AI Industry Daily Briefing — November 3, 2026

The Executive Read

Yesterday’s edition covered the cloud giants that sell the AI buildout. Today’s documents come from the other side of the counter: the companies that buy it, and what they say they get back. Palantir’s release, dated November 2, reports third-quarter revenue of $2.313 billion, up 96% from a year earlier, with U.S. commercial sales of $984 million ahead of U.S. government sales of $953 million. Amazon’s earnings release shows AWS sales up 39% to $46.0 billion, and also shows what it costs: $77.3 billion of equipment purchases in one quarter and a trailing-twelve-month free cash flow outflow of $42.4 billion. Anthropic says its new agreement with the U.S. General Services Administration cuts government prices by 40%, and Apple says a limited set of Apple Intelligence features reaches China in December with Alibaba’s Qwen model inside. Nearly everything here is the company’s own account of itself, in a filing or a post. We note where a figure is a vendor’s claim, where it is a non-GAAP measure (a figure that leaves out items required in standard accounting), and where we could only reach a secondary source. The two threads to watch are customers paying for AI at scale and the cash that suppliers burn to serve them.

Top AI Headlines

Palantir: Revenue Nearly Doubles, With U.S. Commercial Leading

What happened. Palantir’s third-quarter press release, filed with the Securities and Exchange Commission and dated November 2, reports revenue of $2.313 billion, up 96% from a year earlier and 20% from the second quarter. U.S. commercial revenue was $984 million, up 148% year over year. U.S. government revenue was $953 million, up 96%. Income from operations under standard accounting (GAAP) was $1.085 billion, a 47% margin. On Palantir’s adjusted measure, which is not GAAP, it was $1.427 billion, a 62% margin, and adjusted free cash flow was $1.216 billion. For the fourth quarter, Palantir guides to revenue of $2.649 billion to $2.653 billion. For all of 2026 it guides to revenue of $8.530 billion to $8.534 billion and adjusted free cash flow of $4.8 billion to $5.0 billion. CEO Alex Karp said in the release: “The Sovereign AI movement is imposing the most dramatic dichotomy between haves and have nots.” Sovereign AI means a country building or controlling its own AI systems.

Why it matters. Palantir sells software that connects an organisation’s data to AI models and workflows. Its commercial side now exceeds its government side in the quarter’s figures, which suggests private-sector buyers are paying real money for deployed AI, not only running pilots. Two cautions. Every number here is Palantir’s own, and the guidance is a forecast, not a result. The release also does not say how much of the growth comes from price, new customers or expanded use by existing ones; the excerpt we reviewed did not give a customer count.

Business implication. Palantir is a working example of the enterprise AI market that cloud vendors are building capacity for. Buyers comparing platforms should note the margin: a 62% adjusted operating margin implies that software layered over models can be very profitable, at least for this vendor. This is informational, not investment advice.

Sources: Palantir Technologies, Q3 2026 earnings press release, Exhibit 99.1 to Form 8-K (SEC).

Amazon: AWS Grows 39%, and Free Cash Flow Turns Deeply Negative

What happened. Amazon’s third-quarter release, for the quarter ended September 30, reports net sales of $201.9 billion, up 12% from $180.2 billion. Operating income was $28.1 billion, up from $17.4 billion, a 13.9% margin that Amazon calls its highest ever. AWS, Amazon’s cloud unit, had sales of $46.0 billion, up 39% from $33.0 billion, and operating income of $17.8 billion, up 55%, a 38.7% margin. Purchases of property and equipment in the quarter were $77.3 billion, against $35.1 billion a year earlier. Over the trailing twelve months, free cash flow, meaning operating cash flow minus equipment purchases, was an outflow of $42.4 billion, compared with an inflow of $14.8 billion a year earlier. Amazon attributes the swing primarily to AI investment. It expects about $220 billion of cash capital expenditures in 2026. CEO Andy Jassy said AWS expects revenue growth “over 40% both in 2027 and in 2028.” Net income was $20.7 billion against $21.2 billion, a comparison Amazon says is distorted because last year’s figure included $9.5 billion in pre-tax gains on its Anthropic investments.

Why it matters. Amazon’s equipment purchases in one quarter exceeded Alphabet’s $51.4 billion, reported in our last edition. The 2027 and 2028 growth projections are management’s forecasts, not results. The release does not report the size of AWS’s contracted backlog. Some outlets have printed a backlog figure; we could not find it in the release and do not use it.

