Alphabet is not a business in decline. Its revenue grew 24.7% in the third quarter of 2026 to $127.6 billion, Google Cloud doubled, and the Gemini app passed one billion monthly users. The stock is another matter. On 29 October 2026, the day after the third-quarter report, it fell 12.64%, the worst day in Alphabet's history, and by 30 October it stood 30.5% below its May record. Two things broke at once: Search, the business that still pays for everything, grew more slowly than expected, and the cost of the AI build-out pushed free cash flow below zero. Underneath both is something larger: Google is turning from a search monopoly into an AI infrastructure company.

-12.64%29 October 2026, the worst day in Alphabet's history, about $476 billion of market value
-$9.1BFree cash flow in Q3 2026, against about -$2.7 billion expected
+103%Google Cloud revenue growth in Q3 2026, to $30.7 billion
25.6xThe real trailing P/E once paper gains on Anthropic and SpaceX are removed, not 14.6x

This article follows the argument of the video: the crash and the valuation, the businesses behind it, the two causes of the fall (weaker Search and negative cash flow), the cloud and chip business the money pays for, the competition, the question of AGI against products, Google's strategy, its core transformation and its new identity, and finally the outlook. Market figures are anchored to the 30 October 2026 close of $279.84 unless another date is given; financial figures run to the third quarter of 2026.

The Crash

On 28 October 2026 Alphabet closed at $324.89. That evening it reported its third quarter, and the next day it closed at $283.82, down 12.64%. The previous worst day in its history was a fall of 11.63% on 16 March 2020, at the start of the pandemic; the next three on the list are from 2008 and 2023. By our count on the share numbers on its cover pages, about $476 billion of market value disappeared in one session; CNBC put it at "roughly $500 billion". Volume was 4.5 times the 30-day average. The stock slipped another 1.40% the next day, to $279.84.[1][2]

The fall follows a run that few large companies have matched. From its tariff low of $144.70 on 8 April 2025 the stock rose 178.2% in thirteen months, to a record close of $402.62 on 13 May 2026, worth about $4.86 trillion, as the risk of a court-ordered break-up of Search faded and Google Cloud took off. On 30 October it was 30.5% below that record. Big falls are nothing new for Alphabet: in the 2022 bear market it lost 44.3%, from $149.84 in November 2021 to $83.43 in November 2022, and it took until January 2024 to close above the old peak. Every fall of 20% or more since 2020 ended at a new high; the current one has not yet.[1]

The selling did not stop with the earnings. On 5 November 2026 the Nasdaq 100 fell 6.3% in an AI sell-off after Democratic gains in the midterm elections raised the prospect of slower data centre building. Alphabet fell only 2.2%, to $282.57, which suggests that much of the bad news was already in its price.[3]

What the valuation says now

At $279.84 Alphabet trades at 14.6 times its trailing earnings per share of $19.16, which looks like the cheapest of the large technology companies: Microsoft is at 27.3, Nvidia at 25.9, Meta at 21.0 and Amazon at 19.4. The number is misleading. 43% of Alphabet's trailing EPS came from paper gains on its stakes in other companies, mostly Anthropic and SpaceX, booked as the value of those stakes rose. In the first nine months of 2026 these gains were $99.1 billion after tax, 49% of the $201.2 billion of net income. Remove them and trailing EPS is $10.95, a P/E of 25.6, in line with its peers rather than below them. From 2021 to 2025 Alphabet's year-end P/E ranged from 19.3 at the end of 2022 to 29.0 at the end of 2025.[1][4]

The gains run both ways

The gains are not cash. In the third quarter the stakes lost value instead, and reported EPS of $2.12 was below the $2.67 of the underlying business. SpaceX, which listed at $135 in June, closed at $134.72 on 30 October, so the fourth quarter starts with a loss on that stake too.

