Anthropic PreStocks (ANTHROPIC) on Solana

Palantir CEO Alex Karp publicly criticized the token-based pricing models used by both Anthropic and OpenAI during a July 2 appearance on CNBC's "Squawk Box," saying "something has gone completely wrong." Karp framed his remarks as observation rather than attack, arguing that enterprises increasingly view per-token billing as wasteful — characterizing the prevailing sentiment as paying to "get no value" while surrendering intellectual property to model providers.

The criticism echoes broader enterprise pushback against closed AI infrastructure costs. Microsoft CEO Satya Nadella has similarly warned against "tokenmaxxing" for non-frontier use cases, and venture capitalist Chamath Palihapitiya highlighted the contrast between a roughly $105,000 monthly bill for GPT-5.5 Pro versus approximately $2,740 for DeepSeek R1. The pressure on token-based revenue models poses a competitive challenge for Anthropic, whose enterprise API pricing follows the same structure Karp is targeting, as companies shift toward open-weight alternatives or proprietary stacks.

Anthropic is in early talks with Samsung Electronics to manufacture a custom AI chip, joining rivals OpenAI and Google in efforts to reduce dependence on Nvidia, which controls roughly 74% of the global AI chip market. The company currently rents compute from Amazon (Trainium), Google (TPUs), and Nvidia, and the soaring cost of running its largest models has prompted it to explore proprietary silicon. Samsung — which participated in Anthropic's $65 billion funding round in May and operates advanced foundry facilities including a two-nanometer process line — is a leading candidate, with Microsoft and UK startup Fractile also under evaluation. Key details including chip function, server integration, and specifications have not yet been determined.

To build internal hardware expertise, Anthropic has hired Clive Chan, formerly of OpenAI's chip design team. The move follows OpenAI's unveiling of its custom inference chip "Jalapeno" (developed with Broadcom) and positions Anthropic as the latest frontier AI lab seeking to vertically integrate on the compute stack. For holders of Anthropic PreStocks (ANTHROPIC) on Solana, the development signals a longer-term strategic bet on hardware independence that could reduce operational costs and strengthen Anthropic's competitive position in the race to train and deploy ever-larger models.

The Trump administration's Department of Commerce has lifted the export controls it imposed on Anthropic's Claude Fable 5 and Mythos 5 models in mid-June 2026, restoring global access after roughly two and a half weeks of restricted availability. Commerce Secretary Howard Lutnick said the government had "worked closely" with Anthropic to analyze and approve Fable 5, framing the resolution as a step toward strengthening U.S. leadership in AI. Anthropic will redeploy Fable 5 with new classifiers designed to detect and block cybersecurity, biology, and chemistry queries, part of broader commitments to monitor for malicious activities. The administration reserved the right to reimpose restrictions if those commitments are violated.

Anthropic said global access to Fable 5 via Claude.ai, the Claude platform, and Claude Code will resume this week, while Mythos 5 access for select U.S. organizations had already been partially restored following a June 26 government approval. The controls had originally barred even non-citizen Anthropic employees from accessing the models, creating significant operational disruption. With the full lift now in place, Anthropic's flagship models return to the global market under an enhanced safety framework negotiated with the federal government.

Amazon has pushed back on a report from The Information suggesting that a restructuring of its payment arrangement with Anthropic would increase its AI costs. According to the report, Amazon and Anthropic transitioned from computing-hour-based pricing to token-based pricing for access to Anthropic's Claude models, effective next year. Amazon disputed the cost implication directly, stating that "it is incorrect that changes from our expanded collaboration will increase our costs" and describing the partnership as grounded in ongoing technical collaboration.

The denial comes against the backdrop of a deepening financial relationship: Amazon invested $4 billion in Anthropic in 2023 and earlier in 2026 committed up to an additional $25 billion. The restructuring report also noted Amazon is evaluating OpenAI models and its own Nova models as potential alternatives following Anthropic price increases — context that lends weight to Amazon's incentive to clarify the cost picture publicly. For Anthropic's pre-IPO market, the exchange reinforces that the Amazon relationship remains intact even as both parties adjust commercial terms at scale.

STARTRADER launched a CFD instrument ticker ANTHUSD on June 29, 2026, giving retail and institutional clients a way to take positions on Anthropic ahead of any public listing. The product trades 24/7 across the broker's MetaTrader, STAR-APP, and STAR-COPY platforms with up to 5x leverage, requiring no access to private placement rounds or accredited-investor thresholds.

CEO Peter Karsten framed the launch as a direct response to demand for early AI exposure: "AI is shaping the next generation of global industries, and traders want the ability to access these opportunities early." ANTHUSD is a CFD rather than actual equity, meaning traders gain price exposure without holding shares, and is listed alongside a parallel OpenAI instrument (OPENAIUSD). The move widens the pool of participants who can express a directional view on Anthropic's private-market valuation before a traditional IPO occurs.

The Trump administration has authorized Anthropic to make its Mythos 5 AI model available to more than 100 specific U.S. government agencies and companies, including access for non-American employees at those organizations. Commerce Secretary Howard Lutnick directed the approval to Anthropic chief compute officer Tom Brown, stating that "appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model." The authorization comes roughly two weeks after Anthropic suspended access to Mythos 5 and Fable 5 in response to a government export control directive citing national security concerns, which was itself triggered by reports that security researchers had bypassed the models' safety guardrails.

The partial restoration targets organizations operating critical infrastructure and marks a significant development in ongoing negotiations between Anthropic and the Trump administration. Notably, Fable 5 — a version of Mythos with enhanced protections — was not addressed in the directive and remains unavailable. Anthropic's broader legal dispute with the administration, stemming from a Department of Defense determination that labeled the company a supply chain risk, is still in litigation.

Anthropic is aggressively expanding its global data center footprint, with hiring activity pointing to Australia and Japan as the next frontier. Of the 13 open compute roles the company is currently recruiting for — focused on developing and managing AI data centers — eight are based in the Asia-Pacific region: six Australia-based engineer and operator roles, and two Japan roles covering data center deals and electrical engineering. The move follows a wave of U.S. data center commitments in spring 2026 and an April push to hire compute capacity negotiators for Europe.

The infrastructure ambition behind the hiring is significant. Anthropic has signed more than a dozen non-binding letters of intent with U.S. data center developers representing over 1GW of IT capacity, according to reporting by The Information, with Google reportedly arranging financial guarantees for lease payments. The company already has commitments spanning AWS, CoreWeave, Fluidstack, Akamai, and a reported $1.25 billion monthly lease for SpaceX's Colossus data centers. Having filed confidentially for an IPO and raised $65 billion in a Series H round, Anthropic is positioning its own infrastructure control as a long-term cost reduction strategy rather than relying solely on third-party cloud providers.