Business implication. Customers should read two things. AWS is growing faster while spending more than twice what it did a year ago on equipment, so the supply side is still being built. And the free cash flow figure shows the buildout is outrunning the cash the business generates, which lenders and suppliers will watch. We did not verify post-earnings share moves reported elsewhere and do not report them. This is informational, not investment advice.

Sources: Amazon.com, Inc., Third Quarter 2026 results (Amazon investor relations).

Anthropic and the GSA: 40% Off for Government, by Anthropic’s Account

What happened. In an October 27 post, Anthropic announced an expanded agreement with the U.S. General Services Administration, the federal government’s purchasing agency. Under it, government customers get a 40% discount on Claude Enterprise, Claude for Government and the Claude API, with no minimum spending commitment. The discount also extends to state, local, tribal and territorial governments, and, Anthropic says, will reach purchases through cloud providers “in the coming months.” Anthropic estimates the discounts will save the government $1 billion. It says about 500 U.S. federal, state and local government organisations use Claude, weekly active users of Claude for Government are up roughly eightfold since February, and use across federal civilian agencies has grown roughly twentyfold in a year. Claude for Government became generally available on September 30 and, per the post, holds FedRAMP High authorization, the federal security standard for sensitive unclassified data. Agencies can buy through GSA’s Multiple Award Schedule.

Why it matters. The usage figures are Anthropic’s, and the post does not give the baselines behind “eightfold” or “twentyfold.” The $1 billion savings is an estimate that depends on how much agencies buy. The case studies are also Anthropic’s: it describes a Utah pilot in which 71% of 34 developers reported saving 20 or more hours in four weeks, and a state IT team that it says migrated nearly 900 API methods in two days. These are self-reported by Anthropic and the agencies it quotes, and we have not seen independent audits.

Business implication. A published government price floor tends to shape what other large buyers ask for. Rival model vendors that sell through the same schedule will face the 40% reference point. Anthropic filed to list its shares on October 26; we have not drawn figures from that filing. This is informational, not investment advice.

Sources: Anthropic, “Expanding our partnership with GSA to bring Claude to all levels of government” (October 27); Anthropic, S-1 announcement (October 26).

Apple Says a Limited Apple Intelligence Reaches China in December, With Alibaba’s Qwen

What happened. Apple’s China newsroom announced that Apple Intelligence features will be available for testing in China from November 4 in the latest iOS developer beta, and will launch to Chinese users with the public release of iOS 27.3 in December. The release says Apple partnered with Alibaba to bring Qwen, Alibaba’s AI model, into Siri, Writing Tools and Visual Intelligence, including text and image understanding. Listed features include Live Translation, Writing Tools, Visual Intelligence, Clean Up in Photos and priority mail and notification summaries. Apple says the update does not include Siri AI or the other next-generation features it announced at WWDC 2026, its developer conference. Support on other platforms, such as the Mac and iPad, depends on “necessary approvals.” Per the release’s footnote, the features need an iPhone 16 or later, an iPhone 15 Pro or Pro Max, or an iPhone Air, with Siri and device language set to English and Simplified Chinese. MacRumors, citing Apple’s announcement, reports that Image Playground is also not included.

Why it matters. Earlier reporting by Bloomberg, which we have not read directly, said Apple received initial clearance in July. What Apple’s own release shows is the shape of the compromise: older features, a local partner’s model, and no new Siri. We have no benchmark comparing Qwen with the models Apple uses elsewhere.

Business implication. For developers and enterprises in China, iPhone-based AI features will run on a domestic model. For Western model vendors, it is another sign that the Chinese market is being served through local partners. The English-and-Chinese language requirement and the device list limit the first wave of users.

Sources: Apple China Newsroom, “Apple Intelligence features are coming to users in China” (October 2026) (Chinese-language; our reading of it is a machine summary); MacRumors, “Apple Says iOS 27.3 Beta Coming Next Week” (October 30).

Model and Product Updates

OpenAI and ServiceNow widen their partnership. OpenAI’s announcement, dated October 28, says the two companies will bring more OpenAI models and agents into ServiceNow products and connect ServiceNow workflows to ChatGPT and Codex, OpenAI’s coding agent. It builds on a January agreement. A ServiceNow plugin for ChatGPT is described as available now through the OpenAI marketplace, and OpenAI agents will reach ServiceNow workflows through a layer called Action Fabric, which, OpenAI says, applies existing business rules and permissions. The companies also say they are building a joint security product. Most planned features are slated for 2027, so this is a statement of intent more than a launch. We could not open OpenAI’s page ourselves, and the details come from a search summary of it and ServiceNow’s press release; read both before relying on specifics.

