next 12 months, about $14.35. Dividing $279.79 by 19.5 times implies that figure. Every price band below multiplies that figure by the corresponding price-to-earnings ratio. Two indexes provide context. At the report date, the S&P 500 trades at 19.6 times. The technology-heavy Nasdaq-100 trades at 21.2 times.
| Band | P/E Ratio | Implied Share Price | What the Band Means |
|---|---|---|---|
| Bottom | 16–17 times | $230–244 | A return to the valuation during the last search scare |
| Somewhat cheap | 17–21 times | $244–301 | Below fair value. The price at writing of $279.79 (19.5 times) sits here, almost level with the S&P 500's 19.6 times |
| Fair value | 21–23 times | $301–330 | Roughly the Nasdaq-100's 21.2 times and above. The midpoint of 22 times implies $316 |
| Rerating | 23–27 times | $330–387 | Search concerns ease, and the P/E ratio returns to a higher level |
| Somewhat expensive | 27–30 times | $387–430 | The price assumes more than near-term evidence supports |
| Top | 30–32 times | $430–459 | The highs from late 2025 through spring 2026. The multiple was 32.1 times on February 2, 2026 |
The fair-value band is roughly at the Nasdaq-100's level. I do not think Google should trade below this index. On the companies' reported figures, its cloud business is growing fastest among the three largest providers, its margin is rising, and search is still growing.
The multiple has changed since July. When the July report appeared, Google traded at 23.9 times, with consensus earnings per share of about $13.37. That figure is now about $14.35, up 7%, while the stock is down 12.5%. Part of that 7% simply reflects time passing. Both figures cover the next 12 months. Three months later, that 12-month window has moved forward by a quarter, so this does not prove that analysts raised earnings estimates.
In July, my fair-value midpoint was $325, corresponding to roughly 23–25 times earnings. This time it is $316, or 22 times. The multiple is lower because the search question will remain open until the second half of 2027, and financing has become heavier. Until then, my fair-value band stops at 23 times. My method has also changed. In July, I valued the separate businesses and added them together. This time I use consensus earnings multiplied by a P/E ratio, the market's most common approach. My model treated all of this year's capital spending as an immediate cost, although the capacity will earn money for many years. That produced too low a value, below even the S&P 500's multiple.
The most recent concentrated market doubt about Google search came on May 7, 2025. Apple's services chief Eddy Cue testified in the U.S. Justice Department's antitrust case against Google. He said Apple was considering AI search as a replacement for Google as the default in the iPhone's Safari browser. GOOGL dropped 7.3% that day to close at $151.38. Its P/E ratio decreased from 17.8 times the day before to 16.5 times, then stood at 16.8 times the following day. The S&P 500's forward P/E was about 20.5 times, leaving Google roughly 4 multiple points below the market. Over the next four quarters, search revenue grew between 12% and 19%. A year later, the stock was $397.99.
An earlier episode came on February 8, 2023. Google's AI chatbot Bard gave a wrong answer in a promotional video. The stock dropped 7.7% that day, closing at $99.37. Search growth recovered to 11% and 13% in the third and fourth quarters of 2023. A year later, the stock was $145.91. I do not have the P/E ratio for that day, so I use it only as supporting evidence. The 16.2 times recorded on April 8, 2025, was lower, but that was a market-wide tariff selloff. I do not count it as a search scare.
When search came into question, the lowest multiple the market gave Google was roughly 16–17 times. We know the subsequent moves in both episodes only in hindsight and cannot simply repeat them as a forecast. The last scare proved unfounded, but the agent question is different. It challenges the way search charges for its service. So 16–17 times is a reference, not a guarantee. Google now trades level with the broader market. Relative to that market, its discount is not yet as deep as last time.
Today's 19.5 times therefore leaves 13–18% downside to the last search scare's low and roughly 13% upside to my fair-value midpoint. The distances are similar. But there is another reason to remain bullish beyond this comparison. Signed cloud contracts become revenue quarter by quarter. Wall Street expects earnings to keep growing, so that reference floor can move.
For example, according to Investing.com's account, Wells Fargo estimates 2028 GAAP earnings per share of $18.20. GAAP earnings follow accounting standards. If the market still assigns only 16–17 times in 2028, the implied share price is about $291–309, above today's price. This is only an illustration. Applying a forward P/E ratio to 2028 GAAP earnings is not the same measure as the table above. The $18.20 estimate already decreased 4% after earnings and could decrease again. If search keeps slowing, the $14.35 figure will also decrease, shifting every band in the table downward.
