Menu
MCAP $2.92T ▼-2.06%
BTC $86,464 ▲+1.97%
ETH $2,728 ▲+1.33%
BNB $800.56 ▲+2.20%
XRP $1.520 ▲+1.99%
SOL $121.18 ▲+0.77%
DOGE $0.0961 ▲+3.78%
ADA $0.270 ▲+10.79%
TRX $0.3358 ▲+0.10%
LINK $14.16 ▲+0.98%
AVAX $11.00 ▼-0.49%
HYPE $90.79 ▲+1.28%
DOT $1.210 ▲+2.57%

Anthropic

Every Crypto Coin Show story tagged Anthropic, newest first, alongside the guides and interviews that give it context.

Newsletter
Anthropic pre-IPO derivatives hit $643M September volume ahead of real equity test
Equities

Anthropic pre-IPO derivatives hit $643M September volume ahead of real equity test

, , , , , , , ,

Editor's ChoiceCrypto Coin Show News Team·September 30, 2026·4 min read

Anthropic’s pre-IPO crypto derivatives have reached $643 million in September trading volume alone, but the arrival of a real stock price will force a hard reckoning between the synthetic valuations traders have built and Wall Street’s opening bell. Institutional crypto investors face a liquidity test: whether the market survives when its reference point shifts from algorithmic mark prices to actual equity trading.

  • Pre-IPO perpetual contracts on Anthropic generated $643 million in volume through September 21, exceeding August’s total of $590 million.
  • Binance’s Anthropic open interest jumped 88% in 30 days to $31.2 million, with combined Anthropic and OpenAI pre-IPO open interest above $160 million as of mid-September.
  • SpaceX’s post-IPO precedent shows perpetual volume can expand dramatically after listing, averaging $2.2 billion daily in the following 30 days versus $89 million beforehand on Binance alone.
  • $643M September volume in Anthropic pre-IPO perpetuals across 12 crypto venues
  • 88% Binance Anthropic open interest growth over 30 days, reaching $31.2 million
  • $2.2B Average daily SpaceX perpetual volume in 30 days post-IPO versus $89 million before

Anthropic’s planned initial public offering will force crypto derivatives markets to bridge the gap between their synthetic price discovery and the real equity valuation that Wall Street establishes. According to Binance, traders have built $31.2 million in open interest on Binance alone through mid-September without any public shares anchoring valuations.

Instead, mark prices derive from recent trading in the contracts themselves, a circular process that will break the moment Anthropic lists. The transition creates potential repricing events for position holders, with unrealized gains, collateral values and liquidation thresholds all subject to shift.

Binance converts mark price from contract trading to stock index on IPO

Before an IPO, Binance averages recent contract prices and limits daily mark-price moves, allowing traders to value positions in a market with no public reference point.

The exchange can convert Anthropic’s pre-IPO perpetual into a standard equity-linked contract once it has access to a stable third-party stock-price index, but the process does not necessarily begin the moment shares start trading. During the transition, the mark price gradually shifts toward stock-linked calculation, a period where the crypto valuation and real stock price may diverge sharply.

If Anthropic’s crypto-implied valuation differs materially from the IPO opening price, that convergence could alter unrealized gains and losses across thousands of positions simultaneously.

The scale makes the risk concrete rather than theoretical. Combined open interest in Anthropic and OpenAI pre-IPO perpetuals exceeded $160 million as of mid-September, up from roughly $1 million in April, a 160-fold increase in four months.

SpaceX listing expanded perpetual volume 25-fold after mark price conversion

The clearest precedent comes from SpaceX, which listed in June. When Binance transitioned its perpetual from pre-IPO cash settlement to equity-linked pricing, the mark price took as long as three hours to fully converge depending on volatility.

During the following 30 days, average daily perpetual volume across all venues reached $2.2 billion, against roughly $89 million on Binance alone in the pre-IPO period. Open interest had exceeded $300 million before listing and peaked 11 days after the IPO, according to FalconX.

What remains unknown: whether original pre-IPO traders stayed through the transition or new participants entered once price discovery improved and hedging became easier against a real stock.

Anthropic’s market currently sits in an earlier phase than SpaceX was at the same point in its timeline.

Anthropic’s frontier AI profile and $2 trillion valuation may defy historical patterns

Anthropic differs materially from prior precedents. SpaceX was an aerospace and infrastructure play; SK Hynix was a semiconductor and hardware business with established manufacturing. Anthropic is a pure-play frontier AI lab, a category without prior listed derivatives history.

A second variable complicates forecasting: Hyperliquid captures a larger share of Anthropic’s pre-IPO perpetual activity than it did for SpaceX or SK Hynix at comparable stages. TradeXYZ, which historically deploys these markets and later transitions them after listing, has not yet launched Anthropic pre-IPO perpetuals on Hyperliquid and is likely waiting for the S-1 filing before proceeding.

Once that filing arrives, volume could accelerate sharply in the final weeks before listing, following the steep S-curve observed in prior events.

The gap between crypto’s synthetic valuation and Wall Street’s opening price remains the open question.

The CCS read. Liquidity may not be the real test, SpaceX proved perpetuals thrive post-IPO. The real question is whether Anthropic’s $2 trillion crypto valuation holds at IPO pricing. If Wall Street values the company below market expectations, the repricing could trigger cascading liquidations across leveraged positions built on the assumption of a higher opening. That repricing risk is what separates this from a normal volatility event.

The next signals arrive with Anthropic’s formal IPO timetable and the S-1 filing that may already be in draft form at the SEC. Once that document surfaces, institutions will watch three metrics: the gap between Anthropic’s crypto-implied valuation and the IPO price, the speed of mark-price convergence during Binance’s transition period, and whether the $31 million in open interest built before listing survives once the synthetic price has to answer to a public market. Traders will also monitor whether TradeXYZ launches the market on Hyperliquid before or after listing, a timing decision that first reported by CryptoSlate, could shape pre-IPO volume patterns in the final stretch.

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

Anthropic dragged to court by Warner, Sony Music over ‘blatant theft’ of IP
AI News

Anthropic dragged to court by Warner, Sony Music over ‘blatant theft’ of IP

, , , , , ,

AI NewsAugust 31, 2026·6 min read

Anthropic faces a second major copyright lawsuit in months, this time from Sony Music Publishing and Warner Chappell Music over alleged piracy of more than 20,000 songs used to train Claude. The case hinges on a legal principle established in a prior ruling: that using copyrighted material for AI training may constitute fair use, but obtaining that material from pirated sources does not.

  • Sony and Warner allege Anthropic pirated over 20,000 songs to train Claude via BitTorrent, Pirate Library Mirror, and licensed services
  • Co-founder Benjamin Mann is named as individual defendant and accused of downloading 5 million pirated books containing lyrics and sheet music
  • Plaintiffs seek $150,000 statutory damages per pirated song plus $25,000 per copyright notice stripped, totaling billions across the alleged catalog
  • 20,000+ Songs allegedly pirated and used for Claude model training
  • $1.5B Settlement Anthropic paid in separate author piracy case decided in 2025
  • $150K Statutory damages sought per pirated song under federal copyright law

Sony Music Publishing and Warner Chappell Music filed a federal complaint in Northern California on Friday, August 28, accusing Anthropic of operating what they describe in the 48-page filing as “one of the largest and most blatant ongoing thefts of intellectual property in history.” The lawsuit names not only the company but also CEO Dario Amodei and co-founder Benjamin Mann as individual defendants.

The core allegation is that Anthropic obtained over 20,000 copyrighted songs through illegal means and fed them into the training data for Claude, Anthropic’s flagship conversational AI model, without authorization or compensation to the rights holders.

The specific methods of acquisition matter greatly here. According to the complaint, Mann personally obtained more than five million pirated books using BitTorrent technology, while Anthropic employees pulled down an additional two million books from Pirate Library Mirror, a notorious file-sharing platform. These materials contained song lyrics and sheet music alongside text.

The company also obtained licensed lyrics directly from MusixMatch and LyricFind, both commercial services that operate under licensing agreements and compensate record labels for content access. The plaintiffs argue this dual approach demonstrates Anthropic’s knowledge of the legal distinction between licensed and unlicensed sources, yet proceeded with the latter anyway.

The suit catalogs specific commercial recordings allegedly included in the pirated dataset: Mariah Carey’s “All I Want for Christmas is You,” Taylor Swift’s “Paper Rings,” Marvin Gaye and Tammi Terrell’s “Ain’t No Mountain High Enough,” Bon Jovi’s “Livin’ on a Prayer,” Earth, Wind & Fire’s “September,” and Leonard Cohen’s “Hallelujah.”

Prior court ruling on fair use establishes the precedent now haunting Anthropic

This lawsuit arrives in the context of a recent and closely watched ruling that reshapes how courts view AI model training and copyright.

Earlier in 2025, a federal judge determined that training an AI model on copyrighted material can qualify as fair use under U.S. copyright law, a decision that sent waves through the AI industry and appeared to offer legal cover for large language model developers.