Understand the Business to Understand the Crash

Alphabet had revenue of $127.6 billion in the third quarter of 2026. Search and other brought in $63.4 billion, 49.7% of the total, Google Cloud $30.7 billion (24.1%), subscriptions, platforms and devices $14.3 billion (11.2%), YouTube ads $11.2 billion (8.8%), Google Network, the ads Google places on other websites and apps, $7.5 billion (5.8%), and Other Bets, the moonshots such as Waymo, $0.37 billion (0.3%).[4]

Behind those lines sit some of the largest consumer products in the world. YouTube reached a record 15.8% of all US TV viewing in August 2026, against 9.3% for Disney and 7.7% for Netflix. Google counts 350 million paid subscriptions, including YouTube Premium and its Google One AI plans. Android powers 69% of mobile web traffic and Chrome carries 66% of browser traffic. Waymo gave more than 500,000 paid rides a week in April 2026 and was valued at $126 billion. The moonshots are expensive, though: from 2021 to September 2026 Other Bets lost $31.9 billion on $7.7 billion of revenue, about $4 for every dollar it brought in.[5][6][4]

Search is becoming less relevant

The most important change is in the mix. Search was 57.8% of revenue in 2021, then 57.4%, 56.9%, 56.6% and 55.7% in the following years. Over the last five quarters it went 55.3%, 55.4%, 55.0%, 52.8% and 49.7%, the first quarter below half in the data, 8.1 points lower than in 2021. Search did not shrink: it still grew 12% in the third quarter. Cloud simply grew far faster, from 7.5% of revenue in 2021 to 14.6% in 2025 and 24.1% in the third quarter of 2026, up 16.6 points. The ad lines lost ground too: YouTube ads fell 2.4 points, from 11.2% to 8.8%, and Google Network 6.5 points, from 12.3% to 5.8%.[4]

The transformation in one line

Google is fundamentally transforming from a search monopoly into an infrastructure and platform company. Cloud supplied 61.6% of Alphabet's revenue growth in the third quarter of 2026.

Crash Cause 1: Search Growth Weakened

Google is still mostly a search business, so investors react to any weakness there. Search grew, but more slowly: revenue rose 12.0% in the third quarter, from $56.57 billion to $63.35 billion, down from 19.1% in the first quarter and 16.8% in the second. It came in 2.5% below the $64.96 billion consensus reported by CNBC. Paid clicks grew 8%, against 13% in the second quarter, while the cost per click still rose 4%. Against the second quarter, Search revenue rose only 0.1%, from $63.27 billion to $63.35 billion.[4][2]

Google gave four reasons, in this order: currency, the end of duty-free small parcels in the EU, which hit Asian retailers advertising in Europe, the timing of the World Cup, and lower monetisation in AI Mode, its conversational search. That last reason came almost in passing, in Philipp Schindler's words "the lower monetization we were seeing in AI Mode", but it is the one that frightened investors, because it is the only one that grows as more people use AI search. The CFO, Anat Ashkenazi, added that about 3 points of the slowdown came from lapping a strong quarter a year earlier, and that the fourth quarter faces "a couple of points" more.[7]

Is Search getting worse?

Not necessarily. The profit side looks healthy. Google Services, which holds Search, YouTube and the subscriptions, earned an operating margin of 40.7% in the third quarter, up from 38.5% a year earlier, as efficiency gains more than offset the cost of AI. Even with all Alphabet-level costs charged to Services, AI research, fines and legal included, the margin was 35.8% in the first nine months of both 2025 and 2026, by our math. Google says the cost of serving Gemini per unit fell 78% over 2025, and the cost of core AI answers in Search fell 30% after Gemini 3. AI answers are still more expensive than plain search: on Google Cloud's developer price list they cost $4 to $8 per 1,000 queries against $1.50, 2.7 to 5.3 times as much.[4][7][8]

Our reading: Search is getting weaker because of AI, but Google is managing the transition to AI search without giving up its margins.

Do Agents Dream of Sponsored Ads?

