FR
Live

Claude, the essentials — edition of July 23, 2026

Anthropic Settles for $1.5B in Book Case as AMD Commits $5B and 2GW of Chips

By ·

Anthropic closed one of the largest AI copyright settlements on record and sealed a multibillion-dollar chip pact with AMD on the same day Alphabet's earnings showed how much its stake in the company now matters.

Key points

A $1.5 Billion Reckoning Over Training Data

Anthropic has agreed to pay $1.5 billion to resolve claims brought by thousands of authors and publishers, including the publisher of Harry Potter, who argued that Claude was trained on pirated copies of their books. Multiple outlets converged on the same figure and scope, indicating the settlement is finalized rather than merely proposed, and place it among the largest payouts an AI company has made over its training-data practices.

The settlement effectively puts a price on a practice that most AI labs have so far treated as a legal grey area: acquiring copyrighted text without a licensing agreement. For Anthropic, it converts an open-ended liability into a fixed, if enormous, cost — and it hands publishers and authors elsewhere a concrete benchmark for future claims against other AI developers.

Sources: Un accord historique de 1,5 milliard de dollars, et des millions pour l'éditeur de Harry Potter: c'est le prix qu'Anthropic doit payer pour l'utilisation non autorisée de milliers de livres destinés à entraîner son IA · Anthropic rattrapée par la loi : l'entreprise devra payer 1,5 milliard de dollars à des milliers d'auteurs après avoir utilisé des livres piratés pour entraîner Claude · IA et droit d'auteur : Anthropic versera bien 1,5 milliard de dollars aux auteurs et éditeurs

AMD Buys Into Anthropic's Compute Future

AMD is putting $5 billion into Anthropic, and in return Anthropic is committing to buy around 2 gigawatts of AMD GPU capacity — a scale of chip commitment usually associated with Nvidia's dominant supply agreements. Several outlets frame the deal as a direct move against Nvidia's position: it gives AMD a marquee AI customer and gives Anthropic a second major silicon supplier, at a moment when GPU capacity, not just funding, remains the binding constraint on how fast frontier labs can train and serve models.

The agreement fits Anthropic's pattern this year of diversifying compute suppliers rather than relying on a single partner. Locking in gigawatt-scale hardware alongside an equity investment signals that securing chip supply has become as central to Anthropic's strategy as raising capital itself.

Sources: AMD et Anthropic concluent un pacte sur les puces de cinq milliards de dollars · AMD donne 5 milliards à Anthropic, qui lui achète 2 GW de GPU · Intelligence artificielle : Anthropic achète pour 2 GW de capacité énergétique en GPU à AMD en échange d'un investissement de 5 milliards de dollars · AMD investit 5 milliards de dollars dans Anthropic pour défier Nvidia

Investors Start Pricing the Cost of the AI Build-Out

Alphabet, one of Anthropic's investors, reported a record $112 billion in quarterly profit, with results lifted in part by its stakes in Anthropic and SpaceX. Yet the stock fell on the same results, as markets reacted less to the profit itself than to the scale of AI infrastructure spending Alphabet disclosed for the period ahead.

That reaction fits a broader pattern visible the same day: OpenAI reported plans to commit roughly $750 billion to its own AI infrastructure, a figure of the same order of magnitude as the capital now flowing into Anthropic through AMD's $5 billion investment and 2-gigawatt chip commitment. Taken together, the numbers suggest investors are no longer simply rewarding announcements of AI spending — they are starting to weigh whether the returns can keep pace with the capital being committed.

Sources: Gonflé par SpaceX et Anthropic, le bénéfice d'Alphabet atteint le record de 112 milliards de dollars sur un trimestre, mais l'action chute avec les investissements faramineux prévus par le groupe · OpenAI prévoit un investissement colossal de 750 milliards de dollars dans ses infrastructures d'IA · OpenAI flambe : ses dépenses en IA explosent et atteignent 750 milliards de dollars, la pression est maximale

Claude's Product Line Keeps Moving

Away from the balance-sheet news, Claude's voice mode is being extended to the Opus and Sonnet model families, widening the tiers of the assistant where users can interact by speech rather than text. Separately, Claude Opus 5 has begun appearing on independent benchmark trackers such as Artificial Analysis, giving outside observers a first standardized read on how the model compares to competitors.

Neither development changes the shape of Anthropic's business on its own, but both point to a company continuing to ship and refine its core product even as it absorbs a landmark legal settlement and closes a multibillion-dollar hardware deal in the same stretch.

Sources: Claude : mode vocal étendu à Opus et Sonnet · Claude Opus 5

This edition is an original synthesis written by Claude from aggregated news (press, Hacker News, Reddit, GitHub), under the editorial supervision of Héra SASU. Every fact links to its source. See the live feed →

Claude News is published by Héra SASU. Independent media, not affiliated with Anthropic.