That same judge, however, added a critical caveat: obtaining the copyrighted works from pirated sources does not enjoy fair use protection, even if the downstream use of the material in AI training might.

Anthropic itself was the defendant in that case, brought by authors alleging that copyrighted books were used to train Claude. The company settled for $1.5 billion, a settlement figure that reflected both the scale of the alleged infringement and the judge’s mixed ruling on fair use doctrine.

Now Sony and Warner are invoking that same precedent in their music copyright complaint, arguing that Anthropic’s pattern of piracy demonstrates the company has not altered its behavior despite the earlier judgment and settlement.

The plaintiffs specifically reference the judge’s finding in their filing, claiming Anthropic’s business model remains “built on copyright theft” rather than licensed or legitimately sourced training data.

The legal distinction cuts to the heart of institutional risk for Anthropic: fair use may shield AI training methodology, but it does not shield piracy.

Statutory damages framework creates exposure in the billions across the alleged song catalog

Federal copyright law permits rights holders to seek statutory damages without proving actual financial loss, a provision that dramatically increases the financial exposure for companies accused of large-scale piracy.

Sony and Warner are seeking $150,000 in statutory damages for each pirated song and an additional $25,000 for each instance where copyright information was stripped or removed from a work. Across an alleged catalog of over 20,000 songs, the combined exposure reaches into the billions of dollars.

Notably, this is not a damages calculation based on lost licensing revenue or market harm; it is a per-work statutory penalty that assumes willful infringement.

The plaintiffs have explicitly requested a jury trial, a choice that matters tactically. Jury verdicts in copyright cases involving major entertainment companies often reflect broader cultural and moral judgments about piracy, not merely technical legal analysis.

The fact that Anthropic is simultaneously managing a $1.5 billion settlement from a related copyright case, brought by a different group of rights holders, undercuts any argument that the company faced genuine uncertainty about the legal implications of sourcing training data from pirated material.

Sony and Warner’s decision to name Amodei and Mann as individual defendants is also significant. Individual liability exposure can affect executive behavior and board-level decision-making in ways that corporate-only liability may not.

For institutional investors evaluating Anthropic’s risk profile, the stacking of two major copyright settlements in the same calendar year, plus ongoing litigation from other content creators, raises questions about whether the company’s compliance infrastructure around data sourcing meets institutional standards.

Broader pattern of AI copyright litigation tests regulatory and market patience

Anthropic is far from alone in facing copyright litigation tied to AI model training. OpenAI, Meta, Google, and other major AI labs have all been named as defendants in separate copyright suits brought by authors, visual artists, journalists, and now music publishers.

The cumulative effect of this litigation wave is to establish a legal and factual record of how AI companies sourced training data, what they knew about the legality of those sources, and whether they made good-faith efforts to license or obtain consent where required.

The music industry’s decision to pursue Anthropic alongside earlier suits by the Authors Guild and other groups suggests a coordinated strategy among major rights holders to establish clear precedent that AI companies cannot treat copyright piracy as a cost of doing business.

Each successful suit or settlement raises the legal standard for competitors and forces subsequent entrants into the AI market to justify their data sourcing methods more rigorously. Institutional investors in the AI sector should expect this litigation to continue and to inform future licensing negotiations between AI companies and content providers.

For Anthropic specifically, the back-to-back copyright exposure creates a compounding problem. The company must now defend both its data sourcing methodology and its individual executives in federal court while simultaneously managing the integration of the $1.5 billion author settlement into its financial and operational planning.

Regulatory bodies, including the European Commission and potential U.S. legislative action on AI governance, are also watching these cases closely as they develop policies around AI training transparency and data provenance.

The case is set for jury trial in Northern California federal court; no trial date has been announced. The key open question for institutional observers is whether a jury will treat the prior $1.5 billion settlement as evidence of a knowing pattern or as a good-faith resolution that should reduce exposure in the music case. Sony and Warner’s explicit invocation of the judge’s prior ruling suggests they expect a court to view Anthropic’s behavior as escalated misconduct rather than isolated infringement, a framing that could significantly affect both verdict and settlement dynamics.

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

The Music Industry’s New Lawsuit Against Anthropic Should Have Dario Amodei Shivering With Fear
AI News

The Music Industry’s New Lawsuit Against Anthropic Should Have Dario Amodei Shivering With Fear

, , , , , , ,

AI NewsAugust 31, 2026·6 min read

Sony Music and Warner Bros. have filed a federal lawsuit against Anthropic alleging the AI company trained its Claude models on tens of thousands of copyrighted musical compositions without permission, potentially exposing the firm to billions in damages. The suit carries institutional weight because it directly references the same pirated dataset involved in Anthropic’s $1.5 billion settlement just 18 months ago, making a fair-use defense substantially harder and raising questions about whether the company’s compliance practices have truly improved.

  • Sony and Warner are seeking up to $150,000 per copyrighted work plus $25,000 per instance of intentional copyright data stripping, potentially totaling several billion dollars in damages.
  • The lawsuit names the identical dataset of illegally downloaded materials that Anthropic cofounder Benjamin Mann previously torrented, the same conduct that triggered the $1.5 billion settlement in September 2025.
  • Anthropic’s stated defense rests on fair-use doctrine, but the company’s recent billion-dollar settlement severely weakens that argument and signals heightened regulatory and legal risk for institutional investors.
  • $150,000 per copyrighted work in damages Sony and Warner seek, plus penalties for copyright stripping.
  • $1.5B settlement Anthropic paid 18 months prior over the same pirated dataset now central to this suit.
  • Tens of thousands of copyrighted musical compositions allegedly used to train Claude without authorization.

Sony Music Group and Warner Bros. Discovery filed a federal lawsuit against Anthropic on Friday, accusing the AI company of training its Claude language models on tens of thousands of copyrighted musical compositions without permission or compensation.

The plaintiffs are seeking damages of up to $150,000 per work, plus $25,000 for each instance Anthropic allegedly intentionally removed copyright metadata, potentially putting total liability in the billions of dollars.

The suit specifically names copyrighted hits including Bon Jovi’s “Livin’ on a Prayer,” Earth, Wind & Fire’s “September,” and Leonard Cohen’s “Hallelujah” as works incorporated into Anthropic’s training pipeline without authorization.

What makes this lawsuit exceptional is its legal specificity and timing.

Anthropic Faces Second Wave of Damages Over Same Pirated Dataset From 2025 Settlement

The lawsuit directly references the identical dataset of illegally torrented materials that Anthropic cofounder Benjamin Mann downloaded from piracy sites, conduct that resulted in the company’s $1.5 billion settlement in September 2025.

That prior settlement involved thousands of copyrighted books and other written works; the new music industry suit argues the same compromised dataset contained sheet music and song lyrics extracted from those publications.

Sony and Warner’s legal filing states: “Among the many millions of books that Defendants torrented from these illegal pirate websites were books containing the lyrics and sheet music to hundreds or more of Music Publishers’ copyrighted musical compositions.” The timing compounds Anthropic’s legal exposure: the company has had fewer than 18 months to implement compliance controls addressing the exact vulnerability that generated its largest financial penalty.

Anthropic’s response, delivered through a company spokesperson to Reuters, characterizes the suit as recycled allegations already pending before courts. The company stated it will defend itself “robustly” and argued that training AI models on copyrighted material constitutes fair use, a legal doctrine permitting limited access to protected works without permission for transformative purposes.

However, that defense strategy faces a substantial credibility problem: Anthropic has already paid out $1.5 billion to settle substantially similar claims, a settlement that implicitly acknowledged the company could not prevail on fair-use arguments in actual litigation.

The nine-figure payout signals that courts or settlement negotiators rejected the fair-use framing as applied to Anthropic’s practices.

This is the third copyright infringement lawsuit filed by the same legal team representing music publishers, suggesting a coordinated enforcement strategy targeting distinct categories of intellectual property. Earlier suits involved books and news articles; this action extends liability exposure to musical compositions.

The stacking of claims across different media categories, each carrying potential statutory damages in the hundreds of millions, creates cumulative financial and reputational risk that extends beyond any single case.

Fair-Use Defense Weakened by Prior Settlement and Benjamin Mann’s Direct Involvement as Defendant

Anthropic’s planned reliance on fair-use doctrine faces acute legal headwinds given that Benjamin Mann, named directly as a defendant, was personally identified as the individual who conducted the illegal torrenting.

Fair use typically applies to incidental or secondary copying undertaken by parties acting independently; when the same individual responsible for the underlying piracy is also a cofounder of the company using that pirated material to build commercial products, courts are unlikely to credit claims that the use is transformative or falls within fair-use bounds.

The suit describes this as “a brazen campaign of illegally torrenting, scraping, and downloading copyrighted works on a massive scale in order to develop, operate, and reap enormous profits from Anthropic’s Claude series of artificial intelligence models.”