Do AI searches still generate clicks? Less so. In a randomised study published as a preprint in March 2026, users who were forced into AI Mode, which had no ads at the time, clicked 42.7 points fewer ads and 18.8 points fewer website links. Pew found that users clicked a result on 8% of Google searches with an AI summary, against 15% without one. SparkToro counted 68% of US Google searches ending without any click in early 2026, against 60.45% in 2024, on different panels. And the new traffic does not replace the old: in Ahrefs' sample of 76,000 sites, ChatGPT sent 0.21% of all visits, against 39.98% from Google.[9][10][11]

Google's own numbers are more resilient than the studies suggest: paid clicks still grew 8% in the third quarter, and ads in AI Mode only went international in October 2026. The harder question is what happens when AI agents shop on the user's behalf. Agents browse, but they rarely buy yet: HUMAN Security found that only 2.4% of agent browser activity reached checkout pages in May 2026. Serving ads to agents is a complicated puzzle Google still has to solve.[7][12]

Crash Cause 2: Negative Cash Flow

This is the big one, and the same reason Meta fell 7% after its own report a week earlier. Alphabet's capex rose 114% in the third quarter, from $24.0 billion to $51.4 billion, equal to 40.3% of revenue against 23.4% a year earlier. Operating cash flow was $42.3 billion, so free cash flow was minus $9.1 billion, about $6.4 billion worse than the minus $2.7 billion analysts expected according to Yahoo Finance. It was the second negative quarter in a row, after minus $5.9 billion in the second.[4][1]

And it is getting bigger. Alphabet guides 2026 capex to $195-205 billion, up 119% at the midpoint from $91.4 billion in 2025 and about eight times the $24.6 billion of 2021. Every change to the guidance has been upwards: four raises since 2025, from $75 billion to about $85 billion and then $91-93 billion for 2025, and from $175-185 billion to $180-190 billion and then $195-205 billion for 2026. Purchase commitments reached $924 billion at maximum contract limits on 30 September 2026.[7][4]

From buying back stock to raising money

Alphabet stopped buying back its own shares and started raising money instead. It bought back nothing in 2026, after $45.7 billion in 2025 and more than $50 billion a year from 2021 to 2024. In June 2026 it sold $50 billion of stock: $20.7 billion to the public, $10 billion to Berkshire Hathaway and $19.25 billion of convertible preferred shares. It sold another $12.4 billion through an at-the-market program of up to $40 billion. Debt rose 160%, from $48.5 billion at the end of 2025 to $126.1 billion in September 2026, including a 100-year sterling bond. The share count rose 1.6% in 2026, after falling every year from 2021 to 2025. Investors who bought in June, Berkshire included, were down about 20% at the 30 October close. In the first nine months of 2026, 54% of the money Alphabet had came from outside capital rather than its own operations.[4][1]

Investors are scared because costs are rising faster than revenue, and the future revenue the capex is meant to produce is uncertain. Politics added to it: after the Democrats' election gains, there is more uncertainty about limits on data centre building and AI regulation. In the US, Democrats are more inclined to regulate AI, which could jeopardise Google's investments in both the short and the long term.

The AI Race: Google's Cloud Business

What does all the capex pay for? The clearest answer is Google Cloud. Its revenue rose 103% in the third quarter, from $15.2 billion to $30.7 billion, against 39% at AWS and 48% at Azure in the same quarter. It is also profitable now: an operating margin of 36.2%, against a loss of 11.9% in 2021. Cloud supplied 61.6% of all of Alphabet's revenue growth in the quarter.[4][13]

The pipeline is even larger. Google Cloud's backlog, the signed contracts not yet booked as revenue, grew 262% in a year, from $155 billion in September 2025 to $560.8 billion in September 2026. Just over 55% of it, about $308 billion by our math, turns into revenue within 24 months. Anthropic's IPO filing alone commits at least $114.5 billion to Google.[4][14]

TPU: Google's own chip

Much of that demand is for Google's own AI chip, the TPU (Tensor Processing Unit), which Google has developed "for a decade" as its alternative to Nvidia's GPUs. Google sells it as one package, a full system of chips, networking, storage and software, not just a chip, according to the CFO. Customers can get it in two ways: rented in Google Cloud, or, since 2026, bought as hardware for their own data centres, with most of that revenue arriving in 2027. Google says it has more than 100 significant TPU customers, including Anthropic and Apple. Its inventory rose 446% in 2026, from $2.4 billion to $13.3 billion, driven by TPU systems that Thomas Kurian, head of Google Cloud, says are all tied to signed contracts.[7][4]