Anthropic widens its free-credit program for scientists. In an October 28 post, Anthropic says its AI for Science program, which gives free Claude credits to researchers, has supported hundreds of projects since May 2025. The expansion names 20 partners, among them the Alfred P. Sloan Foundation, the Simons Foundation, Schmidt Sciences, the U.S. National Science Foundation, the Wellcome Sanger Institute and the Francis Crick Institute. Anthropic also says it recently launched a discounted Team plan for scientists. The post is a program announcement and carries no results about scientific output.

Microsoft details North Korean groups’ use of AI. Microsoft Threat Intelligence’s October 29 post rates AI adoption across North Korean threat actors as high, but says AI-enabled operations “remain human-directed and have not become fully autonomous.” It places three groups, Jade Sleet, Coral Sleet and Jasper Sleet, at the most advanced stage, which it calls “operationalized.” It says two others, Onyx Sleet and Pearl Sleet, have used ARTEX, an open-source framework in which AI agents run penetration tests (authorised break-in simulations), in operations since late October 2026. Specific findings include Onyx Sleet creating more than 100 PDF lure files in August and September and Diamond Sleet returning to oil and gas targets for the first time since 2023. These are Microsoft’s assessments of groups it tracks, and the post is the company’s own report.

Sources: OpenAI, “OpenAI and ServiceNow expand partnership across enterprise workflows”; ServiceNow press release; Anthropic, “AI for Science expands to foundations and research institutions” (October 28); Microsoft Security Blog, “Tracking AI adoption in North Korean threat actor tradecraft” (October 29).

Regulation and Policy Watch

UK regulators say existing law already covers AI agents. The LexisNexis legal news service reports, in an item dated October 30, that the UK Competition and Markets Authority updated its AI agents guidance to include a joint statement from the Digital Regulation Cooperation Forum, the group made up of the CMA, the Financial Conduct Authority, the Information Commissioner’s Office and Ofcom. Per that report, the statement says existing legal frameworks apply to personal AI agents, that an agent’s autonomy does not reduce a business’s responsibility, and that risks include fraud, inappropriate sharing of personal data and misleading recommendations. Firms are told to explain how agents act for users, how authority is granted, how personal data is used and whether commercial interests shape recommendations. Regulators say they will use enforcement powers “where appropriate.” We have not read the statement itself; LexisNexis does not link it, and we could not find the gov.uk page. Treat the details as secondary until we can.

China and Apple. The Apple release above is also a regulatory story. Apple’s own text says other platforms depend on “necessary approvals,” which is the only regulatory statement we can attribute to Apple. Reports on which Chinese permits were involved come from secondary sources we could not verify and are left out.

What we did not find. We searched for new action from the European Commission, U.S. federal agencies and China’s regulators dated since October 30 and found nothing we could source to a primary document. Earlier items stay on the calendar in the Watchlist below, including the November 15 close of China’s comment period on its draft export-control catalogue, which we last covered on November 2.

Sources: LexisNexis UK, “CMA updates AI agents guidance with DRCF joint statement” (October 30), secondary; Apple China Newsroom.

Emerging Startup Radar

Flapping Airplanes: $1 billion at a reported $6 billion valuation. TechCrunch reported on October 21 that the year-old AI lab raised a $1 billion Series A co-led by Index and Lux, with earlier backers Sequoia, GV and Kleiner Perkins and others, including Nvidia, also investing. TechCrunch says the $6 billion valuation was first reported by Bloomberg; it is not a company-confirmed figure. The company builds models designed to need far less training data, and describes a “megakernel” software approach that keeps a graphics chip busier by running training and inference in a single program. TechCrunch notes the valuation is four times that of a roughly $180 million seed round earlier this year, which valued it at $1.5 billion. It is a research bet with no product or revenue figures in the report.

Legora: $550 million at $8.8 billion. TechCrunch reported on October 21 that the legal AI company’s Series E was led by Atreides Management, with Baron Capital, Bessemer, General Catalyst and Menlo participating. TechCrunch says annual recurring revenue now exceeds $200 million, up from $100 million in April. Those revenue figures are the company’s, as relayed by TechCrunch. The report says Legora’s main rival, Harvey, raised $550 million at a $15.5 billion valuation about a month earlier. Legal is one of the first professional fields where startups are showing revenue at this scale, but the revenue claims are not audited.