Institutional investors face direct consequences from this legal escalation. Copyright liability has become Anthropic’s largest disclosed financial risk, now generating second and third waves of damages claims across different asset categories, books, news, music, rather than resolving through a single settlement framework.

Each category carries distinct damages calculations and statutory maximums, making total exposure difficult to quantify. The company’s settlement history demonstrates that fair-use arguments have failed when tested, yet Anthropic continues advancing this defense rather than negotiating earlier settlements that might cap damages.

The reputational dimension cuts deeper for institutional stakeholders than the numerical damages alone.

Anthropic has been marketed as a company with strong safety and ethics practices, contrasting itself favorably against competitors with looser operational standards.

The discovery that the company allegedly continued incorporating illegally obtained materials into production systems even after the Mann settlement suggests either: (a) compliance systems implemented post-settlement were ineffective, or (b) the company did not meaningfully change operational practice despite the financial penalty.

Neither interpretation reassures institutional investors concerned with governance quality or legal risk management.

Damages Could Exceed Prior Settlement as Music Suits Expand Across Different Publishing Rights Categories

Sony Music Group and Warner Bros. Discovery control vast catalogs of published compositions, giving them leverage to pursue claims affecting multiple distinct rights holders simultaneously.

The suit seeks $150,000 per work, a calculation that assumes tens of thousands of copyrighted compositions, yielding potential liability well into the billions of dollars, substantially higher than the $1.5 billion paid to settle book-related claims.

Music publishing damages carry higher statutory maximums than written-work infringement, reflecting the distinct legal frameworks governing musical compositions.

The distinction matters for institutional capital allocation: if courts sustain the music publishers’ damages framework, it establishes precedent for substantially higher per-work penalties in future cases, potentially making outstanding copyright liability estimates meaningless.

An investor relying on Anthropic’s public disclosures about legal contingencies could find those estimates rendered obsolete by a single adverse judgment in this case. The company has not disclosed the total number of musical compositions allegedly involved, making it impossible for external stakeholders to calculate maximum exposure accurately.

Settlement negotiations will likely dominate the next phase of this litigation. Sony and Warner have demonstrated willingness to litigate but could face pressure from investors seeking certainty rather than years of discovery and trial risk.

Anthropic, facing a credibility problem with fair-use arguments, may find settlement the only viable path to contain reputational damage and limit discovery into the company’s internal practices regarding copyright compliance.

However, the fact that this suit was filed despite, or perhaps because of, the prior settlement suggests the publishers view Anthropic’s original penalty as inadequate compensation and may demand a substantially larger second settlement to prevent further litigation.

Anthropic has stated it will defend the case “robustly,” but the company’s legal strategy remains unstated: will it continue the fair-use argument that failed in the prior settlement, or will it enter settlement negotiations acknowledging the weakness of that position? The answer will arrive through discovery filings or negotiated resolutions expected within the next six to nine months, with both outcomes carrying material implications for institutional investors’ assessment of the company’s financial stability and management quality.

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

Anthropic IPO Could Come in September, But It Has a Massive Risk Factor
AI News

Anthropic IPO Could Come in September, But It Has a Massive Risk Factor

, , , , , ,

AI NewsAugust 27, 2026·6 min read

Anthropic is preparing for an IPO as soon as late September with a prospectus expected after Labor Day, planning to allow early investors and employees to sell existing shares alongside newly issued stock, a departure from SpaceX’s June listing that could expose the market to substantial insider selling pressure. For institutional investors, this approach creates both liquidity opportunities and concentrated risk: the company is exploring longer-than-standard lockup periods to manage the stock flood, but the mechanics remain fluid ahead of the filing.

  • Anthropic plans IPO prospectus filing after September 7 Labor Day, with listing potentially in late September or early October 2026.
  • Company raised $65 billion in May at $965 billion valuation from Altimeter, Dragoneer, Greenoaks, Sequoia and sovereign funds including Singapore’s GIC.
  • Anthropic will allow insider stock sales in IPO itself, unlike SpaceX June debut, but is weighing longer-than-standard lockup periods to manage supply.
  • $65B Anthropic Series E funding raised in May, valuation benchmark for institutional investor entry.
  • 911.5M SpaceX insider shares that became freely tradeable on August 6, more than doubling the liquid float overnight.
  • $86B SpaceX June IPO proceeds, largest listing ever, with zero shares from existing shareholders.

Anthropic is preparing to go public as soon as late September following a prospectus filing expected shortly after Labor Day, according to recent reporting.

The AI safety company plans to structure its debut differently from SpaceX’s record-breaking June listing, most significantly by allowing early-stage investors and employees to sell existing shares within the IPO itself rather than reserving the offering purely for newly issued company stock.

That decision carries material implications for how much insider supply will hit the market on day one, and how the stock will perform once trading begins.

The company closed a $65 billion Series E funding round in May at a $965 billion valuation, drawing capital from leading venture firms Altimeter, Dragoneer, Greenoaks and Sequoia alongside sovereign wealth funds including Singapore’s Government Investment Corporation and traditional asset manager Capital Group.

Those institutional backers now hold substantial positions, and the ability to monetize them in the IPO itself has become a key negotiating point.

A secondary offering component, allowing insiders to sell alongside the primary offering that raises new capital for Anthropic, is standard in many mature tech IPOs, but Anthropic’s scale and the concentration of its early backers make the mechanics unusually material.

SpaceX’s June IPO Set Precedent by Blocking All Insider Sales at Debut

SpaceX’s June listing established a radically different model. Founder Elon Musk and the company’s venture backers conducted no secondary offering whatsoever. The underwriters, led by major investment banks, priced 555.5 million newly issued shares at $135 each, raising roughly $75 billion directly for the company.

The underwriting syndicate then took another 83.3 million shares, bringing total listing volume to 638.9 million shares and approximately $86 billion in proceeds, still the largest IPO in history by dollar value. Every cent went to SpaceX’s balance sheet.

That structure allowed Musk to retain roughly 82.4% of voting power after the listing, safeguarding his control through supervoting shares, a two-tier equity class designed to preserve founder authority despite dilution from the public float. Anthropic has adopted the same supervoting mechanism for its founders, according to recent reports, ensuring similar control preservation.

The difference lies in liquidity for existing shareholders.

The SpaceX approach locked out all early investors from selling on day one. Venture capitalists, sovereign wealth funds and employees held their positions for the standard 180-day lockup period, during which they could not trade shares. On August 6, roughly 911.5 million insider shares, worth approximately $116 billion at then-current valuations, became freely tradeable all at once.

That represented a flood of supply more than double the 638.9 million shares sold in the original offering, causing the liquid float to expand from 4.9% of the company to 11.8% overnight. The stock still gained 6.1% that day, suggesting the market absorbed the supply, but the concentration and timing created acute risk.

Anthropic’s Secondary Component Offers Earlier Cash But Requires Longer Lockups

Anthropic appears to be choosing the opposite path on secondary sales. Rather than forcing all insiders to hold until the standard 180-day window expires, the company is planning to let select early shareholders and employees sell shares directly in the IPO offering itself. That sale would be priced upfront with specific buyers already identified, creating certainty around valuation and volume.

Insiders who opt into this secondary component would receive cash on IPO settlement day.

To offset the supply risk of allowing insider sales at debut, Anthropic is reportedly weighing lockup periods that exceed the standard 180-day norm. The mechanics would work as follows: insiders who sell in the secondary offering receive cash immediately but then face extended restrictions on selling any remaining shares.

Those who do not participate in the secondary would face the conventional lockup timeline. This two-tier approach aims to smooth the supply curve: some insider selling happens immediately but at a controlled, priced volume; the remainder remains locked for longer, preventing a second wave of supply shock like the one SpaceX experienced on August 6.

The institutional investors backing Anthropic, firms like Altimeter, Dragoneer, Greenoaks and Sequoia, are large enough and sophisticated enough to manage both outcomes. But the timing matters. Secondary offerings allow funds to reduce concentrated positions and rebalance portfolios.

A longer lockup forces them to hold longer, tying up capital in a single name and creating duration risk. The prospectus filing after Labor Day will spell out the exact terms, including lockup periods and which insiders are permitted to sell at the IPO.

Prospectus Filing Will Clarify Lockup Terms and Secondary Volume

The prospectus must be filed and reviewed by the SEC before trading can begin. Current timing suggests the document will arrive shortly after September 7, Labor Day, with a likely IPO date in late September or early October.

That filing will disclose Anthropic’s financial performance, risk factors, capital structure and the specific mechanics of its lockup policy and secondary offering eligibility. For institutional investors, the prospectus is the authoritative source, not market rumor or preliminary reports.

Key questions remain open until that filing arrives. How many shares will the secondary component permit existing holders to sell? Will the supervoting structure allow all shareholders equal treatment, or will founders and certain early investors face different restrictions?

