Anthropic piles up computing bills; Marvell's laser bet rattles investors

Good morning.

Investors are being asked to swallow some very big claims.

Anthropic put about $895.4 billion in purchase commitments and leases in Monday's amended IPO filing. Its risk factors suggest it may be near its own threshold for AI speeding up AI progress.

Marvell agreed on Monday to buy Lumentum for an implied enterprise value of $115 billion in stock. The market had its own response before the bell.

Palantir CEO Alex Karp has an explanation involving the physics department. His company's third-quarter revenue grew 96% from a year earlier, and Monday's analyst call took a turn.

Stay curious,

Marcus Schuler

Anthropic listed about $895.4 billion in purchase commitments and leases in Monday's amended IPO filing, substantially all related to compute and infrastructure.

Third-quarter 2026 revenue reached roughly $17 billion. We reported the $16.8 billion investor figure on Oct. 15. Sequential growth slowed to around 47%, from 144% in the second quarter.

These figures do not establish when the commitments come due. The $104 billion net loss for the nine months through Sept. 30 came mostly from fair-value changes in convertible notes and derivative liabilities.

Why This Matters:

• IPO investors are approaching a roadshow with the filing's price range and share count still blank.
• Anthropic's risk factors say it does not anticipate safely using Claude to "fully autonomously design and build successive generations of frontier models" without human supervision.

Marvell agreed on Monday to buy Lumentum in an all-stock deal with an implied enterprise value of $115 billion.

The purchase would bring Lumentum's lasers and optical components into Marvell. Investors have reason to scrutinize the price: Lumentum shares fell 6.5% before Monday's bell. Marvell would pay $3.2 billion if regulators block the deal.

Deep Analysis follows enterprise AI spending into budget lines and model routing in Matt Mullen's Oct. 27 piece.

The routing split of "95%-5% in favour of open weight/open source models over commercial, frontier models" comes from one unnamed platform's enterprise clients. Mullen quotes Torsten Slok: "mainstream corporate America does not need the latest model and is unwilling to pay for it."

• Ukraine's Air Alert app sent fake radiation warnings late Sunday; Russia later said its own alert app was hacked, and neither side's claims are independently confirmed.
• Supermarkets in six European countries suffered a suspected cyberattack overnight into Monday, with refrigerator and freezer temperatures reportedly reaching 50C (122F) in several shops.
• Salesforce announced Monday a definitive agreement to buy Cresta, whose voice agents handle customer calls for Marriott, Comcast and Alaska Airlines.
• Alphabet's €6 billion bond sale on Monday drew just over €17 billion in orders across seven tranches.
• Trajectory raised a $45 million Series A at a $300 million valuation, led by Sequoia, with Nvidia also investing.
• DeepSeek released Creator Mode for plugin building, warning, "a single prompt may not reproduce exactly what you see in the video."

Turn a strategy memo into a falsifiable bet.

A strategy memo can sound persuasive while hiding the assumption that decides whether the plan works. Use an LLM to expose that assumption and define an early warning signal.

Your raw input:

The prompt:

Why this works:

The prompt forces the model to connect a recommendation to evidence and a measurable claim. The quote requirement keeps the critique anchored in the memo instead of producing generic risk language.

What to use:

Use a reasoning model with a long context window for a full memo. A mainstream general model works for documents under 10 pages.

Plansom turns a rough objective into editable goals, sub-goals and tasks that remain in a shared workspace, a more useful result than a one-off chatbot outline when a team must track execution. As checked October 20, 2026, the Free plan costs $0 and includes 5 goals, 100 tasks and 1 AI Plan per month; Pro costs $28 per user per month and adds integrations and expanded AI features.

How to use it:

1. Open AI Plan in the left panel.
2. Choose the built-in prompt template or enter a short description of the result you need.
3. Answer the clarifying questions about the goal.
4. Review the generated sub-goals and tasks, then add or delete items that do not fit.
5. Open the goal, update its title, description or owners, and save it.

Where it falls short: The Free plan has no integrations and permits only 1 AI Plan per month as checked October 20, 2026.

Palantir CEO Alex Karp compared enterprise software demand to dating on Monday's analyst call. Palantir's third-quarter revenue rose 96% from a year earlier to $2.313 billion.

Our take: A CEO explaining enterprise software demand as a date has apparently decided procurement needs a varsity jacket. Karp said, "We've gone from being the most attractive person in the physics department to being the captain of the football team kind of thing." He added, "I mean, not that I did that."

The customer is worried about intellectual property; the salesman is casting himself as the football captain. "This is like first date into bed kind of thing," Karp said on Monday's call, then: "Maybe that's not the right metaphor for this, but it's a very different jam." Somebody appears to have asked for separate checks. U.S. commercial contract value grew 11% in the third quarter, after 153% the previous quarter.

Marcus Schuler

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai