In one minute
- Anthropic reports $895.4 billion in commitments, $16.956 billion in quarterly revenue and a $567 million operating loss. The new document expands the information available about its planned IPO. [1]
- The New York Times and three other newspapers sue Brave; Vujčić identifies AI valuations as a financial risk, according to Reuters. [2–4]
- OpenAI announces Grid Protectors, higher vulnerability rewards and LifeVal. Norway will receive its report on AI in higher education on 6 November. [6–11]
Lead story
BUSINESS · FINANCING · INFRASTRUCTURE
S-1/A prospectus filed with the SEC on 2 November 2026.
Anthropic expands its disclosures on commitments and quarterly results
Juss og KI examined Anthropic's IPO plans on 29 October in “Claude has won over developers. Will Anthropic win over investors?”. The new development is the amended prospectus, which incorporates third-quarter results and offers a closer look at the obligations underpinning growth. [1]
The company reports $895.4 billion in purchase commitments and leases, including short-term agreements and leases that have not yet commenced. Almost all relate to compute and infrastructure. Approximately $447.5 billion is expected to be incurred before the end of 2030, with substantially all the remainder between 2031 and 2045; $69.8 billion falls due in the twelve months following 30 September 2026. [1]
The total includes agreements entered into after that date. It is neither money already spent nor bank debt immediately payable. Furthermore, around $213.8 billion relates to arrangements with termination rights or conditions before commencement, such as financing, facilities or approvals. Enforceability depends on each contract. [1]
Revenue rises, while the net loss requires a closer reading
Between July and September, Anthropic generated $16.956 billion in revenue, compared with $11.555 billion in the previous quarter. It recorded a $567 million operating loss and $9.863 billion in compute and infrastructure expenses. [1]
Its net loss was $23.291 billion, but this includes a $21.936 billion charge from changes in the value of convertible notes, derivative liabilities and agreements for future equity, or SAFEs. That accounting loss should not be read as cash outflow during the quarter or confused with the operating result. The quarterly figures are unaudited. [1]
Capacity contracts carry risks into the future
The prospectus warns that Anthropic secures capacity well in advance and may have limited scope to adjust its commitments if demand changes. Much of the business depends on customer consumption; securing capacity does not guarantee it can be used profitably. [1]
Some providers depend on third-party financing. Delays in financing, construction or electricity supply can affect available capacity. Defaults under certain commercial agreements may also trigger default provisions in financing arrangements. These are disclosed risks, not evidence that they have materialised. [1]
Why it matters: assessing an AI company also requires examining payment schedules, contractual conditions and who finances its facilities. An aggregate commitment figure alone cannot establish the risk of insolvency.
Limits of verification: these are figures reported by Anthropic to the SEC. The amended prospectus neither completes the IPO nor establishes a final valuation. [1]
Other news
1. The New York Times and other newspapers sue Brave over its use of their articles
COPYRIGHT · 2 NOVEMBER
The New York Times filed case 1:26-cv-09549 against Brave Software in the Southern District of New York. Daily News, Chicago Tribune and Orlando Sentinel filed a separate action, 1:26-cv-09554. They allege unauthorised copying and distribution through Brave's search index, Ask Brave answers and the interface that supplies content to other AI companies. They also allege circumvention of paywalls. [2][3]
The complaints name providers including Anthropic, Perplexity and xAI, but those companies are not defendants in these proceedings. The newspapers' allegations remain unresolved. The dispute concerns both obtaining information and reusing it to answer queries: indexing, training and redistributing content raise distinct questions. [2][3]
Why it matters: contracting with a search intermediary does not automatically settle the rights to the material it delivers.
2. Vujčić identifies AI valuations as a risk to financial stability
ECONOMY · EUROPE · 2 NOVEMBER
According to Reuters, ECB Vice-President Boris Vujčić identified high AI stock valuations, sovereign debt market stress and potential losses at non-bank financial intermediaries as major risks. Speaking at an event in Ireland, he warned of Europe's exposure to a technology correction through investment funds. [4]
The ECB confirms that its next Financial Stability Review will be published on 18 November. The comments flag concerns; they are not the full report or a declaration of crisis. The implications for Europe depend on investors' positions, leverage and liquidity needs if losses occur. [4][5]
Why it matters: AI expansion connects technology companies, financiers and investors. Concentration of exposures deserves as much attention as valuations.