What is the exact length of the extended lockup periods, and does it apply uniformly or in tranches? Will the company provide guidance on ongoing capital allocation, profitability, and competitive positioning against OpenAI, Google and other AI leaders? The prospectus answers these questions and establishes the baseline from which the stock will be valued on day one.

Anthropic’s valuation in May at $965 billion already reflects heavy investor confidence. A September IPO at or above that level would make the company one of the largest software debuts in history. But the secondary offering component means more shares will be issued than would occur in a pure primary offering, diluting existing stakeholders unless the company prices accordingly.

The market will price all of this information into the opening quote.

The prospectus filing after Labor Day will be the definitive moment. It will specify the lockup timeline, the volume of secondary shares, the exact supervoting structure, and whether Anthropic management provides forward guidance on revenue, profitability or runway. Institutional investors should plan to file detailed comments during any SEC review period and to model both a bullish scenario (where the secondary is limited and lockups are merely extended) and a bearish scenario (where secondary volume is large and near-term supply remains elevated). The company’s own statements about lockup policy post-filing will signal management confidence in the stock’s ability to absorb insider selling without sharp depreciation.

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

Amazon Caught Destroying Rare Books to Train AI
Institutional

Amazon Caught Destroying Rare Books to Train AI

, , , , , ,

AI NewsAugust 17, 2026·5 min read

Amazon’s systematic destruction of rare books at a Nevada warehouse facility after scanning them for AI training data reveals the scale and legal vulnerability of the tech industry’s intellectual property sourcing practices. For institutional investors evaluating AI infrastructure spending and regulatory risk, this precedent-dependent model creates material exposure to future copyright reform and supply-chain disruption.

  • Amazon workers at Nevada facility VGT3 strip book spines and scan pages before systematic destruction as routine job function.
  • 1,000 rare books ordered per shipment through anonymous Biblio marketplace channels, targeting out-of-print titles with few remaining copies in circulation.
  • 2023 federal court ruling established that book destruction after digital conversion constitutes fair use, creating legal foundation for industry-wide scaling.
  • 1,000 Rare books destroyed per shipment at Nevada warehouse facility for AI model training
  • 2023 Year federal court ruled book destruction after scanning qualifies as fair use under copyright law
  • VGT3 Amazon warehouse facility designation where systematic spine stripping and page scanning operations occur

Amazon has established a systematic operation at its Nevada warehouse complex to purchase rare and out-of-print books through anonymous marketplace channels, scan their contents for artificial intelligence training datasets, and destroy the physical originals.

The discovery emerged from investigative tracking of a 1,000-volume shipment ordered through Biblio, a marketplace specializing in used and rare titles, which was transported to the facility designated VGT3 within Amazon’s Las Vegas operations.

Workers at the site perform routine tasks of stripping book spines and scanning pages as assigned job functions with formal production metrics, indicating this is a scaled manufacturing process rather than a pilot program.

The operation reflects an industry-wide trend that accelerated following a 2023 federal court ruling that deemed the practice of converting physical books to digital form and then destroying the originals to constitute “transformative” fair use under copyright law.

Amazon’s Nevada Warehouse Operates Rare Book Destruction as Formalized Production Process

The VGT3 facility operates with structured workflows and assigned production rates, with workers performing sequential tasks that move books from intake through destruction. Some staff scan ISBN barcodes upon arrival, while others cut book spines to enable rapid page-by-page scanning before the materials are disposed of.

The warehouse’s operational division logo depicts a Tyrannosaurus rex consuming a book, an image that literalizes the destruction occurring on the facility floor in stark visual terms. Workers describe their assignments in straightforward operational language, suggesting institutional normalization of the process across the workforce.

The targeting of rare titles with few remaining copies in circulation indicates systematic curation rather than random bulk acquisition.

Anonymous purchasing through commercial resale marketplaces obscures Amazon’s identity as the end buyer, allowing the company to acquire materials without triggering seller awareness or public disclosure. The 1,000-book-per-shipment volume, combined with ISBN barcode cataloging, suggests an effort to digitize entire categories of the global printed book inventory.

This scale of operation requires sustained capital expenditure, logistics coordination, and warehouse labor allocation integrated into Amazon’s standard operational structure, pointing to a deliberate long-term investment in this sourcing channel for training data.

2023 Court Ruling Created Legal Foundation for Industry-Wide Book Scanning and Destruction

A federal court decision in 2023, emerging from litigation involving AI training company Anthropic, established that transforming physical books to digital copies and then destroying the originals does not violate copyright law if the transformation serves a different purpose from the original work.

The ruling classified this practice as “transformative” under fair use doctrine, creating explicit legal cover for technology companies to source copyrighted material through acquisition and destruction cycles.

This precedent directly enabled Amazon and other technology firms to scale book destruction operations with reduced litigation risk, since the copyright holder’s original economic interest in the physical book is eliminated through destruction, and the digital reproduction serves a distinct AI training purpose.

The ruling’s language around “transformative” use redefined how courts evaluate derivative works created from copyrighted material, shifting the analysis away from whether the original copyright holder was compensated and toward whether the new use creates different value.

For Amazon and similar operators, this interpretation meant they could legally purchase books through standard commercial channels, extract their contents, and destroy the physical originals without licensing agreements or payments to authors or publishers.

The decision reflects an evolving legal framework in which copyright protection becomes subordinate to technology company data sourcing strategies, provided the original physical work is rendered unusable.

No compensation to copyright holders occurs under this model, since the destruction of the original physical book eliminates any competing economic claim to the digitized content.

Institutional Investors Face Emerging Regulatory Risk as Copyright Reform Becomes Foreseeable

The systematic destruction of rare books to train AI models creates a specific regulatory vulnerability for technology companies and their investors: copyright law reform targeting this exact practice is now foreseeable.

Authors, publishers, and preservation advocacy groups have limited legal recourse under the 2023 ruling, but legislative response at federal and state levels remains entirely open.

Several countries, including the European Union member states, have moved to restrict or prohibit AI training on copyrighted material without explicit licensing, creating jurisdiction-specific compliance challenges for technology companies operating globally.

For institutional investors in AI infrastructure companies, this represents material downside risk to the capital efficiency of training data acquisition models.

If the 2023 court ruling is overturned or superseded by statute, technology companies would face sudden increases in licensing costs for copyrighted material, reduced access to certain rare and out-of-print materials (since destruction would become prohibited), and potential litigation exposure for prior book destruction operations.

The valuation premium currently assigned to technology companies based on low-cost training data acquisition could compress if legal precedent shifts. Investors should evaluate whether individual holdings have diversified training data sourcing strategies that do not depend entirely on the continued legality of book destruction under fair use doctrine.

Amazon’s warehouse operation creates direct legal exposure to copyright reform, since it produces documented evidence of systematic destruction of rare cultural materials.

ISBN Cataloging and 1,000-Volume Shipments Suggest Global Inventory Targeting Strategy

The use of ISBN barcode scanning at the Nevada facility indicates systematic effort to catalog and track specific book titles being destroyed, rather than random bulk processing.

Combined with the 1,000-volume-per-shipment volume through anonymous marketplace channels, this suggests Amazon may be executing a targeted acquisition strategy focused on specific categories of rare or out-of-print material. The anonymity of marketplace purchases prevents sellers and preservation organizations from identifying patterns in what material is being removed from circulation.

For institutional investors, this operational structure raises questions about the long-term availability and pricing of rare intellectual property.

If Amazon and competing technology companies systematically acquire and destroy rare books at scale, librarians, researchers, and collectors may face reduced availability of certain out-of-print titles and increasing scarcity premiums for surviving physical copies.

This creates secondary effects on academic institutions, heritage preservation organizations, and collectors who depend on stable access to rare materials. Technology companies’ training data sourcing practices may directly reshape information access markets outside the AI industry.

Investors should monitor pending copyright legislation at federal and state levels, particularly any bills restricting AI training on copyrighted material without licensing, as well as statements from author advocacy groups and publishing industry representatives signaling intent to challenge the 2023 ruling through legislative remedy or future litigation.

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

Google Cloud eyes enterprise distribution in $100M deal with self-improving AI builder Mirendil
Institutional

Google Cloud eyes enterprise distribution in $100M deal with self-improving AI builder Mirendil

, , , , , ,

AI NewsAugust 6, 2026·5 min read

Mirendil, an AI lab founded by former Anthropic researchers, has secured a multi-year Google Cloud partnership valued at over $100 million to power its self-improving AI systems. For institutional investors and enterprise customers, the deal signals both a validation of recursive self-improvement as a near-term commercial technology and Google’s strategic pivot toward infrastructure plays that lock in hardware adoption through software partnerships.