The technical case rests on scale and price. One TPU pod links 9,216 chips, against 72 GPUs in one Nvidia NVLink group, and on Google Cloud's list prices a TPU costs 24% to 26% less per chip than Nvidia's B200. In October 2026 Google and Meta open-sourced TorchTPU, which runs PyTorch models on TPUs, so customers can leave Nvidia more easily. Google is working to build cheap, scalable and open compute for data centres.[15]

Cloud Competition

Cloud is a competitive market led by Amazon and Microsoft. The difference is what depends on it. For Microsoft, Amazon and Meta, the core business does not depend on cloud AI the way Google's Search does. AWS earned 63% of Amazon's operating income in the third quarter ($17.8 billion of $28.1 billion), and Azure brought in 36% of Microsoft's quarterly revenue ($33.2 billion of $92.3 billion), but neither company's main product needs cheap AI to survive. Meta has no cloud business at all and uses its MTIA chips in-house. Google needs Cloud and its TPUs to work to protect the margins of Search in the AI era; on its fourth-quarter 2024 call it spoke of "our obsession with cost per query", and Gemini 3 and all earlier models were trained on TPUs, according to SemiAnalysis.[13][15]

With aggressive spending, Google could overtake its rivals within one to two years. In the third quarter Google Cloud had $30.7 billion of revenue, Azure $33.2 billion and AWS $46.0 billion. If each kept its current growth rate, with AWS at Amazon's own 40%, Google Cloud would reach about $62 billion in the third quarter of 2027, against $49 billion for Azure and $64 billion for AWS. To pass Azure by then Google needs growth above 60%; to pass AWS, above 110%. That is our illustration, not a forecast. Spending points the same way: over the last twelve months Amazon spent $215 billion of capex, Google $160 billion and Microsoft $136 billion, all company-wide, and by Epoch AI's count Google owned 5.47 million H100 equivalents of AI chips at the end of 2025, against 3.59 million at Microsoft and 2.52 million at Amazon.[13][16]

Google is also first to market in selling its chips. On Amazon's October call, Andy Jassy said only "effectively two" companies have a cloud and a chip strong enough to sell outside. Amazon plans to sell Trainium outside AWS from late 2027, and Oracle starts offering Google's TPUs in its cloud in 2027. Google's bet is AI efficiency at the scale of its own products, such as an AI summary with every Google search at low cost: AI Overviews had more than 2 billion monthly users in mid-2025 and AI Mode more than 1 billion in mid-2026, and tokens processed through Google's APIs grew 4.6 times in a year, to nearly 32 billion a minute.[13][7]

Uniquely positioned

Google already leads in owned compute, is first among the cloud providers to sell its own chips outside, and has the fastest growth of the three largest clouds. None of the three discloses cloud-only spending, so the comparison of capex is company-wide.

AGI or Products?

Why does Google want so much compute: frontier models, artificial general intelligence or infrastructure? The people tell part of the story. Researchers left in 2026 and Google became more product-focused. The stock fell 4.99% on 22 June 2026 after Gemini co-lead Noam Shazeer left for OpenAI and Nobel laureate John Jumper for Anthropic, and 4.03% on 5 August 2026, when Demis Hassabis stepped back from running DeepMind and Jeff Dean left after 27 years. According to Zeki data reported by Fortune, DeepMind hired about two people for every leaver in the third quarter of 2026, against twelve in the second quarter of 2023, and 25% of its leavers went to Anthropic, 21% to Meta and 14% to OpenAI. None of them gave a public reason; engineers told Fortune of "frustration at falling behind on coding benchmarks" and "chasing pre-IPO equity".[17][1]

The new structure points to a pragmatic focus on revenue rather than AGI and AI gimmicks: one team now runs the models, the Gemini app and the developer teams, under Koray Kavukcuoglu. Google still names AGI as the first priority for its TPUs: "That is the foundation for everything we do", it said on the second-quarter call. In practice, AI products look more important than AI research.[7]

AGI Dreams

Is Google still in the race at the frontier? In October 2026 it launched Gemini 4 Argon and announced Gemini 4 Carbon on 28 October, though Carbon is not yet generally available. On 5 November Carbon ranked first on the Artificial Analysis index with 60 points, against 59 for Claude Opus 5.5. Benchmarks are often of little relevance, though, and the record is mixed: in July 2026 Google started its "most ambitious pre-training run yet" for Gemini 4, while Gemini 3.5 Pro, promised at I/O in May 2026 for June, missed three release dates, according to Fortune. Gemini 5 would be the first model built mostly after the exits, and so the real test of the new team.[18][17]

