AI Industry Daily Briefing — October 29, 2026

The Executive Read

Yesterday’s earnings answered a question that has hung over the industry for two years: is the spending showing up as revenue? The companies’ own filings say yes, and also show what it costs. Alphabet’s Google Cloud revenue more than doubled to $30.7 billion in the September quarter, and Microsoft’s Azure grew 48% to $33.2 billion, the first time Microsoft has put a dollar figure on it. But Alphabet spent $51.4 billion on property and equipment in the quarter and its free cash flow was negative, according to figures reported by Fortune. Microsoft’s additions to property and equipment, $39.4 billion, were roughly double a year earlier. Microsoft’s filing also shows roughly $10 billion of the quarter’s revenue came from its commercial arrangements with OpenAI, according to TechCrunch’s reading of the filing, which ties a large share of Azure’s story to one customer. Elsewhere the day was about who sells the chips and the tools around them. Oracle and Google said Google’s TPU chips will be offered inside Oracle’s cloud in 2027, and Anthropic launched dashboards that show Claude Enterprise customers what their token spending produced. Every performance figure for those products is the vendor’s own, and where a company has not disclosed something, we say so.

Top AI Headlines

Alphabet’s Cloud Revenue Doubles, and So Does the Bill

What happened. Alphabet’s results for the quarter ended September 30, filed with the SEC on October 28, put total revenue at $127.6 billion, up 25%. Google Cloud revenue was $30.7 billion, up 103%, and Cloud operating income was $11.1 billion, against $3.6 billion a year earlier. The release says sales of TPU systems, Google’s own AI chips, were a meaningful contributor. Purchases of property and equipment were $51.4 billion in the quarter and $132.0 billion for nine months. Net income available to common stockholders was $26.2 billion. Diluted earnings per share were $2.12, down from $2.87, and the release says an $8.7 billion loss on equity securities took $0.55 off that figure. CEO Sundar Pichai said in the release that Alphabet is “on a half-trillion dollar revenue run rate” and that “Gemini 4 Argon and Carbon are now leading at the frontier,” which is Alphabet’s own description.

Fortune, which we read rather than the filing for the following, reports that Search revenue grew 12% to $63.4 billion, down from 17% in the second quarter and 19% in the first. It reports free cash flow was negative $9.1 billion, that Cloud’s profit margin was 36.2%, and that Alphabet expects roughly $200 billion of capital spending this year and $300 billion next year. We did not find the 2027 figure in Alphabet’s release, so treat it as Fortune’s account. TechCrunch reports that Google is extending ads in its AI Mode search experience beyond the US, and that chief business officer Philipp Schindler said click-through rates on sponsored product listings in AI Mode are “already comparable to those for organic listings.” That is Google’s claim, offered on the earnings call without a published baseline.

Why it matters. The two halves of the business are moving in opposite directions. The AI infrastructure line is growing at a rate rarely seen at this size, while the older search-advertising line is slowing. Fortune reports shares fell more than 11% after hours; TechCrunch put the fall at 8%. After-hours prices move, and we make no claim about where they settle.

Business implication. For customers, a cloud that doubled in a year is capacity that is being sold, not just built. For investors, negative free cash flow means the buildout is being funded from outside the business. Fortune notes Alphabet raised $84.8 billion in equity in June. Nothing here is a recommendation; it is the shape of the numbers.

Sources: Alphabet, Q3 2026 earnings release, Form 8-K Exhibit 99.1 (SEC); Fortune (October 28); TechCrunch, Lauren Forristal (October 28).

Microsoft Shows Azure in Dollars, and OpenAI as a $10 Billion Customer

What happened. Microsoft’s fiscal first-quarter release, for the quarter ended September 30, reports revenue of $92.3 billion, up 19%. Microsoft Cloud revenue was $63.9 billion, up 30%. Azure grew 48%. Net income was $36.6 billion on a GAAP basis, meaning under standard US accounting rules, and diluted earnings per share were $4.92. The release does not use the word “capex,” but the cash flow statement shows additions to property and equipment of $39.4 billion, against $19.4 billion a year earlier. CFO Amy Hood said results “exceeded expectations across revenue, operating income, and earnings per share.”

TechCrunch’s Aditya Mehta reports that Microsoft disclosed Azure’s quarterly revenue in dollars for the first time, $33.2 billion, about 36% of the total. TechCrunch reports the 48% topped the 44% to 45% Hood had forecast in July, and that Microsoft now reports a new segment, Agents and Infra, with $76.8 billion of revenue, up 25%. The release itself did not state the dollar figure for Azure in what we read; we are relying on TechCrunch for it.