  • Mirendil raised $200 million at $1 billion valuation in June; Google deal represents roughly 50% of seed capital committed to single infrastructure agreement
  • Partnership grants access to Google TPUs and Nvidia GPUs with managed training clusters, emphasizing hardware flexibility and workload optimization across chips
  • Self-improving AI already showing measurable commercial traction: Anthropic reports 80% of code merged in May 2026 written by Claude models, up from single digits in early 2025
  • $100M+ Google Cloud commitment as percentage of Mirendil’s entire seed funding round
  • 80% Anthropic code merged by AI models as of May 2026, versus prior single-digit baseline
  • $1B Mirendil valuation at seed stage, one of largest AI startup seed rounds on record

Mirendil, the AI research lab founded by former Anthropic researchers, has committed to a multi-year partnership with Google Cloud worth more than $100 million, securing computational infrastructure for its self-improving AI models.

CEO Behnam Neyshabur confirmed the deal, which provides access to Google’s Tensor Processing Units (TPUs) and Nvidia graphics processors alongside managed training clusters.

The agreement represents a substantial portion of the company’s recently closed seed funding and underscores growing institutional confidence in recursive self-improvement as a near-term technology capable of automating research workflows across medicine, biology, chemistry, and materials science.

Mirendil commits half its seed capital to single Google Cloud infrastructure deal

The $100 million Google commitment amounts to roughly 50 percent of Mirendil’s $200 million seed round, which closed in late June at a $1 billion valuation. That round, led by Andreessen Horowitz with participation from Kleiner Perkins and Nvidia, represented one of the largest seed financings for an AI startup to date.

By allocating such a substantial portion of early capital to a single infrastructure partner, Mirendil’s founders are signaling both the computational intensity of their research agenda and confidence in Google’s ability to deliver the horsepower required for recursive self-improvement at scale.

The concentration of spend reflects the economics of frontier AI development, where training and inference costs dominate operational budgets. For Mirendil, the Google arrangement locks in multi-year access to heterogeneous compute, both TPUs and Nvidia GPUs, at a time when capacity remains constrained and pricing volatile.

Cofounder Harsh Mehta emphasized that the company’s models are architected to distribute workloads efficiently across different chip types, lowering costs not only for Mirendil’s internal research but for enterprise customers licensing its systems.

Self-improving AI moves from research concept to measurable productivity metric

Recursive self-improvement refers to AI systems that iteratively refine and redeploy themselves, becoming progressively more capable without constant human retraining or intervention. Mirendil’s stated ambition is to build models capable of performing the full range of work conducted by a frontier research laboratory, from hypothesis generation through experimentation and iteration.

CEO Neyshabur, a former Anthropic research scientist and co-inventor of the Sharpness-Aware Minimization optimization algorithm, articulated the vision plainly: “You can have a self-improving AI where you can point a problem at it, and it keeps getting better with time.” His example was Alzheimer’s research, a domain where self-improving systems could autonomously pursue open-ended scientific goals without constant human guidance.

Anthropic has already demonstrated measurable traction with this model. As of May 2026, more than 80 percent of code merged into Anthropic’s systems was written by its Claude models, compared to single-digit percentages in early 2025.

The company acknowledged that this metric likely overstates true productivity gains, yet confirmed that the acceleration is genuine: engineers are now merging approximately eight times as much code per day as they were a year prior.

This data point carries weight for institutional investors evaluating whether self-improvement remains theoretical or has entered the realm of practical enterprise deployment.

Mirendil’s cofounder Harsh Mehta contributed to Anthropic’s internal automation efforts, giving the company direct operational experience with self-improving workflows. That institutional knowledge, combined with access to Anthropic’s research legacy, positions Mirendil to compress the timeline from prototype to production deployment.

The Google partnership accelerates that path by removing infrastructure barriers.

Google positions self-improving AI as hardware lock-in and enterprise upsell opportunity

Google’s $100 million commitment is not purely philanthropic. For the cloud giant, Mirendil represents a strategic customer whose software layer can drive adoption of Google’s own silicon. Neyshabur noted that Mirendil’s workload optimization capabilities help customers extract more performance from Google’s chips, providing Google with a competitive pitch against rival cloud providers.

In a market where raw chip performance is increasingly commoditized, software that can orchestrate heterogeneous compute more efficiently becomes a differentiator worth funding directly.

Amin Vahdat, Google’s senior vice president and chief technologist for AI and infrastructure, framed the partnership in systemic terms. Progress in AI infrastructure, he indicated, is no longer driven solely by chip-level performance but by the ability to orchestrate entire systems, balancing compute, memory, networking, and scheduling across complex workloads.

Mirendil’s software architecture, tested against both TPUs and Nvidia GPUs, offers Google a vehicle to demonstrate that thesis to enterprise customers. When those customers adopt Mirendil’s models, they simultaneously adopt Google Cloud as the optimal deployment platform.

The deal also grants Google early exposure to recursive self-improvement at scale. Should Mirendil’s research succeed, Google gains both a strategic partner and a template for understanding how self-improving systems operate within its cloud infrastructure.

That knowledge compounds in value as self-improving AI moves from research labs into commercial deployments across pharmaceuticals, materials science, and research institutions.

Institutional investors face new metrics for evaluating self-improving AI runway and unit economics

The Mirendil-Google deal establishes a practical pricing baseline for self-improving AI infrastructure partnerships. At $100 million over multiple years, the deal implies annualized commitment in the tens of millions of dollars for a company still in early-stage product development.

Investors evaluating AI infrastructure plays must now assess whether early-stage AI firms can sustain multi-year infrastructure commitments while simultaneously building enterprise sales organizations and navigating regulatory scrutiny.

Anthropic’s disclosed metrics on code generation and engineer productivity suggest that self-improving systems do accelerate output measurably, even as the company cautioned against overstating gains.

For institutional investors, the critical question is whether Mirendil’s models will achieve similar or superior productivity metrics in domains like drug discovery or materials science, where economic value per unit of acceleration is orders of magnitude higher than software development.

A self-improving system that accelerates therapeutic discovery by even 10 percent could justify substantially larger infrastructure spending.

The heterogeneous compute architecture, spanning both Google TPUs and Nvidia GPUs, also signals a broader industry shift. Rather than betting on a single chip manufacturer, Mirendil’s design assumes a multi-vendor landscape and optimizes for portability.

This approach reduces lock-in risk for enterprise customers and suggests that the AI infrastructure market is moving toward abstraction layers and scheduler-level optimization, not raw silicon dominance.

Watch for Mirendil’s first disclosed customer deployments and performance benchmarks in pharmaceuticals or materials science over the next 12-18 months. Those metrics will determine whether Mirendil’s self-improving architecture translates competitive advantage in research workflows into enterprise revenue sufficient to justify its $1 billion valuation and the implied $50+ million annual spend on Google infrastructure. Google’s willingness to expand or renew the commitment at higher levels would signal internal validation of recursive self-improvement’s commercial viability; conversely, any public shift to alternative compute partners would suggest the technology faces unforeseen deployment challenges.

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

U.S. officials accuse China’s Moonshot of stealing from Anthropic’s Fable in Kimi K3 production
AI News

U.S. officials accuse China’s Moonshot of stealing from Anthropic’s Fable in Kimi K3 production

, , , , , ,

AI NewsJuly 22, 2026·5 min read

The White House has accused Chinese AI startup Moonshot of systematically stealing Anthropic’s proprietary Fable model to develop its new Kimi K3 system, marking an unprecedented escalation in U.S.-China AI competition that could reshape how regulators police cross-border model development. The accusation centers on large-scale model distillation routed through obscured channels and powered by sanctioned NVIDIA chips, raising questions about enforcement of export controls and the enforceability of intellectual property claims in AI.

  • Michael Kratsios, White House Office of Science and Technology Policy director, publicly accused Moonshot of covertly copying Anthropic’s Fable model for Kimi K3.
  • Anthropic tracked 3.4 million Claude conversations routed through hundreds of fabricated accounts to Moonshot between January and February 2025.
  • Moonshot’s Kimi K3 reached 2.8 trillion parameters, the largest open-weight model released to date, with full weights scheduled for public release July 27.
  • 3.4M Claude conversations Anthropic traced to Moonshot accounts versus prior undisclosed baseline activity.
  • 2.8T parameters in Kimi K3, exceeding all prior open-weight model releases by scale.
  • July 27 deadline for Kimi K3 full model weights public release per Moonshot announcement.

On Wednesday, Michael Kratsios, the White House’s chief technology policy official, posted a detailed allegation on X accusing Moonshot AI of running an industrial-scale operation to extract training data from Anthropic’s Claude model and use it to build Kimi K3, a new large language model set for broader public deployment this week.

Kratsios stated that his office had obtained intelligence showing Moonshot constructed a purpose-built internal platform to conduct what he termed “large-scale, covert industrial distillation” against U.S. models, cycling through multiple access routes to mask the activity from detection.

The accusation centers on model distillation, a technique in which outputs from a stronger, more capable AI model are fed into a weaker one to train it to replicate the stronger model’s behavior.

The timing of Kratsios’s statement is significant: Moonshot released Kimi K3 to limited users on July 16 and suspended new sign-ups within two days due to overwhelming demand.

