Anthropic

Recent news

• Sep 4, 2026

  Self-Improving AI Could Drive Innovation – But Strain Data Centers

  The article discusses how recursive self-improvement (RSI) in AI could affect data center operations, but it does not announce a new project or deal.

  • The piece is analysis/commentary on RSI, focusing on implications for power, cooling, orchestration, governance, and certification in data centers.
  • It cites Anthropic saying Claude authored more than 80% of its code merged in May 2026, and references Q-CTRL, Uptime Institute, Dell’Oro, Phaidra, Emerald AI, etalytics, and Vigilent as examples or commentators.
  • No investment, contract, or funding figures are provided; the article is a forward-looking industry analysis rather than a primary announcement.
• Aug 28, 2026

  At $125 a Month, Cable TV Nearly as Expensive as Typical Monthly Home Electric Bill, Report Finds

  Gov. Morrisey is reported as being encouraged about a proposed $45 billion Nscale-Anthropic data center deal in West Virginia.

  • The item references a $45 billion data center deal involving Nscale and Anthropic in West Virginia.
  • The text is a headline-style reference to a recent or ongoing announcement, not a full primary-source release; no further project specifics are provided in the excerpt.
• Aug 26, 2026

  🤖 La Machine #89: Is Hugging Face About to Become AI’s Next $13B Prize?

  The newsletter republishes and comments on an earlier Mistral AI appearance before French lawmakers, while also summarizing a separate Mistral-HUMAIN partnership announced by Mistral AI.

  • Arthur Mensch told a French parliamentary inquiry on May 12 that Europe has about two years to secure a meaningful share of AI value creation, warning that AI is now the core of the cloud business and that Europe must focus on high-value AI services rather than low-margin commodity cloud services.
  • The issue also references Mistral AI’s separate strategic partnership with HUMAIN for sovereign AI infrastructure, localized models and enterprise applications across Saudi Arabia and the wider region, and mentions broader newsletter items including a €2bn Microsoft AI data center in Alsace and other AI and data-center-related stories.
• Aug 19, 2026

  Federal Trade Commission Chairman Sees AI Antitrust Risks Upstream

  FTC Chairman Andrew Ferguson said the biggest competitive risks in artificial intelligence are in chips, power, and construction rather than among AI models themselves.

  • Speaking at the Technology Policy Institute’s Aspen Forum in a fireside chat with Brian Sullivan of CNBC, Ferguson said competition among large language model developers appears healthy, citing near weekly releases from Anthropic, OpenAI, Google, and Meta.
  • Ferguson said the main concern is input monopolies in chips, energy, and construction, warning against monopoly power upstream and acquisitions that could create a chokepoint in the AI supply chain; he also said FTC enforcement is focused on false marketing claims rather than AI uses deemed merely potentially illegal.
• Aug 19, 2026

  List of AI companies turning to old and rare books for AI training

  Amazon has been reported to be buying old, obscure and rare books and sending them for destructive scanning at its Las Vegas facility VGT3, according to an investigation by 404 Media.

  • 404 Media said it tracked a rare book with a device to Amazon facility VGT3 in Las Vegas; workers reportedly remove spines, scan pages, and discard physical copies.
  • Amazon said it buys books through commercial channels, but did not say how many books it purchases or whether it preserves rare titles; booksellers in Europe and Ireland reported suspicious bulk orders, including 5,000 obscure titles to Kennys.
  • The article also compares Amazon’s practice with similar AI-book uses by Anthropic, Meta, OpenAI, and Google; it cites Anthropic’s Project Panama, Meta’s discussion of buying Simon & Schuster and paying $10 per book, and Anthropic’s later $1.5 billion settlement over pirated books.
• Aug 13, 2026

  As AI power demand soars, Musk turns to natural gas

  The Wall Street Journal reports that SpaceX is increasingly relying on natural gas to power its expanding AI infrastructure.

  • SpaceX is building massive AI data centers and a $16.8 billion chip-manufacturing facility in Texas, and is installing its own natural-gas turbines and power plants because grid connections can take years.
  • The report says SpaceX’s data-center power capacity could rise from about 2 GW to 10 GW or even 15 GW by end-2027; it also says SpaceX and xAI have used gas turbines in Tennessee and Mississippi, and that SpaceX has secured cloud agreements with Google and Anthropic.
• Aug 5, 2026

  Essential Summer Reads #1: How Juisci Is Using AI to Save Doctors 8 Hours a Week and Fight Medical Misinformation

  Juisci has announced a €5.5 million funding round to strengthen its R&D and expand into the U.S. market.

  • The Paris-based startup says it now helps 200,000 healthcare professionals in 60 countries access scientific research, using AI to select, summarize, and recommend peer-reviewed studies.
  • The round was led by Ring Capital with Big Pi Ventures and Bpifrance; CEO Robin Roumengas said the platform saves doctors hours each week by giving them key points and links to the original source.
• Aug 2, 2026

  China’s AI strategy is built for the world beyond America

  Policy Circle has published an analysis arguing that China is pursuing a low-cost, open-weight AI strategy to expand industrial adoption and diplomatic influence.