3. OpenAI launches Grid Protectors to cut future campus demand at critical times
ENERGY · INFRASTRUCTURE · 2 NOVEMBER
OpenAI proposes agreements with utilities and public authorities to adjust consumption when the grid is under strain. Its planned campus in Effingham County, Georgia, will serve as a pilot. The agreement with Georgia Power provides for up to 1,000 megawatts of flexible demand response: consumption that could be reduced when needed. The utility confirms that planned capacity and its right to restrict supply under specified circumstances. [6][7]
OpenAI commits to paying the project's infrastructure and electricity service costs. The new development is the national programme: Georgia Power had already announced its agreement on 22 July. Arrangements elsewhere remain to be finalised. These mechanisms concern future facilities; no demand reductions already carried out can be attributed to the programme. [6][7]
Why it matters: flexibility can help manage peaks. Its value will depend on how much consumption can be reduced, for how long and under what enforceable obligations.
Briefs
OpenAI raises vulnerability rewards and prepares new tests. It announces up to $30,000 for critical flaws in core products and up to $172,000 with bonuses for exceptional findings. Another internal testing round with Cathedral researchers is being planned, rather than already completed. Paying for vulnerability reports encourages discovery but does not establish that flaws have been fixed or replace independent security assessment. [8][9]
OpenAI introduces LifeVal to evaluate agents on everyday tasks. The evaluation covers care, education, finance, shopping and access to government services. The provider says it draws on time-use studies and a demographically representative survey of 3,995 people. Scores for individual examples do not establish an overall real-world success rate. Finding information, preparing an application and validly submitting it are different actions: LifeVal does not establish legal reliability or grant authority to act for another person. [10]
Norway announces the handover of its report on AI in higher education. The Malthe-Sørenssen committee will deliver NOU 2026:14, Kunstig intelligens i høyere utdanning: I kunnskapens tjeneste, on 6 November. The Ministry of Education and Research announced this on 2 November. This is notice of the handover; it does not establish that the final recommendations have been published or adopted as government decisions. [11]
The signal in the noise
AI expansion is also shaped by contracts. Capacity commitments, content licences and electricity flexibility allocate costs and responsibilities for years ahead. To assess a promise, ask when it becomes binding, what conditions limit it and what evidence will demonstrate performance.
On the radar
- ADNOC announces a partnership with G42 and OpenAI. On 2 November, the energy group announced a multiyear agreement to connect operational and commercial decisions using AI. The promised productivity benefits are the partners' expectations. [12] - Palantir reports quarterly revenue of $2.313 billion, up 96% year on year. According to its 2 November results, US commercial revenue exceeded US government revenue for the first time. These are corporate, unaudited quarterly figures. [13]
Sources
[1] SEC · Anthropic, S-1/A, 02-11-2026 · quarterly results and contractual obligations
[2] The New York Times v. Brave Software · complaint, 02-11-2026 · PDF
[3] Daily News, Chicago Tribune and Orlando Sentinel v. Brave Software · complaint, 02-11-2026 · PDF
[5] ECB · Financial Stability Review · release schedule
[6] OpenAI · Introducing the Grid Protectors Program · 02-11-2026
[7] Georgia Power · Georgia Power to serve OpenAI project in Effingham County
[8] OpenAI · Rewarding defenders with bigger bug bounties · 02-11-2026
[9] OpenAI · Preparing our next security sprint to harden OpenAI · 02-11-2026
[10] OpenAI · LifeVal · 02-11-2026
[11] Government of Norway · notice of the report handover on AI and higher education · 02-11-2026
[12] ADNOC · Partnership Agreement with OpenAI and G42 · 02-11-2026
[13] Palantir · Q3 2026 results · release furnished to the SEC · 02-11-2026