The company announced that full model weights would enter the public domain by July 27, a date that has now become the focal point of a potential regulatory standoff over whether the model can be freely distributed or whether the U.S. government will attempt to block its release.

Anthropic’s February discovery of 3.4 million fabricated accounts accessing Claude

The White House accusation builds directly on a report Anthropic published in February 2025, in which the company said it had tracked more than 3.4 million conversations involving its Claude model back to Moonshot through hundreds of accounts that appeared to be falsely created.

Anthropic stated that some of the account profiles matched publicly listed senior staff members at Moonshot, suggesting the activity was coordinated and not the result of routine user testing or competitive research.

Anthropic said the campaign systematically probed Claude’s known strengths in reasoning, coding, and vision tasks, the three categories most relevant to building a competitive general-purpose language model.

The scale and targeting of the data collection indicated to Anthropic’s investigators that the activity was designed to extract enough information to allow Moonshot engineers to train a model that could replicate Claude’s capabilities, rather than simply auditing Claude’s safety features or benchmarking its performance.

Model distillation itself is not inherently illegal. Researchers and companies routinely use outputs from public-facing models to train smaller, more efficient versions; this is considered a standard engineering practice when conducted at small scale and without concealment.

The distinction Kratsios draws is between that legitimate practice and what he characterizes as covert, industrial-scale theft of proprietary technology using deliberate obfuscation and sanctioned hardware.

Moonshot’s use of Thailand-based NVIDIA GB300 servers to evade U.S. export rules

According to Kratsios’s statement, Moonshot conducted the distillation work using NVIDIA GB300 servers stationed in Thailand. The GB300 is among NVIDIA’s most advanced and powerful chips, and U.S. export control regulations bar the sale of chips of that capability tier directly to China, a restriction intended to limit China’s ability to train state-of-the-art AI models independently.

By routing the work through Thailand, a country with fewer export control restrictions on advanced semiconductors, Moonshot would have circumvented the intent of the U.S. restrictions while remaining in a technical gray zone of legal violation.

This detail carries significant implications for U.S. export policy. Over the past three years, the Commerce Department has repeatedly tightened restrictions on advanced chip sales to China, citing national security concerns.

The use of third-country infrastructure to work around those rules, if substantiated, would suggest that export controls alone are insufficient to protect sensitive U.S. AI research and that enforcement requires international cooperation or secondary sanctions on companies that help circumvent the primary restrictions.

The accusation also raises questions about the vulnerability of cloud-based AI services to state-sponsored or well-resourced corporate espionage. Anthropic’s ability to detect the activity through account patterns and conversation logs is noteworthy, but it also underscores that large-scale extraction may be possible to hide in plain sight if the attacker has sufficient resources and planning.

Kimi K3’s July 27 release deadline and competing claims over safety and control

Moonshot announced that Kimi K3’s full model weights will be released publicly by July 27, just five days after Kratsios’s accusation. The model stands at 2.8 trillion parameters, making it the largest open-weight model ever released, surpassing prior releases like Meta’s Llama series.

Open-weight models, where the underlying trained parameters are published and freely downloadable, are harder to contain or regulate once released, since anyone with sufficient computing resources can run, modify, or fine-tune the model.

Anthropic has signaled that it views the public release of Kimi K3 as a material risk, arguing that models derived from copied training data lose the safety guardrails and alignment work that the original developers embedded.

Once weights are public, the company stated, they can be modified, combined with malicious training data, or deployed in uncontrolled environments beyond the reach of any policy or governance framework.

The White House has not explicitly announced enforcement action to block the release, but Kratsios’s public accusation on Wednesday appears designed to establish a political and legal record for potential future intervention.

U.S. regulators have limited direct authority over Moonshot, a Beijing-based company, but they could theoretically press NVIDIA or other U.S. technology providers to deny Moonshot access to infrastructure, or could coordinate with allies to impose restrictions on companies that distribute or build upon Kimi K3 weights.

The critical open question is whether the U.S. government will attempt to block Moonshot’s July 27 release or take secondary measures against companies that use or distribute the model. Anthropic has made clear its view that the release poses a safety and security risk, and the White House accusation suggests the government is preparing a rationale for intervention, but no explicit regulatory action has been announced. The outcome will test whether the U.S. can enforce IP claims and export policy in AI development across jurisdictions, or whether open-source distribution has become effectively beyond government control once code leaves U.S. soil.

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

Elon Musk’s AI Startup Is a Complete Disaster Behind the Scenes
Blockchain

Elon Musk’s AI Startup Is a Complete Disaster Behind the Scenes

, , , , , ,

AI NewsJuly 16, 2026·6 min read

Elon Musk’s xAI, recently merged into SpaceX ahead of a planned public offering, is hemorrhaging talent and operational coherence as internal dysfunction undermines its ability to compete with established rivals like Anthropic. Institutional investors evaluating SpaceX’s IPO prospectus will need to assess whether xAI’s technical and leadership challenges represent a material risk to the parent company’s valuation and near-term profitability.

  • All eleven of xAI’s cofounders departed the company following its February merger with SpaceX ahead of the planned IPO.
  • Michael Nicolls, promoted to lead xAI after the SpaceX merger, explicitly set “matching Claude’s performance” as the startup’s near-term objective in internal memos.
  • xAI sold computing capacity to rival Anthropic while attempting to develop competing AI systems, signaling resource constraints and operational desperation.
  • 11 xAI cofounders departed following merger with SpaceX and before IPO filing
  • 30% Planned workforce reduction in March, executed with minimal HR coordination or notice
  • February Timeline when xAI-SpaceX merger began triggering mass cofounder exodus

Elon Musk’s AI startup xAI is experiencing severe operational dysfunction following its integration into SpaceX, with the merger triggering a mass departure of leadership and the onset of systemic management failures. All eleven of the company’s cofounders left following the February merger announcement and subsequent completion, according to multiple accounts of internal dynamics.

The exodus coincided with SpaceX’s preparation for a historic IPO, leaving xAI stripped of its original technical and strategic vision at a critical moment when it should have been scaling operations and securing market position against competitors.

The operational impact has been immediate and measurable. In March, xAI announced plans to reduce its workforce by up to 30 percent, but executed the cuts with such poor coordination that affected employees learned of their termination through informal channels rather than official notification, according to Bloomberg reporting.

The human resources department, already understaffed, could not process paperwork fast enough to manage either the layoffs or new hiring simultaneously. Promising candidates who advanced through interview rounds received no follow-up communication for weeks, creating a reputation problem for recruitment at a moment when xAI needed to retain and attract top engineering talent.

Musk’s Personal Obsession with Claude Redefines xAI’s Strategic Roadmap

Michael Nicolls, promoted to president of xAI after the SpaceX merger, made clear in internal communications that the startup’s near-term mission centered on matching Anthropic’s Claude rather than pursuing independent innovation.

In a memo to staff after assuming leadership, Nicolls wrote: “Our near-term goals are to match the performance of Claude,” according to reporting on the company’s internal direction.

This framing represented a significant strategic pivot, rather than defining xAI’s own competitive advantages, the company positioned itself as reactive, with every Claude product update triggering demands from Musk that Grok should deliver equivalent functionality.

The Claude-fixation pervaded xAI’s technical culture and infrastructure. Internal Slack channels were named after Anthropic’s flagship product. Project planning documents regularly referenced Claude’s architecture and capabilities.

Engineers reported that keeping pace with Anthropic’s release schedule had become an explicit performance metric, a departure from typical AI development where companies set their own research agendas and timeline.

Some staff began to question whether Musk, whose prior successes centered on automotive and aerospace engineering, possessed sufficient domain expertise to lead an AI research organization where the technical landscape differs fundamentally from hardware engineering.

That skepticism deepened as Musk pursued one of his most ambitious technical projects: Macrohard, an AI agent system designed to emulate the capabilities of an entire software engineering company. Experts have expressed significant doubts about whether such a system is even theoretically feasible at present.

Employees familiar with large language model development flagged concerns that Musk’s previous success in other domains had not translated into understanding the specific constraints and possibilities of contemporary AI systems.

Cofounder Exodus Triggered by SpaceX Merger Leaves xAI Leaderless

The February timing of xAI’s announced merger with SpaceX coincided precisely with when the company’s original founding team began resigning. Every cofounder ultimately departed, representing a vote of no confidence in the post-merger strategic direction.

The departures were not announced as a coordinated statement but occurred individually over several weeks as the IPO process accelerated, suggesting the cofounders reached independent conclusions that their continued involvement was untenable.

The cofounder exodus reflects both a breakdown in governance and a fundamental misalignment between Musk’s vision for the merged entity and the technical judgment of xAI’s original leadership team. Mergers of AI-focused subsidiaries into larger industrial companies have historically created tensions between research autonomy and corporate efficiency pressures.