Among 26 firms' post-earnings actions, 22 cut their price targets, 3 left them unchanged, and 1 downgraded its rating. None raised a target. Price targets are analysts' estimates of what the stock should be worth a year ahead.
JPMorgan considers the search headwinds “more navigable than permanently structural.” MoffettNathanson retained its Buy rating but said this was the first quarter in a long time that gave the bearish argument something to work with. Bernstein has a Neutral rating and put it plainly. Search is why investors own Google. The explanations make sense, but the market has no patience for them now. Bernstein also thinks Google could return to the front rank as capacity comes online early next year and advertising in AI interfaces gains scale.
Some firms cut the multiple directly. These multiples use 2027 earnings, so they are not the same measure as the 19.5 times next-12-month earnings used above. According to Investing.com, Scotiabank reduced its multiple from 30 times 2027 earnings to 24 times. Wells Fargo reduced its multiple from 27.5 times to 25 times. Revised price targets range from DA Davidson's $300 to $460 at Mizuho and TD Cowen. Citizens, which downgraded its rating, gave no target, leaving 25 targets among 26 firms. My $316 midpoint is below 24 of those 25 targets. It estimates fair value today without restoring search's multiple in advance.
The cost of rented capacity will enter the accounts, its use is increasing, and TPU system sales are expanding. I will watch the depth of the margin decline. I will also watch whether management describes it as a temporary rental cost or a lasting change in the product mix.
About $175B in cash and bonds gives Google considerable room. I will watch how much operating cash flow covers and how much comes from debt and new shares. I will also watch whether ATM sales keep exceeding employee tax payments and whether buybacks remain at zero. The more incremental spending relies on operating cash flow and debt, the less dilution shareholders bear. Dilution means more shares and less profit attributable to each one.
As AI Mode ads launch outside the United States, I expect paid-click growth to move back toward query growth. The first disclosure of AI Mode monetization figures would matter more than any single quarter's growth rate. I also expect services margins to hold around 40% after excluding one-time items.
I will watch three things. Do ads actually appear in agent workflows? Does management begin discussing revenue per task? Does Google Cloud begin separately discussing revenue from agents? This question will not end in any one quarter. It will accompany every earnings report through 2027.
Suppose this bullish view has proved wrong by spring 2028. I suspect the likeliest outcome looks like this. No single business collapses. Agents and AI interfaces gradually take commercially motivated searches, while Google's new charging methods fail to keep pace. Search growth settles in the single digits. AI service costs and depreciation weigh on services profit. Cloud keeps growing, but an increasing share comes from TPU hardware, accompanied by hardware repurchase commitments and related-party contracts. Those counterparties are companies in which Google holds stakes. The market values that business as hardware. More new shares help fund the 2027 construction program, gradually diluting value per share. None of these developments is fatal alone. Together, they would make today's price look less cheap.
A large share of capital spending is chasing $561B in signed contracts. Cloud revenue has doubled while its margin rose to 36.2% over a year. Today's search slowdown reflects currency, a tough comparison, and lower earnings from AI Mode. The answer will have to wait until the second half of 2027. Who pays in the agent era will take longer to answer. Google has Gemini and sells computing capacity, but nobody can yet calculate the scale. Cloud margins, July's deciding factor, passed the first round this quarter. Contracts are still awaiting delivery. At $279.79, the price has already come down. I remain bullish.
Paxis Alpha Research prepared this report using Google's third-quarter 2026 earnings release, Form 10-Q, and the third-quarter and earlier earnings calls. Sources also include public market valuation data and post-earnings media and sell-side coverage. Publication date: October 30, 2026.
This research is for information only. It is not investment advice or a recommendation to buy or sell any security. The valuation bands describe what different prices imply. They are not trading instructions. Readers should conduct their own due diligence and consult a licensed financial adviser before making any investment decision. Past performance does not indicate future results.
A revenue recovery does not establish a profit recovery. The report awaits the lower comparison base in Q3 2027, but a rebound in year-over-year growth alone cannot distinguish that effect from better AI monetization. Even if paid clicks catch up with query growth, incremental advertising revenue still has to cover incremental compute costs. The report’s revenue and click data cannot establish profit per query; that calculation is still needed to test the bullish thesis.
Impact: Even after search growth recovers, profit per query still needs verification.