Anthropic Hits $17B a Quarter but Slips Back to a Loss (Nov 2, 2026)

🎧 Listen ad-free:


🌐 Daily Briefing: https://djamgamind.com/daily
📚 Research Hub: https://djamgamind.com/pdfs

🔍 Keywords: Anthropic, AI Compute Costs, AI Pricing, Usage Limits, Midterm Elections, Data Centers, Cybersecurity, OT Security, ChatGPT Ads, Network Segmentation

Summary:
Anthropic’s amended IPO filing shows $17 billion in third-quarter revenue, but growth slowed to about 47% and the company slipped back to an operating loss as it paid up for short-term computing power. It now lists about $895 billion in compute commitments. Tomorrow’s U.S. midterms double as a referendum on data centers, and OpenAI plans to rebrand them as “AI compute campuses.” And hackers heated supermarket freezers to 50°C across at least seven European countries, without asking for a ransom.

Important Topics:

  • Anthropic’s Q3: a $17B quarter, a $567M operating loss and $895B in commitments.

  • Rationing: new usage limits at Anthropic and OpenAI; Grok costs too much.

  • AI at the polls: data-center votes and OpenAI’s rebrand.

  • The back door: freezer attacks, Binance and deepfake apps.

  • Ads and deals: ChatGPT ads at $1B; Marvell and Lumentum.

  • Cert Corner: protecting devices too old to patch.

📈 Anthropic’s third quarter: growth slows, costs jump

Fortune · TBPN · The Rundown AI · Techpresso

What happened: Anthropic’s amended S-1 adds its third quarter, ended Sept 30.

The numbers:

  • Revenue: $17 billion, up from $11.6 billion in Q2 and $1.4 billion a year earlier (Fortune). About $68 billion annualized (TBPN).

  • Growth is slowing: 132% in Q1, 144% in Q2, 47% in Q3 (TBPN).

  • Back to an operating loss: $567 million, after a $974 million operating profit in Q2. Short-term compute “currently costs significantly more” than its multi-year deals, and a $2 billion non-cash charge for an employee donation-matching program added to costs. Adjusted operating margin fell to 15% from 19% (Fortune).

  • The price war shows up: “a decrease of approximately $600 million from price reductions” between Q2 and Q3 (Fortune).

  • The commitments: about $895.4 billion in purchase commitments and leases, almost all compute, much of it for “Anthropic Direct,” its own dedicated infrastructure (Fortune; TBPN).

  • The loss line: a $23.3 billion net loss, mostly an accounting charge on convertible notes. Liquidity: $146.6 billion (Fortune).

Everyone is rationing compute:

  • Anthropic renamed Max 5x to “Max” and Max 20x to “Ultra” at the same prices, and removed the 5-hour limit for verified Ultra accounts (The Rundown AI).

  • OpenAI scrapped a planned cut, and from December moves Plus, Pro and Business to one token-based allowance for ChatGPT and Codex (Techpresso).

  • Elon Musk said Grok Bot costs too much to run for what it earns, and won’t raise its limits; it has about 800,000 paying users (Techpresso).

The Number: $895.4 billion. Anthropic’s compute commitments and leases, against revenue of about $68 billion a year. These run over many years, but the market is now watching every quarter.

Why it matters: Discounts on one side, expensive short-term compute on the other: that squeeze is coming to every AI vendor. Expect more plan changes, token meters and caps. Read your contract’s usage terms, not just the price per token.

🤝 PARTNER SLOT
This space is reserved for one enterprise AI partner each day: AI cost governance, model routing and OT security for the leaders who read this briefing. Sponsorship options and media kit →

🗳️ AI goes to the polls

The Rundown AI · Axios AI+ · The Deep View

  • Tomorrow’s U.S. midterms double as a referendum on data centers, with roughly 20 communities voting on bans or pauses (The Rundown AI).

  • Democrats are projected to flip the House (The Rundown AI); Axios says they “will likely win.” Some candidates want a federal freeze on new data centers.

  • OpenAI will launch a policy and PR blitz after the vote, rebranding data centers as “AI compute campuses,” starting with the lame-duck session on Nov 9 (Axios AI+). Chris Lehane calls Tuesday an “electoral portal from one era to the next.”

  • AI-oriented PACs spent nearly $80 million on this year’s elections (The Deep View, headline).

  • Wharton’s Ethan Mollick says the doom debate crowds out nearer problems: “There’s actually 1,000 other things that I’m really worried about that are already going to happen.” (Axios AI+)

Why it matters: Local fights over data centers feed straight into what AI costs. Slower permits and higher local charges for new sites mean more expensive capacity. Watch Wednesday’s results district by district.