In this case, the integration into SpaceX, a company operating under IPO-related disclosure and governance requirements, appears to have shifted decision-making authority toward Musk’s personal preferences and away from the technical judgment that had guided xAI’s founding strategy.

Grok’s Nonconsensual Image Generation and Unfulfilled Pay-for-Data Scheme

Additional controversies eroded internal morale and raised questions about the company’s technical governance. In early 2025, Grok was discovered to have generated millions of nonconsensual intimate images of real people, including minors, using xAI’s infrastructure. The incident troubled employees and raised serious questions about content safety systems and model training protocols.

The scale of the problem, millions of images, suggested that content filtering mechanisms were either absent or fundamentally ineffective.

Separately, xAI promised employees cash compensation in exchange for providing personal tax returns to use as training data for Grok. The company never fulfilled the payment commitment, creating a secondary trust problem on top of the reputational damage from the nonconsensual image generation scandal.

These governance failures, inadequate content filtering and broken promises to staff, compound the technical and strategic uncertainties that already characterized the organization.

xAI’s Sale of Computing Capacity to Anthropic Signals Resource Desperation

In a striking reversal of its competitive positioning, xAI sold computing capacity to Anthropic, the very company it was attempting to emulate and outpace. The transaction occurred as the SpaceX IPO timeline tightened and xAI’s internal dysfunction deepened.

For investors evaluating SpaceX’s overall business profile, the sale of GPU capacity to a direct AI competitor raises a simple but uncomfortable question: why would a company pursuing a serious competitive position against Anthropic monetize its scarcest strategic resource by selling it to that rival?

The transaction reflects a broader resource constraint. All major AI laboratories, Anthropic, OpenAI, Anthropic and others, are experiencing acute shortages of GPU compute.

Training and operating large language models requires sustained access to high-end semiconductor capacity. xAI’s decision to sell rather than retain this capacity suggests either that the company had accumulated excess inventory relative to its near-term needs, or that it required immediate liquidity to cover operational costs amid the post-merger disruption.

Neither scenario inspires confidence in the venture’s financial durability or technical strategy.

IPO Prospectus Will Determine How Deeply Markets Assess xAI’s Viability

SpaceX’s pending IPO filing will require detailed disclosure of xAI’s financial performance, technical roadmap, and competitive positioning.

Institutional investors evaluating the broader SpaceX investment case will need to determine whether xAI represents a material business unit, a research-stage venture with uncertain return prospects, or a distraction that diverts capital and management attention from SpaceX’s core aerospace operations.

The current dysfunction, mass cofounder departures, 30 percent workforce cuts, inability to hire, reputational damage from content moderation failures, and computing asset sales to competitors, will be difficult to obscure in SEC filings or investor presentations. Whether SpaceX’s IPO disclosure addresses xAI’s operational status directly or buries it in consolidated financial tables may itself become a marker of management confidence in the unit’s viability. Institutional crypto and AI investors will be watching whether SpaceX’s underwriters and counsel permit detailed xAI disclosure, and if so, how the market prices SpaceX equity given the

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

Why’s Terawulf stock surging today?
Equities

Why’s Terawulf stock surging today?

, , , , , ,

AI NewsJuly 6, 2026·5 min read

TeraWulf locked in $19 billion in contracted revenue over two decades from AI infrastructure leader Anthropic, fundamentally shifting its business model from volatile Bitcoin mining to predictable, long-term AI data center leasing. The deal validates the company’s infrastructure pivot while addressing institutional investor concerns about revenue stability and competitive positioning in the high-margin AI compute sector.

  • TeraWulf signed a 20-year lease with Anthropic valued at approximately $19 billion in total revenue
  • NASDAQ:WULF shares surged 17% to $24.14, reflecting 117% year-to-date gains since January 2025
  • First Anthropic services begin H2 2027; full 401-megawatt capacity operational by early 2028 at Hawesville facility
  • $19B Total contracted revenue from Anthropic over the full 20-year lease term
  • 117% Year-to-date stock appreciation since the start of 2025 through announcement day
  • 401MW Critical IT capacity at Hawesville campus when fully operational in early 2028

TeraWulf announced Monday that it has secured a two-decade lease agreement with Anthropic, the maker of the Claude family of large language models, to operate a purpose-built AI infrastructure campus at its Justified Data site in Hawesville, Kentucky.

The deal is expected to generate approximately $19 billion in total revenue for TeraWulf and represents a watershed moment for the company’s transformation from Bitcoin mining into the higher-margin, more stable AI data center market. The announcement drove TeraWulf’s stock up 17 percent to $24.14 by late Monday morning, extending its year-to-date rally to 117 percent.

Anthropic Campus Provides $19 Billion Revenue Foundation Over Two Decades

The Hawesville facility will be developed in stages, with Anthropic’s first services commencing in the second half of 2027 and the full 401 megawatts of critical IT capacity becoming operational in early 2028. That capacity level positions the site to support large-scale AI model training workloads, addressing the intense computational demands of frontier large language models.

For institutional investors monitoring data center supply and demand dynamics, the deal demonstrates that Anthropic, now valued at roughly $40 billion in private markets, is willing to commit to long-term, fixed-capacity arrangements at scale.

The contract’s 20-year term is the critical element here. Bitcoin mining, TeraWulf’s original business, generates volatile returns tied directly to cryptocurrency prices and mining difficulty, making revenue projections unreliable.

By contrast, a fixed lease with a major AI model developer locks in predictable cash flows across two decades, substantially reducing business risk and improving the company’s ability to service debt and fund future expansion.

Institutional equity analysts have flagged revenue visibility as a key valuation concern for data center operators; this deal substantially addresses that concern for TeraWulf.

The choice of Anthropic as an anchor tenant carries additional weight. The company competes directly with OpenAI and other large language model developers and has secured commitments from Google, Salesforce, and other major enterprise customers.

By betting on Anthropic’s continued prominence in AI over the next 20 years, TeraWulf is implicitly wagering on a specific vendor’s survival and competitiveness, a real risk, but one mitigated by Anthropic’s current market position and funding depth.

TeraWulf Exits Texas Joint Venture to Control Full Ownership and Operations

On the same day, TeraWulf announced the sale of its 50.1 percent stake in an AI data center joint venture located in Abernathy, Texas, to an investor group led by Fluidstack, its co-investor in the project. While the company did not disclose the transaction’s financial terms, it stated that it realized a premium over the approximately $450 million it had invested in the venture.

The sale represents a deliberate shift in capital deployment strategy, prioritizing sites where TeraWulf maintains complete operational control and direct customer relationships over shared ownership structures.

This move reveals management’s conviction that owning infrastructure outright, rather than sharing it with joint venture partners, is the superior long-term position. Joint ventures introduce governance friction, complicate customer negotiations, and limit pricing power.

By exiting Abernathy and redploying the sale proceeds into wholly owned platforms, TeraWulf gains the flexibility to negotiate customer terms more aggressively and avoid disputes with partners over capacity allocation or pricing.

For institutional investors, it signals management clarity about the competitive dynamics of AI infrastructure and confidence in TeraWulf’s ability to build owned assets at scale.

The pairing of the two transactions reveals a coherent capital strategy. TeraWulf is simultaneously locking in long-term, predictable revenue (the Anthropic lease) while shedding partnerships that dilute ownership and control.

In the company’s own language, the moves “enhance TeraWulf’s long-term revenue visibility, strengthen its financial position, and further align the Company’s capital with infrastructure platforms where it maintains direct ownership, customer relationships, and operational control.”

Bitcoin Mining Revenues Overtaken by AI Data Center Transition in Q1 2025

The Anthropic deal caps a dramatic business model pivot. In Q1 2025, TeraWulf generated $21 million in revenue from high-performance computing hosting services while deriving roughly $13 million from Bitcoin mining operations, the first quarter in which AI data center revenues exceeded mining revenues.

The company’s original footprint included Bitcoin mining facilities in New York and Pennsylvania; those assets are now being repurposed or complemented by AI infrastructure builds designed to accommodate far larger, longer-term customers with different technical requirements.

The transition has carried steep near-term costs. In Q1 2025, TeraWulf reported net losses of $427.63 million as it invested heavily in buildout and development. The company ended the quarter with approximately $3.1 billion in cash and an equivalent amount in long-term debt, creating a leveraged position that makes long-term revenue certainty essential to financial sustainability.

Anthropic’s contract directly addresses that imperative by providing decades of predictable cash inflow.

TeraWulf is not alone in this transition. Numerous Bitcoin mining operators, including Core Scientific, Hut 8, and others, have pivoted toward offering power and space to AI firms as a way to stabilize revenues and escape the cyclicality of cryptocurrency price swings.

Data Center Demand Surge Underpins AI Infrastructure Expansion Across Industry

TeraWulf’s ability to attract Anthropic and other AI customers reflects a structural undersupply of high-performance data center capacity. The International Energy Agency has projected that data center electricity consumption will nearly double from current levels to approximately 945 terawatt-hours annually, driven primarily by AI model training and inference workloads.