The frontier is uncertain, but the product side looks very good. The Gemini app grew from 90 million monthly users in October 2024 to more than 1 billion in August 2026, up 1,011%, by Google's count; Sensor Tower counted 662 million in May 2026, so the scope differs. Gemini had 25% of chatbot web traffic in September 2026 against 56.4% for ChatGPT, up 19 points from 6% in March 2025, according to Similarweb. Pew found that 24% of US adults use Gemini, against 44% for ChatGPT, in February 2026.[19]

Overall Strategy

So what is happening to Google, and what is its strategy? Its core search business is going through an AI transformation. What used to need backlinks and algorithms now needs expensive inference on every search. Google's own agents are replacing Google Search as the front end, while Search becomes the back end. Hundreds of thousands of brands can now sell through Google's agent checkout, built on its Universal Commerce Protocol (UCP), plus a Universal Cart, according to the third-quarter call.[7]

That makes cost efficiency, its own AI models and ownership of the supply chain a must to protect margins. The cloud business is essential, which is why Google invests more aggressively than Microsoft and plans close to Amazon for 2026: $195-205 billion against about $220 billion of company-wide capex. Being first to sell its chips outside shows the urgency. If Google keeps up the growth and the investment, it could pass Microsoft in cloud, and eventually Amazon.[13]

Core Strategic Transformation

Google used to have a near monopoly in search, and by the numbers it still does: 90.5% of worldwide search in 2026, according to StatCounter. In the US its share fell 2.7 points, from 88.1% in 2021 to 85.4% in 2026, while Bing rose from 6% to 9.6%. In April 2025 Safari searches fell for the first time in 22 years, as Apple's Eddy Cue testified in court. Yet in the third quarter of 2026 query growth was the highest in five years, according to Sundar Pichai.[20][7]

That raises two open questions. Is Google Search still a tech monopoly if it is the back end serving agents? And will Gemini's results be best because of Google Search, or does search itself become irrelevant and everything comes down to how the agent presents the answer?

From click to conversion

Google's search segment is shrinking in relative terms because Cloud grows so fast, but the business itself is changing too. Google is shifting from serving search results alone to serving AI alongside them, powered by those results. That makes ads and leads less valuable on their own and the purchase more important for the advertiser. Google is gradually moving down the value chain from the click to the conversion in its products, while it is still paid per click.

How can Google earn money from economic activity in the age of AI? The early evidence is mixed. Walmart's Daniel Danker said its checkout inside ChatGPT converted three times worse than click-outs to Walmart's own site. Like Meta, Google is trying to take a share of the transaction, but this is a new and unproven market: Mark Zuckerberg's plan for Meta is to "take a very small cut of whatever the transaction is", while Google has disclosed no agentic commerce revenue at all.[21]

Sales, ads and leads are moving through AI, and it is not clear how much of the value an AI provider can capture. In Invoca's July 2026 data, 49% of phone calls referred by ChatGPT were leads, the highest share of any channel, but the volume was "very low". eMarketer expects $32 billion of US AI advertising in 2026, but less than $1 billion of it inside standalone chatbots. ChatGPT's ads had an annual run rate of $1 billion in August 2026, about $250 million a quarter, or 0.3% of Google's $82 billion of quarterly ad revenue. The Gemini app carries no ads yet. Is the game in inference and training, or also in the chatbot layer beyond subscriptions and credits? Can Google learn to monetise agents the way it monetised search results?[12][22]

For Google

It owns the most-used search engine, a fast-growing chatbot, its own chips and models, and the agent checkout plumbing. Paid clicks still grow and query growth is at a five-year high.

Against

AI answers cost more to serve and earn less per query, users click less, agents rarely buy yet, and no one, Google included, has shown how much of a transaction an AI provider can keep.