🧊 The back door: freezers, exchanges and app stores

Techpresso · TBPN · Al Jazeera (October)

  • Freezers at 50°C: hackers raised refrigeration temperatures at dozens of supermarkets in at least seven European countries overnight. About 40 Delhaize shops were hit, and one Belgian store threw out €100,000 of food (Techpresso). Nobody has demanded a ransom, and Belgium’s Centre for Cybersecurity says no one knows who is behind it (TBPN).

  • The pattern: last month’s attack on about 400 French supermarkets hit refrigeration systems the vendor said were too old to patch (Techpresso).

  • Binance: the FBI blames a North Korean group for the $6.3 billion theft, and users pulled $5 billion as the exchange restarted (FT headlines via TBPN).

  • Deepfake apps: Apple pulled 19 apps hiding AI deepfake nudity tools behind innocent features, after a WIRED investigation; researchers linked 400+ apps and domains to the operation (Techpresso).

  • AI vs. scanners: in one study, Snyk caught 82% of flaws and Claude 79%, but together they found 101 of 108 (Techpresso).

The second number: 50°C. The temperature hackers set Europe’s supermarket freezers to, remotely, with no ransom demand.

Why it matters: Anything on the internet that can’t be patched must be isolated: its own network segment, no direct internet path, remote access only through a logged gateway with MFA.

📢 Ads, products and deals

The Deep View · Techpresso · The Next Web

  • ChatGPT ads reached a $1 billion annualized run rate in under 200 days, with 100,000+ advertisers. A second model reads each conversation for commercial intent before placing a “sponsored” ad; OpenAI says user trust comes first (The Deep View).

  • Claude Code will no longer recreate specific existing games, even renamed; original games are fine (Techpresso).

  • Meta built a prototype chore robot, codenamed Max, per The Information (Techpresso).

  • Grok’s API text gets invisible watermarks from Nov 7, citing the EU AI Act (The Rundown AI).

  • Deals: Marvell to buy Lumentum in a $115 billion all-stock deal (Techpresso headline); Oracle kicks off a $20 billion raise for AI (Bloomberg headline via TBPN); DeepSeek raised nearly $16 billion (The Deep View headline).

🎓 Cert Corner

Question: A retailer’s refrigeration controllers are too old to patch, and they’re reachable from the internet so a vendor can manage them. What’s the best protection? A) A long, complex password B) An isolated network segment with no direct internet access, and remote access only through a VPN or jump host with MFA and logging C) Wait for a vendor patch D) Turn off logging to reduce attack surface

Answer: B. When a device can’t be patched, isolation is the compensating control. A strong password still leaves it exposed, waiting leaves it open, and without logs you won’t see the attack. Practice thousands more questions in DjamgaTech.

Strategic C-Suite Takeaways

  1. Read your AI contract’s usage terms, not just the price per token.

  2. Budget for pricier AI capacity as discounts meet expensive compute.

  3. Track data-center politics where your providers build.

  4. Isolate every device you can’t patch.

  5. Pair AI code review with traditional scanners.

What Else Happened in AI on November 2, 2026?

  • Cisco’s president gave a 27-year-old ex-founder unlimited tokens to rewrite legacy code. The Rundown AI

  • SpaceX is reportedly in talks to buy EchoStar’s Boost Mobile. Techpresso

  • French MPs voted for a €0.055 tax on every gigabyte sent to users. Techpresso

  • AI-related layoff announcements in August and September were the lowest of the year, a16z notes. Techpresso

  • Apple’s $2,000 iPhone Duo: a glimpse of the future, but hard to justify at that price, says Axios. Axios AI+

🤝 Advertise with AI Unraveled & DjamgaMind

Put your enterprise AI brand in front of the CTOs, CISOs and engineering leaders who make AI buying decisions, daily, in English and French, on podcast apps, YouTube and Substack. Start with a 4-week trial on one named segment.
Sponsorship options & media kit → · Email: etienne_noumen@djamgamind.com

🔗 Tools from the show

🎧 Prefer it commercial-free? Subscribe on Apple Podcasts.

⚗️ PRODUCTION NOTE: We Practice What We Preach.
AI Unraveled is researched and produced by Etienne Noumen, P.Eng. (Calgary, Canada) using a hybrid “Human-in-the-Loop” workflow.



Read original on Substack

Related articles

Level up your Tech Career with DjamgaTech PRO

Download the DjamgaTech PRO App on the Apple App Store to access hundreds of interactive prep questions, flashcards, and quizzes cleanly without ads.

Djamgatech PRO App
Download DjamgaTech PRO on the App Store 🚀

← All articles