FuriosaAI. Our November 2 edition reported the South Korean chip designer’s roughly $620 million Series D from CRN Asia. Wowtale, a trade outlet covering Asian startups, also gives the round as about $623 million at a post-money valuation near $2.89 billion, dated October 29. That is a second trade source, not a company release we read.

Sources: TechCrunch, Flapping Airplanes (October 21); TechCrunch, Legora (October 21); Wowtale (October 29). Trade press only; no company releases read.

AI Infrastructure and Market Signals

The spending ledger, from the filings. Putting the week’s primary documents side by side: Alphabet’s equipment purchases were $51.4 billion in the quarter and its free cash flow was negative $9.1 billion. Amazon’s purchases were $77.3 billion in the quarter, and its trailing-twelve-month free cash flow was negative $42.4 billion. The two companies define the periods differently, one a single quarter and the other a trailing year, so the free cash flow numbers should not be compared directly. Amazon expects about $220 billion in 2026 cash capital spending and, per its own statements, will spend more in 2027. Amazon’s AWS and Alphabet’s Cloud segment grew 39% and 103% respectively. The pattern across both is the same: revenue is growing at the pace of the spending, but cash is going out faster than it comes in.

Palantir’s sovereign-AI line. Karp’s quote in Palantir’s release ties its results to national governments buying AI they control. We did not find a figure in the release breaking out non-U.S. government sales, so the claim remains a statement of strategy, not a measured trend.

Enterprise restrictions. Bloomberg’s headline on November 2 says Goldman Sachs “restricts traders and bankers from using some AI tools.” The article is behind a paywall and returned an error to our tools, so we cannot report its contents. A search summary of it said the bank limits some advanced coding tools to engineers; we treat that as unconfirmed and link the report as Bloomberg’s own.

Market backdrop. We could not find a primary-source reading of today’s or Monday’s market session and report none.

Sources: Amazon investor relations, Q3 2026 results; Alphabet, Q3 2026 exhibit 99.1 (SEC); Palantir, Q3 2026 Exhibit 99.1 (SEC); Bloomberg, Goldman article (paywalled; reporting is Bloomberg’s).

Public Investment Watchlist

Informational only; this is not investment advice.

• Palantir (PLTR). Q3 revenue of $2.313 billion, up 96%; full-year revenue guidance of $8.530 billion to $8.534 billion; adjusted free cash flow guidance of $4.8 billion to $5.0 billion. All from the company’s release.
• Amazon (AMZN). AWS sales of $46.0 billion, up 39%; 2026 cash capital expenditure expected at about $220 billion; trailing free cash flow negative $42.4 billion. Fourth-quarter guidance is net sales of $240.0 billion to $247.0 billion and operating income of $28.5 billion to $33.5 billion.
• Alphabet (GOOGL) and Microsoft (MSFT). Covered in yesterday’s edition; no new filings since.
• Anthropic (proposed Nasdaq listing, ticker ANT). Anthropic’s own post says it filed a Form S-1 on October 26 for a proposed IPO on the Nasdaq Global Select Market, that the registration statement “has not yet become effective,” and that the share count and price range are not yet set. Goldman Sachs and Morgan Stanley are named as sources of the preliminary prospectus. We report no financial figures from the filing.
• Apple (AAPL). Announced the China rollout of limited Apple Intelligence features, as above. No financial disclosure was attached.

Watchlist

• November 4: Apple’s first iOS developer beta with Apple Intelligence features for China is due, per Apple’s release.
• November 5 to 6: The World AI Regulation Summit takes place in London, according to its organiser’s listing; we have not verified the agenda.
• November 15: China’s comment period on its draft export-control catalogue closes.
• December: Public release of iOS 27.3 with the China features, per Apple. The EU’s ban on AI-generated non-consensual sexual imagery is scheduled to apply, per the Commission’s page cited November 2.
• Anthropic S-1 amendments: Price range and share count have not been published.
• OpenAI: We could not open its newsroom today. We have not verified reports of new model releases and carry none.

---

A note on today’s reporting: OpenAI’s newsroom, Bloomberg and The Neuron’s digest returned errors to our tools, so some items rest on search summaries, which we flag in the text. Nothing from earlier editions is withdrawn today.