New Identity

Google's identity is moving from search monopoly to AI infrastructure and economy competitor, from the one of one to one of many. In cloud it is the third player, with 15% of the market against 28% for AWS and 20% for Azure, according to Synergy Research for the second quarter of 2026.[23]

It is also moving from builder to backer. At 30 September 2026 Google had $75.7 billion of guarantees on partners' data centre financing, so that third parties can build sites that run Google TPUs: $48 billion already signed and $27.7 billion more promised, $10 billion of it for Crux AI, Google's TPU cloud venture with Blackstone. The guarantees make the money cheaper: data centres with Google guarantees borrow at about 7.1%, against 9.3% for those built around Nvidia chips, according to the Financial Times via Channel Insider.[4][24]

If Google can show that TPUs are cheaper and more efficient to run in AI data centres, it gains a strong competitive advantage. The evidence so far points both ways. SemiAnalysis estimates that an Ironwood TPU costs Google 44% less all-in than an Nvidia GB200 server, and in its InferenceX test Ironwood produced 50% more tokens per dollar than Nvidia's B200. Against that, in the only public MLPerf head-to-head, on one image model, Nvidia's B200 was 2.5 times faster per chip than Google's older Trillium TPU, and SemiAnalysis also found Nvidia's GB300 about 30% better on performance per dollar than Ironwood in the middle of the latency range.[15]

TPU ahead

44% lower all-in cost per Ironwood than per GB200 server for Google itself, and 50% more tokens per dollar than B200 at 20 tokens per second per user (SemiAnalysis).

Nvidia ahead

2.5 times faster per chip for B200 against Trillium in MLPerf, and about 30% better performance per dollar for GB300 against Ironwood mid-range (SemiAnalysis).

Outlook

Four of the last eight earnings reports sent the stock down the next day, and every large drop came with capex or cash flow news: minus 7.29% after the fourth quarter of 2024, when the first $75 billion capex guide came out, minus 7.13% after the second quarter of 2026, when capex was raised again, and minus 12.64% now. The fourth quarter of 2025 closed only 0.54% lower but opened 6.25% down on the first 2026 guidance. The good days came with Cloud: plus 9.96% after the first quarter of 2026. The average move after earnings is 5.5%, but the last three were 10%, 7% and 13%.[1]

In the short run, cash flow stays negative. To reach its guidance Google needs $63-73 billion of capex in the fourth quarter, up 126% to 162% on a year earlier, against a record quarterly operating cash flow of $52.4 billion, so free cash flow is set to stay negative next quarter by our math. The S&P consensus of minus $18.2 billion for 2026 implies about minus $13.4 billion in the fourth quarter. If Search weakens further, the stock is in trouble: the 12.64% fall followed a 2.5% miss in Search revenue, and by our math the lapping effect adds 2.2 points of headwind in the fourth quarter and 2.4 points in the first quarter of 2027.[25][4]

Google is now sprinting to lead in AI infrastructure and will not stop investing aggressively. It says only that 2027 capex will "increase significantly"; outside expectations, reported by CNBC and Fortune and from BofA's Justin Post, are around $300 billion. Only regulation could slow the capex now, and even that is uncertain: 76.9% of Google's long-lived assets sit in the US, but agents are already here, and slower AI regulation would not save Search. AI does not need to be smarter to replace search; what is left is user behaviour and better products. 45% of US adults already use AI for local recommendations, up from 6%, according to BrightLocal.[25][26]

Our view: the long-term upside is very large, but so is the volatility, driven by Search weakness, capex and cash flow. The average analyst target of $400, from 69 analysts on TradingView, is 43% above the 30 October close. That is the analysts' view; ours is that Google has the strongest hand in AI infrastructure and the hardest transformation of its core business, both at the same time.[1]

Good

Google Cloud up 103% at a 36.2% margin, a $560.8 billion backlog, the most owned AI compute, first to sell its own chips, a 40.7% Services margin, a Gemini app with over a billion users and a P/E of 25.6 on gain-free earnings.

Bad

Search slowing to 12% with AI Mode earning less, free cash flow of minus $9.1 billion, capex of $195-205 billion and rising, debt up 160%, buybacks stopped, shares diluted, research leaders gone, and no proven way to earn from agents.