  • China is backing capable, inexpensive, modifiable AI systems such as Qwen, DeepSeek, Kimi, and Z.ai, while the State Council’s “AI Plus” programme aims to direct AI into scientific research, manufacturing, consumer products, public services, and government.
  • The article says the 15th Five-Year Plan for 2026–30 adds AI agents, multimodal systems, embodied AI, and swarm intelligence, and cites an official ambition for AI-related industries to exceed 10 trillion yuan by 2030.
  • It also says Xi Jinping offered 5,000 AI training and seminar places to developing countries over five years at the 2026 World Artificial Intelligence Conference, alongside proposed cooperation centres with BRICS, ASEAN, the African Union, and Latin American countries.
• Aug 2, 2026

  Sarvam’s AI Arsenal

  Sarvam AI is expanding from Indian-language models into a full-stack enterprise AI platform spanning infrastructure, models, and applications.

  • The startup unveiled Sarvam Epoch and said it now wants to provide everything Indian enterprises need to build, deploy and scale AI, including foundation models, infrastructure, coding copilots, voice agents, document intelligence and workplace applications.
  • Sarvam said Sarvam Inference runs open-weight models on India-hosted infrastructure; it is building a trillion-parameter model, scaling compute to 10,000 accelerators, and has opened a San Francisco office while appointing Devendra Singh Chaplot.
  • The article also reports prior funding and valuation context: Sarvam entered India’s unicorn club after raising $234 Mn in a $300 Mn Series B at a $1.5 Bn valuation; the company’s total funding is listed as $275.0 Mn+.
• Jul 31, 2026

  Microsoft doubles down on multi-model AI as it builds a Copilot super app

  Microsoft has announced it is building a Copilot “super app” that will be rolled out this quarter, alongside continued expansion of its multi-model AI and cloud capacity strategy.

  • The new platform will combine chat, Cowork, long-running Autopilot agents, and Microsoft Scout and will be wired into Agent 365, IT Ops, SecOps, FinOps, CRM, ERP and other business processes.
  • Microsoft also said it added 88 data centers in FY 2026, including 31 across five continents in the past quarter, added another gigawatt of capacity this quarter, and is on track to roughly double overall capacity in two years.
  • The article is a report on Microsoft’s earnings call and quotes Satya Nadella and Amy Hood; it also describes usage-based pricing changes and model-swapping strategy across Microsoft AI products.
• Jul 24, 2026

  The Gigawatt Buildout Faces the Execution Test

  The article is an analysis and commentary piece about the data center and AI infrastructure market, highlighting several recent announcements and deals rather than a single new announcement by one entity.

  • Alphabet raised its 2026 capital spending forecast to $195 billion-$205 billion, while OpenAI said it is developing Project Camellia in Effingham County, Georgia with 3.2 GW of power to be delivered from 2028 through 2032.
  • The piece also cites the $40 billion acquisition of Aligned Data Centers by BlackRock, MGX and GIP, Hut 8‘s 15-year, 352 MW lease in Texas, and several financing and permitting pressures, including a possible $7 billion collateral requirement for Oracle in Wisconsin and $12 billion of financing for Meta’s Texas project.
• Jul 23, 2026

  After the AI Rush, Can Data Centers Reclaim Sustainability?

  The article argues that data center sustainability is under renewed pressure as AI-driven growth increases power, cooling, and water demands, while regulation and community opposition are pushing operators back toward efficiency and environmental stewardship.

  • Regulatory and community pressure: Greenpeace criticism, moratoria in Singapore, Amsterdam, Dublin, and a one-year New York State moratorium on projects over 50 MW are cited as examples of tighter oversight and permitting constraints.
  • Industry response and current context: Operators have long used PUE and renewable PPAs, but the article says AI is shifting priorities toward performance; it cites Frontier’s $1.8 billion permanent carbon removal commitment and an additional $915 million in funding, plus Amazon’s claim it is 75% of the way to water neutrality by 2030 and has cut Northern Virginia water use by 42% year over year.
• Jul 17, 2026

  Right-Sizing Data Centers: How to Match Footprint to Real-World Demand

  The article explains a framework for right-sizing data centers and argues that bigger is not always better.

  • It compares the size spectrum from edge and micro data centers to hyperscale campuses, emphasizing that deployments should be sized around actual business needs.
  • It discusses factors for planning footprint, including IT load, server classes, rack density, cooling, materials, and phased expansion; examples cited include AWS’s 1,200-acre Project Rainer in New Carlisle, Indiana, and CoreSite’s LA1 in Los Angeles.
• Jul 16, 2026

  TeraWulf’s $19B Anthropic Lease Puts Its Brownfield AI Strategy to the Test

  TeraWulf has announced a 20-year, $19 billion lease with Anthropic for approximately 401 MW of critical IT capacity at its Justified Data campus in Hawesville, Kentucky, while also agreeing to sell its 50.1% interest in the 168-MW Abernathy data center joint venture in Texas.

  • Anthropic lease: TeraWulf says the agreement is expected to generate about $19 billion in contracted revenue over 20 years; initial capacity is slated for 2H 2027 and the full 401 MW ramp by early 2028.
  • Abernathy sale and site details: TeraWulf will sell its 50.1% interest in Abernathy for about $530 million, paid in three installments through April 2027; the Hawesville site includes 250+ buildable acres, multiple high-voltage transmission lines, an energized substation, and direct transmission access.