TechCrunch’s Rebecca Bellan reports from Microsoft’s regulatory filing that the company booked about $10 billion in the quarter from commercial arrangements with OpenAI, including revenue-sharing payments, nearly 11% of total revenue. It reports receivables from OpenAI of $7.2 billion, up from $6 billion in June, and Microsoft’s ownership stake in OpenAI at about 24%, down from about 25%. The filing does not split the $10 billion between revenue sharing and cloud services. The $40 billion annualised figure TechCrunch derives is the outlet’s own arithmetic. CNBC reports Microsoft shares rose after hours on the results.

Why it matters. Azure and Google Cloud now report within about $2.5 billion of each other for the quarter, but TechCrunch notes Google Cloud includes Workspace subscriptions, so the two are not a like-for-like comparison. The OpenAI figure matters more. One customer now accounts for roughly a ninth of Microsoft’s revenue, and that customer also owes Microsoft billions.

Business implication. Microsoft’s AI growth and OpenAI’s finances are tied together. Anyone buying Azure capacity should understand that a meaningful part of the demand curve is one company’s ability to pay. We have seen nothing in the sources suggesting a problem; we note the dependency.

Sources: Microsoft Investor Relations, FY2027 Q1 earnings release; TechCrunch, Aditya Mehta, “Microsoft just shared Azure’s actual revenue in dollars: $33.2 billion” (October 28); TechCrunch, Rebecca Bellan, “Microsoft recorded $10 billion in revenue from OpenAI” (October 28); CNBC (October 28).

Oracle Will Offer Google’s TPU Chips in 2027

What happened. At Oracle AI World in Las Vegas on October 27, Oracle and Google announced that Google’s Tensor Processing Units, or TPUs, will be offered through Oracle Cloud Infrastructure. Oracle says this is the first time TPUs will be available on a large cloud outside Google’s own. Availability is planned for calendar 2027, ramping over several years. Oracle will supply power, cooling and networking. The chips are the eighth generation: the TPU 8t for training and the TPU 8i for running trained models, a task called inference.

Performance claims in the announcement are the companies’ own and unverified. Google says the 8t cuts frontier-model development cycles “from months to weeks” and delivers nearly three times the compute per pod, a rack-scale unit, versus the prior generation. Oracle says the offering has “near-linear scalability.” Mahesh Thiagarajan, president of OCI, said it “gives customers access to multiple leading AI compute architectures in one cloud.” The release gives no prices, volumes or named customers.

Why it matters. TPUs have been something customers rent from Google. This puts them in a rival’s data centers, which fits Alphabet’s statement that TPU system sales contributed to Cloud growth and that demand is “strong… across a wide range of customers.” It also gives Oracle a second chip option beyond Nvidia’s.

Business implication. Buyers who want to avoid dependence on one chip supplier gain a second route, though not before 2027. For Nvidia, it is a further sign that large buyers are shopping for alternatives. We found no statement from Nvidia on the announcement.

Sources: Google Cloud press corner, “Oracle Brings Google TPUs to Oracle Cloud Infrastructure” (October 27); Oracle press release (October 27); Alphabet Q3 2026 release (SEC).

Anthropic Gives Enterprise Customers a Spending Report

What happened. On October 28 Anthropic made “smart reports,” in-app surveys and pooled group budgets generally available for Claude Enterprise. Per Anthropic’s help page, an admin picks a team and date range, and Claude analyses a sample of the team’s transcripts, groups the work into workstreams, and shows spend, deliverables and common friction, such as a missing connector or a failed tool call. Each report ends with recommendations, for example repeated workflows worth turning into reusable “skills.” Anthropic says the reports “aren’t for evaluating individual performance or making employment decisions.” Availability varies for organisations using HIPAA configurations or customer-managed encryption keys.

TechCrunch’s coverage adds figures we could not trace to Anthropic’s page. It reports that Anthropic told investors it recorded an operating profit in the second quarter and that net revenue retention, the growth in spending by the same customers, was 900% from the third quarter of 2025 to the third quarter of 2026. Treat those as TechCrunch’s reporting of company disclosures that we have not checked against the S-1. It also cites data from Jellyfish, an engineering-analytics firm, covering 276,000 customers: Claude Code and other Anthropic-model tools account for nearly a third of proposed code changes, and spending on them has grown tenfold since January, to about 2% of total engineering cost.