Sources and chart methodology

  1. TradingView daily closes for GOOGL and GOOG, 2004 to 30 October 2026, split-adjusted and not dividend-adjusted, plus the TradingView quote on 5 November 2026 and the analyst snapshot of 31 October 2026 (69 analysts, average target $399.96). Falls are close to close from a running high. Market value uses the share counts on Alphabet's latest cover page. Peer P/E ratios are TradingView's on 30 October 2026, as reported. Earnings reactions are the first regular session after each release.
  2. CNBC on Alphabet's worst day, 29 October 2026, for "roughly $500 billion", the prior record and the Search consensus of $64.96 billion.
  3. NBC News on the 5 November 2026 tech sell-off.
  4. Alphabet SEC filings: Q3 2026 10-Q, the FY2021 to FY2025 10-K filings and the quarterly earnings releases. Free cash flow is operating cash flow minus purchases of property and equipment. The P/E excluding gains removes the after-tax effect of gains and losses on equity securities given in each release. Revenue shares, segment margins and the share of growth from Cloud are our math. The 2021 Cloud margin is the like-for-like recast in the FY2023 10-K.
  5. Nielsen's Gauge for August 2026.
  6. StatCounter mobile operating system share and browser share, September 2026, which measure web traffic, not devices in use.
  7. Alphabet earnings call transcripts, Q4 2024 to Q3 2026 (Yahoo Finance transcriptions), for the Search reasons, lapping, AI cost reductions, capex guidance, TPU descriptions and customers, AI Mode and AI Overviews users, tokens, query growth, agent checkout and the AGI quote.
  8. Google Cloud developer price list for search and AI answer queries.
  9. Wang et al., randomised study of AI in search, March 2026 draft, a preprint not yet peer reviewed.
  10. Pew Research Center on clicks with AI summaries and SparkToro on zero-click searches; SparkToro's 2024 and 2026 figures come from different panels.
  11. Ahrefs on traffic from ChatGPT and Google.
  12. HUMAN Security's May 2026 benchmark, via PPC Land, and Invoca's 2026 lead conversion report.
  13. Peer filings and calls: Amazon's Q3 2026 earnings release and call (AWS growth and operating income, "effectively two", Trainium), Microsoft's September 2026 10-Q (Azure revenue on its new definition) and Oracle's 2027 TPU plan. The Q3 2027 cloud figures are our illustration at constant growth rates, not a forecast. Capex is company-wide over the last twelve months.
  14. Anthropic's public S-1, 26 October 2026, for its minimum commitment to Google.
  15. SemiAnalysis: TPUv7 analysis, November 2025 and InferenceX on TPUs, September 2026; MLCommons MLPerf results; Google Cloud list prices for the per-chip comparison with B200; InfoQ on TorchTPU, October 2026.
  16. Epoch AI data on AI chip owners, end of 2025, in H100 equivalents.
  17. Fortune: DeepMind talent data from Zeki, DeepMind's identity crisis and the June departures, plus Fortune on the Gemini 3.5 Pro delays.
  18. Artificial Analysis on Gemini 4 Carbon and its Intelligence Index v5 leaderboard, 5 November 2026.
  19. Gemini users from Google's calls and Google's August 2026 post, with the October 2024 and March 2025 figures from a court slide via 9to5Google; Sensor Tower's May 2026 estimate; Similarweb on chatbot traffic; Pew Research Center, February 2026.
  20. StatCounter search engine share, worldwide and US; Fortune on Eddy Cue's testimony.
  21. MarTech on Walmart's ChatGPT checkout, reporting WIRED; Mark Zuckerberg on the Sources Podcast, 8 September 2026.
  22. eMarketer's forecast of US AI advertising, 2026; OpenAI on ChatGPT's ad run rate, August 2026. The 0.3% compares the quarterly equivalent with Google's Q3 2026 ad revenue.
  23. Synergy Research data for Q2 2026, via Tech Insider.
  24. Channel Insider on Google's TPU financing network, citing the Financial Times.
  25. Fortune on the Q3 2026 report and CNBC on 2027 capex expectations; S&P consensus via StockAnalysis, not confirmed as updated after Q3. Implied fourth-quarter capex, free cash flow and lapping are our math.
  26. BrightLocal Local Consumer Review Survey 2026.