BusinessMarkets, money & strategy
Alphabet shares fall more than 10% premarket after first Q3 double miss since 2022 as CFO links search slowdown to AI Mode
Alphabet missed Q3 revenue and EPS forecasts for the first time since late 2022, and its CFO tied slower search ad growth to AI Mode. Shares fell over 10% premarket as $51.4 billion in capex drove free cash flow negative.
From the CNBC · Technology article of October 29, 2026 and cited sources: Q3 revenue was $127.6B against a $128.2B estimate, and adjusted EPS was $2.67 against $3.02. [1] [4] [7] [8]
SOURCES[1] Alphabet drags on Nasdaq futures as the stock faces a… cnbc.com[4] Reuters via Investing.com: Alphabet stock sinks as search… investing.com[7] Alphabet Q3 2026 earnings release (SEC 8-K) sec.gov
AT Pulse research deskPublished · Updated · 5 min read · AI-assisted analysisAI-assisted, not human-edited. Methodology Corrections Who owns this site
- What happened
- Q3 revenue was $127.6B against a $128.2B estimate, and adjusted EPS was $2.67 against $3.02. [1] [4] [7] [8]
- Source caveat
- Two days after the Gemini 4 Argon API release, which supports up to about 1M output tokens through long decode, Google announced Gemini 4 Carbon on the earnings call. [5] [8] [11] [12]
- Original source
- CNBC · Technology article, independent perspective
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- October 29, 2026 article published
- Covered
- Oct 29 edition
- Evidence
- 12 sources across 11 sites
- 1Alphabet drags on Nasdaq futures as the stock faces a…cnbc.com
- 2Alphabet earnings miss reflects first real sign that AI is…cnbc.com
- 3Google shares tumble 10% after $51bn AI outlay deepens…ft.com
- 4Reuters via Investing.com: Alphabet stock sinks as search…investing.com
- 5Google AI for Developers: Gemini 4 Argonai.google.dev
- 6gori.me: Gemini 4 Argongori.me
Reach: not measured yet
Q3 revenue was $127.6B against a $128.2B estimate, and adjusted EPS was $2.67 against $3.02. Search revenue of $63.35B missed for a second straight quarter. CFO Anat Ashkenazi linked a larger-than-expected slowdown in search ad growth to heavier use of AI Mode. Google also says search query growth is at a five-year high, so the pressure appears to come from lower revenue per query rather than lost usage. [1] [4] [7] [8]
Spending is outpacing cash generation. Q3 capex of $51.4B pushed free cash flow to −$9.1B, after −$5.9B in Q2. The Q3 figure also absorbed a $5.2B European Commission fine and a build-up of TPU inventory. Guidance calls for $195–205B in 2026 capex and a significant increase in 2027. Google Cloud grew 103% and has a $561B backlog, but investors focused on Search and cash burn. [4] [9] [10]
Two days after the Gemini 4 Argon API release, which supports up to about 1M output tokens through long decode, Google announced Gemini 4 Carbon on the earnings call. Carbon's access is gated, starting with cyber defenders and US government testers, because Google says it is especially strong in cybersecurity. No Carbon model card or technical report is available. Its #1 Artificial Analysis Intelligence Index score of 59 is cited by Google and not independently confirmed in the reviewed sources. [5] [8] [11] [12]
Alphabet's CFO tied slower search ad growth to AI Mode even as query volume rose. Implication: AI-style search may earn less per query, and record infrastructure spending leaves less room for error if ads in AI Mode take time to monetise.
Executive brief
Alphabet reported its first miss on both revenue and earnings per share since Q4 2022. On the same call, its CFO linked a "larger-than-expected deceleration" in search advertising growth to heavier use of AI Mode, Google's chatbot-style search. Investors had long feared that AI would start eating into Search, and this is the first quarter in which Alphabet's own explanation points that way. Shares fell more than 10% before the market opened on October 29. A session drop above 9.77% would be the stock's worst post-earnings reaction on record (CNBC). Google Cloud more than doubled. Capital spending of $51.4 billion pushed free cash flow to −$9.1 billion (Reuters via Investing.com).
What changed and event timeline
Gemini 4 Argon ships
Google released its flagship reasoning model, pitched at software engineering, maths and agents. It can generate up to about 1 million output tokens with a "long decode" feature (;).
Q3 results show a double miss
Revenue was $127.6B against a $128.2B estimate. Adjusted EPS was $2.67 against $3.02 expected. Search revenue was $63.35B against a $64.96B estimate (;).
Gemini 4 Carbon announced on the earnings call
Pichai called Carbon Google's strongest frontier model. It is going first to trusted cyber defenders and to US government testing ().
After hours: shares slide about 10%
Investors focused on weaker Search and cash burn rather than Cloud's 103% growth ().
Premarket sell-off drags Nasdaq futures
Alphabet was down more than 10%. D.A. Davidson's Gil Luria said this "may be the quarter we feared." Microsoft rose after Azure grew 48% ().
Capabilities and access
- Gemini 4 Argon (
gemini-4-argon): 1,048,576-token context window. Default output is 65,536 tokens, rising to about 1M with long decode. - Gemini 4 Carbon: access is staged, starting with cyber defenders and US government testers. Broad API availability has not been announced (Google CEO remarks).
- TPU 8t / 8i: Oracle and Crux AI will distribute TPU systems in external data centres.