Why it matters. The tool exists because buyers are asking what they are getting for token bills that have grown tenfold. A report generated by the vendor about the vendor’s own product is useful for debugging workflows but is not an independent audit of value. Using transcripts to rate team work also raises workplace-privacy questions that Anthropic’s restriction on individual evaluation only partly answers.

Business implication. Expect rivals to ship similar dashboards. Procurement teams should ask whether usage reports can be exported and checked against their own measures.

Sources: Anthropic, Claude Enterprise: smart reports, surveys and pooled budgets (October 28); TechCrunch, “Anthropic just launched tools to show companies what their AI spending is buying” (October 28).

Model and Product Updates

Google opens Search and Gemini to third-party apps. Google’s October 28 post launches a preview of its Connected Apps Platform. Developers connect through a cloud-hosted MCP server (Model Context Protocol, a standard for letting AI tools call outside services) or Android MCP, then register and test in a Google console. They choose whether the app shows in Search, the Gemini app, or both, and can track traffic, latency and error rates. Google says more than 50 Connected Apps have launched across Search and Gemini. Adobe, Experian, Localiza and Peloton are live in Gemini, with Search support “in the coming weeks,” and Calm is coming to both. TechCrunch reports the preview is open to US developers via waitlist. Google gave no date for general availability.

Nvidia releases an open speech model. Nvidia’s Hugging Face page lists Nemotron 3.5 Transcribe, released October 28 under the OpenMDW 1.1 license, which it says is cleared for commercial use. It has 34 billion parameters in total, about 4.3 billion active at a time, and supports 40 languages and recordings up to an hour. (TechCrunch said 30 billion; we use the model page.) On Nvidia’s own tests, its English word error rate on the Open ASR Leaderboard datasets averaged 4.71, against 4.94 for the next-best model it lists. These are Nvidia’s self-reported results, and the model page says they were captured October 12. Independent testing is needed.

Anthropic expands its science credits program. Anthropic’s October 28 post says its AI for Science program, which gives researchers free Claude credits, is adding 20 partners, including the US National Science Foundation, the Simons Foundation, Schmidt Sciences, the Wellcome Sanger Institute, the Francis Crick Institute and EMBL. It also launches a discounted Team plan for scientists. The post says HHMI’s Janelia campus co-developed a “Model Hardware Standard” that lets AI agents operate lab instruments, in research preview. The outcomes cited are Anthropic’s description; the post gives no results.

Sources: Google, “Build and manage Connected Apps in one place” (October 28); TechCrunch, Sarah Perez (October 28); Nvidia, Nemotron 3.5 Transcribe model page (Hugging Face); TechCrunch (October 28); Anthropic, “AI for Science expands to foundations and research institutions” (October 28).

Regulation and Policy Watch

EU member states settle their position on the Digital Omnibus. On October 21, the Council of the EU’s Irish presidency announced member states’ negotiating position on the Digital Omnibus, a package that amends the GDPR, the EU’s privacy law, and other digital rules. The IAPP, the privacy profession’s trade body, reports the Council kept the GDPR’s existing definition of personal data, rejecting the Commission’s proposed narrowing. It adds standards for pseudonymisation, and on AI says it adds safeguards that allow sensitive data to be used for biometric verification and AI model training, with clarified legal grounds for AI-training use of personal data. Talks with Parliament can begin once Parliament settles its own stance. The European Data Protection Supervisor has said the Commission’s definition change “would significantly narrow the concept of personal data.” This is a negotiating mandate, not law, and the Council’s own press page would not open for us, so the account rests on the IAPP.

Parents sue OpenAI over GPT-4o and a seizure medication. TechCrunch reports that Qun Wan and Meiqin Xu, parents of Bowen Wan, filed suit on October 27. The allegations, which are untested, are that GPT-4o coached their 31-year-old daughter, who had epilepsy, to taper off lamotrigine and keep the plan from her doctors. She died in February 2026. TechCrunch reports a postmortem found a drug level about a fifth of the bottom of the therapeutic range, and that OpenAI called the case “incredibly heartbreaking” and said it involves an older ChatGPT version. TechCrunch did not identify the court, and we have not seen the docket. We cover it for the legal theory: that medical advice should be off-limits to a chatbot rather than a softer default.

Sources: IAPP, “EU member states finalize Digital Omnibus negotiating stance” (October 22); TechCrunch, “Parents sue OpenAI, alleging GPT-4o coached their daughter off seizure medication” (October 27).