That growth trajectory supports the economics of long-term leasing arrangements and provides confidence that demand will not suddenly evaporate.

The scarcity of available compute capacity is acute. Major AI labs including OpenAI, Google DeepMind, Meta, and Anthropic have all publicly signaled that infrastructure availability is a constraint on their ability to scale research and commercialize new models.

For TeraWulf, the timing of the Anthropic deal also reflects favorable financing conditions. The company is less dependent on volatile equity markets to fund expansion if it can use contracted revenues to secure debt financing.

Banks and institutional debt investors are more willing to lend against predictable, long-term lease revenues than against speculative mining operations, further reducing TeraWulf’s cost of capital and competitive disadvantage relative to larger data center operators like Equinix or CoreWeave.

Phased Delivery Timeline Extends Revenue Recognition Through 2028 and Beyond

The Hawesville facility’s staged rollout, first services in H2 2027 and full capacity by early 2028, means TeraWulf will begin recognizing Anthropic revenues in roughly two years. That timeline provides critical breathing room for the company to manage its leverage and reduce reported losses as AI services revenue ramps.

It also creates quarterly guidance visibility for analyst consensus, a key metric that institutional equity investors use to assess management’s execution risk.

The phased approach

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

Enterprises pull back on AI spending as token bills outpace productivity gains
Blockchain

Enterprises pull back on AI spending as token bills outpace productivity gains

, , , , , ,

AI NewsJune 26, 2026·6 min read

Enterprise AI spending is contracting sharply as chief financial officers demand measurable returns on API bills that have outpaced actual productivity gains, threatening growth projections at OpenAI and Anthropic that depend on expanding token consumption. The reversal marks the end of what the industry calls “tokenmaxxing”, the practice of maximizing AI tool usage regardless of business value, and signals a fundamental shift in how large organizations evaluate artificial intelligence investments.

  • Amazon dismantled an internal AI usage leaderboard after determining it generated busywork rather than productive output, with leadership explicitly warning staff against using AI without clear justification.
  • Uber depleted its entire 2026 AI coding budget in four months and now enforces a $1,500 monthly spending cap per employee on AI tools after uncontrolled consumption threatened budgets.
  • OpenAI and Anthropic face a structural threat as enterprises shift to cheaper models and open-source alternatives, with IDC forecasting 70% of leading AI enterprises will use multiple models by 2028.
  • $1,500 Monthly spending cap per Uber employee on AI coding tools, down from unlimited consumption
  • 70% Of leading AI enterprises predicted to use multiple models by 2028, versus single-provider reliance today
  • 4 months Time required for Uber to exhaust entire 2026 AI coding budget before implementing controls

The pivot away from aggressive AI adoption represents one of the most consequential reversals in enterprise technology spending this year. After months of mandatory AI tool rollouts and aggressive adoption campaigns, major corporations are now systematically auditing their usage and cutting back on the expensive APIs that power large language models.

The shift reflects a hard reckoning between the theoretical productivity gains promised by AI vendors and the actual business impact observed in practice.

CFOs at major organizations are no longer accepting rising token bills as a cost of doing business and are instead demanding evidence that AI investments are actually completing work, saving time, or reducing rework, not simply consuming more computational resources.

Amazon Dismantles AI Leaderboard After Finding System Encouraged Busywork Over Results

Amazon’s decision to dismantle an internal leaderboard tracking employee AI usage crystallizes the problem facing enterprises. The company’s leadership concluded that the system, designed to encourage adoption, was instead driving workers to use AI tools on trivial tasks simply to increase their standing on the leaderboard.

An Amazon senior vice president told staff bluntly: “Please don’t use AI just for the sake of using AI.” This message directly contradicts the adoption-at-all-costs messaging that dominated corporate AI strategy throughout 2024 and early 2025.

Amazon’s experience is not isolated; it reflects a pattern emerging across the enterprise sector.

Uber provides a more dramatic example of uncontrolled AI spending spiraling into crisis. The company burned through its entire 2026 AI coding budget within four months before implementing emergency controls.

In response, Uber imposed a $1,500 monthly spending cap per employee on AI coding tools, a hard constraint that would have seemed impossible just months earlier when unlimited AI adoption was viewed as competitive necessity.

Meta likewise sent an internal memo to approximately 6,000 employees warning of an “exponential increase” in AI usage and flagging that the company faced billions of dollars in internal AI infrastructure costs.

Accenture Reverses AI Adoption Stance as Spending Becomes “Unpredictable” and Unquantified

The contradiction at Accenture underscores how rapidly corporate positions on AI are shifting. The consulting giant previously told employees they could “risk losing out on promotions” if they failed to adopt AI tools, an aggressive mandate designed to drive maximum usage. Within months, that pressure has inverted.

Leaked audio from an internal meeting captured an Accenture executive acknowledging that AI spending is “becoming very unpredictable” and that leadership across finance, operations, and IT levels continues to ask whether the company is “getting value from what we’re spending.”

This confession reveals the core problem: enterprises adopted AI spending without establishing clear metrics for success. CFOs are discovering that high token consumption does not correlate with business outcomes, and in many cases, correlates with waste. The spending is unpredictable because usage patterns are disconnected from measurable work completion.

IBM’s analysis, authored by Adam McDaniel and Markus Eisele, argues that both “tokenmaxxing” and its opposite, excessive cost-cutting focused on minimizing tokens regardless of utility, represent failures of governance. IBM advocates instead for what it terms “valuemaxxing,” which measures completed tasks, time saved, and rework avoided rather than tokens consumed.

This framework reorients corporate AI spending away from consumption metrics and toward business outcomes, which is precisely what CFOs are now demanding to see.

OpenAI and Anthropic Face Threat as Enterprises Shift to Cheaper Models and Self-Hosted Infrastructure

The spending contraction threatens the financial models of the companies that benefited most from the tokenmaxxing boom. OpenAI crossed $25 billion in annualized revenue earlier this year while placing its valuation at $1 trillion.

Anthropic is valued several billion dollars lower but operates under the same assumption: that enterprise customers will continuously increase their token consumption as adoption deepens. Both companies are burning substantial cash on compute infrastructure, research, and hiring while betting that profitable growth will eventually follow from market expansion.

That bet is now under pressure.

Enterprises are already responding to cost pressure by fragmenting their AI vendor relationships. Instead of committing to a single flagship model from OpenAI or Anthropic for all workloads, companies are now reserving expensive, capable models for complex work requiring state-of-the-art performance and routing routine tasks to cheaper alternatives.

Some organizations are deploying open-source models on their own infrastructure, eliminating per-token charges entirely and replacing them with capital expenditure on compute hardware. The International Data Corporation projects that by 2028, 70% of leading AI-driven enterprises will operate with multiple AI models across their operations rather than relying on a single provider.

This structural shift from single-vendor dependency to multi-model strategies undermines the growth projections that OpenAI and Anthropic have built into their business plans. A company consuming tokens across five different models and platforms will almost certainly consume fewer tokens from any single vendor than a company locked into one.

Distributed adoption also means that smaller, more efficient models, whether from competitors or open-source communities, will capture increasing share of workloads that large language models currently handle at significant cost.

CFO Scrutiny Forces Recognition That AI Adoption Metrics Must Connect to Business Value

The underlying shift is structural and unlikely to reverse quickly. CFOs who approved unlimited AI spending during the hype cycle are now facing direct accountability for whether those expenditures produced measurable returns. Board-level scrutiny of AI spending is intensifying precisely as the novelty of the technology is wearing off.

Companies that adopted AI aggressively without establishing ROI frameworks are now in the position of having to justify continued spending to skeptical finance leadership.

This accountability creates a new discipline around AI deployment. Instead of asking “how can we use AI more,” enterprises are asking “what work does AI actually complete better and cheaper than our current process?” The answer to that question is narrower than the tokenmaxxing era assumed.

For routine customer service, basic code generation, and standard report writing, small models and open-source alternatives now deliver sufficient quality at a fraction of the cost of premium APIs.

For genuinely complex reasoning, specialized domain work, and novel problem-solving, flagship models still command premium pricing, but those workloads represent a smaller percentage of total enterprise AI usage than executives initially believed.

The contraction is already observable in API consumption trends among large enterprises.

Accenture’s leaked acknowledgment that spending remains “unpredictable” and that senior leadership cannot yet articulate clear value reflects the reality that many enterprises lack the governance infrastructure to manage AI costs effectively.

Some organizations are implementing formal approval workflows for new AI tool deployments, requiring business case documentation and ROI projections before approving API spending. Others are moving to fixed budgets for AI tools rather than allowing uncapped consumption.

These mechanisms will reduce total token consumption because they force teams to justify spending against actual business requirements rather than adopting AI tools speculatively.

The critical

Get this in your inboxThe Crypto Coin Show newsletter covers the policy and market moves institutional crypto investors are pricing in.

Subscribe

Load More