Read the full section
- Gemini 4 Argon (
gemini-4-argon): 1,048,576-token context window. Default output is 65,536 tokens, rising to about 1M with long decode. Takes text, image, video, audio and PDF input and returns text. Supports thinking, function calling, structured outputs and code execution (Google docs). - Gemini 4 Carbon: access is staged, starting with cyber defenders and US government testers. Broad API availability has not been announced (Google CEO remarks).
- TPU 8t / 8i: Oracle and Crux AI will distribute TPU systems in external data centres. Close to 100 customers already use TPUs (earnings call transcript).
Technical analysis for researchers and developers
- Long output (up to 1M decode tokens) makes Argon suited to whole-repository generation and long agent traces.
- Carbon came about two weeks after Argon. No method has been published (Google CEO remarks).
- No technical report or model card for Carbon appeared in the reviewed sources, so its evaluations cannot be reproduced yet.
Read the full section
- Long output (up to 1M decode tokens) makes Argon suited to whole-repository generation and long agent traces. Expect cost and latency to grow with output length (Google docs).
- Carbon came about two weeks after Argon. Pichai credited AI-assisted training for the faster release pace and expects "recursive self-improvement" (AI helping build the next models) to speed things up over the next 12 months. No method has been published (Google CEO remarks).
- No technical report or model card for Carbon appeared in the reviewed sources, so its evaluations cannot be reproduced yet.
- Google says AI Mode queries average three times the length of traditional searches. That implies more inference compute per query (Google CEO remarks).
Claims and evidence
- Carbon ranks #1 on the Artificial Analysis Intelligence Index, with a score of 59.
- Argon led 12 of 18 public benchmarks against GPT-6 Astra and Claude Opus 5.5. This is Google's own comparison, relayed by the press (gori.me).
- Internal teams shipped 35% more code with 30% fewer errors using Argon. This is an internal metric with no published methodology (Google CEO remarks).
Read the full section
- Carbon ranks #1 on the Artificial Analysis Intelligence Index, with a score of 59. Google cites this third-party leaderboard, but the reviewed sources contain no independent confirmation (Pichai on X).
- Argon led 12 of 18 public benchmarks against GPT-6 Astra and Claude Opus 5.5. This is Google's own comparison, relayed by the press (gori.me).
- Internal teams shipped 35% more code with 30% fewer errors using Argon. This is an internal metric with no published methodology (Google CEO remarks).
- AI Mode slowed search ad growth. Attributed to CFO Anat Ashkenazi (Reuters via Investing.com). Chief business officer Philipp Schindler said ad experiments in AI Mode are "promising" (transcript).
Context and prior work
- Search revenue missed forecasts for the second straight quarter (CNBC).
- Free cash flow was −$5.9B in Q2, which Reuters reports was the first negative quarter in Alphabet's 22 years as a public company.
- The Q3 cash figure also absorbed a $5.2B European Commission fine and a build-up of TPU inventory (transcript).
Read the full section
- Search revenue missed forecasts for the second straight quarter (CNBC).
- Free cash flow was −$5.9B in Q2, which Reuters reports was the first negative quarter in Alphabet's 22 years as a public company. Q3 deepened that to −$9.1B (Reuters via Investing.com).
- The Q3 cash figure also absorbed a $5.2B European Commission fine and a build-up of TPU inventory (transcript).
Limitations, safety and contested findings
- Usage and revenue point in different directions. Google says search query growth is at a five-year high, yet Search revenue missed.
- Release timing is questioned. Bernstein asked whether Google rushed the Gemini 4 models out to offset softer financials (Reuters via Investing.com).
- Cyber capability drives the gated rollout. Google says Carbon is especially strong in cybersecurity, which is why access starts with defenders.
Read the full section
- Usage and revenue point in different directions. Google says search query growth is at a five-year high, yet Search revenue missed. The pressure looks like lower monetisation per query, not lost usage (Google CEO remarks).
- Release timing is questioned. Bernstein asked whether Google rushed the Gemini 4 models out to offset softer financials (Reuters via Investing.com).
- Cyber capability drives the gated rollout. Google says Carbon is especially strong in cybersecurity, which is why access starts with defenders. It also reports gains in alignment and safety but has published no evaluation details (Herald Business).
- The "record" is not yet settled. It depends on how the October 29 session closes.
Business and practitioner implications
- Advertisers: Alphabet is expanding ads in AI Overviews and AI Mode, including internationally.
- Cloud buyers: the Cloud backlog is $561B, and Google expects to recognise just over 55% of it as revenue within 24 months.
- Investors: 2026 capex guidance is $195–205B, with a "significant" increase planned for 2027 (transcript).
Read the full section
- Advertisers: Alphabet is expanding ads in AI Overviews and AI Mode, including internationally. Schindler says sponsored and organic product listings in AI Mode now get similar click-through rates (transcript).
- Cloud buyers: the Cloud backlog is $561B, and Google expects to recognise just over 55% of it as revenue within 24 months. TPU access is spreading beyond Google's own data centres (transcript).
- Investors: 2026 capex guidance is $195–205B, with a "significant" increase planned for 2027 (transcript). Microsoft, by contrast, guided full-year capex down to $170B, which hit neocloud stocks such as CoreWeave and Nebius (CNBC).
The source trail.
Sources (12)
Alphabet drags on Nasdaq futures as the stock faces a record post-earnings loss
Article text retrieved; extracted text may omit tables or interactive elements.
cnbc.comAlphabet earnings miss reflects first real sign that AI is taking a bite out of search
Related coverage; assess separately
cnbc.comGoogle shares tumble 10% after $51bn AI outlay deepens cash burn
Related coverage; assess separately
ft.com