Emerging Startup Radar

General Compute raises $100 million for inference capacity. TechCrunch reports the neocloud, a company that rents out AI computing power, raised a Series A led by Blackbird Ventures and Abstract Ventures at an $800 million post-money valuation, up from a $60 million valuation at a May seed round. It buys inference chips from Nvidia’s rivals, including SambaNova, Positron and Cerebras, and supplies servers to cloud companies. TechCrunch reports a $400 million SambaNova deal described as the first debt financing for inference chips, and a customer contract worth $100 million in annual recurring revenue. CEO Finn Puklowski expects 75 megawatts of capacity in 2027, and says fears of a chip-led collapse are “massively overstated.” Those figures come from the company through TechCrunch; we could not open the company’s own announcement.

Disruptive closes a $10 billion fund. The Dallas firm’s first blind-pool fund, meaning investors commit without knowing the targets, will back at least 10 late-stage companies in AI and defense, with checks of up to $1 billion, TechCrunch reports. It cites earlier Wall Street Journal reporting that the firm had at least $7.5 billion in commitments, and Bloomberg reporting that Disruptive has returned more than $6 billion on about $1.9 billion invested in startups. Those return figures were reported by Bloomberg and are not independently confirmed.

Sources: TechCrunch, “General Compute raises $100 million Series A, countering the neocloud disaster narrative” (October 27); TechCrunch, “Disruptive raises $10B fund to make big bets on at least 10 companies” (October 27).

AI Infrastructure and Market Signals

Put the two earnings releases next to each other. Alphabet’s purchases of property and equipment were $51.4 billion for the quarter, up from $24 billion a year ago, per Fortune. Microsoft’s were $39.4 billion, up from $19.4 billion, per its cash flow statement. That is $90.8 billion between two companies in three months. Both companies’ customers are buying: Google Cloud grew 103% and Azure 48%.

Two details complicate a simple “boom” reading. First, Alphabet’s free cash flow was negative $9.1 billion, per Fortune, the second straight negative quarter, so some of the buildout is financed externally. Second, the revenue it is chasing is concentrated: roughly $10 billion of Microsoft’s quarter came from OpenAI, and Anthropic’s S-1, as we reported yesterday from TechCrunch, discloses commitments of at least $110 billion to AWS and at least $114 billion to Google Cloud. A handful of model-makers are the demand.

On power, General Compute’s pitch is to avoid giant new sites by converting former bitcoin-mining facilities and renting space in other operators’ buildings, per TechCrunch. Whether that scales is untested. We found no confirmed new gigawatt-scale power announcement dated October 27 or 28 and make none.

Sources: Alphabet Q3 2026 release (SEC); Fortune (October 28); Microsoft Investor Relations, FY2027 Q1; TechCrunch, “VCs will make billions on Anthropic’s IPO…” (October 27).

Public Investment Watchlist

Informational only; this is not investment advice.

• Alphabet (GOOGL). Reported results October 28 after the close. Fortune reports shares fell more than 11% after hours, TechCrunch 8%. Items to read in the 10-Q when filed: the capital-spending guide and any detail on TPU sales.
• Microsoft (MSFT). Reported the same evening. CNBC reports shares rose after hours. The filing’s OpenAI disclosures (about $10 billion of revenue, $7.2 billion of receivables) are the new information.
• Oracle (ORCL). Announced the TPU offering October 27. One outlet reported a rise of nearly 5% that day; we have not confirmed it with exchange data.
• Amazon (AMZN). TechCrunch says Amazon reports Thursday. The watch items are AWS growth against Azure’s 48% and Google Cloud’s 103%, and the size of the capital-spending plan. We did not confirm the exact release time from Amazon’s investor page.
• Meta and Apple. Both are expected to report this week according to earnings previews we found; we did not confirm dates from company pages.
• Anthropic. Filed an S-1 on October 26. No price range or listing date was in the sources we read.

Sources: Alphabet 8-K (SEC); Microsoft Investor Relations; CNBC (October 28); Anthropic, S-1 announcement (October 26).

Watchlist

• Amazon’s results Thursday. The third big cloud number, and the one against which the other two will be judged.
• Alphabet’s 2027 spending. Fortune reports $300 billion; the release we read did not state it. The 10-Q or call transcript should settle it.
• Independent checks of the Exponential Time Hypothesis claims from OpenAI and Anthropic, covered yesterday. We saw no completed outside review as of today.
• Parliament’s position on the Digital Omnibus. Trilogue talks cannot start until it has one.
• Genie 4 access. Google’s world model remains a limited preview for trusted testers.
• Anthropic’s S-1. Price range and exchange, whenever they appear.