Anthropic Firms Up Pricing, Chips and Data Centers Ahead of a Possible IPO

A narrow Claude Code fix with a telling cause

Anthropic's own release notes for Claude Code 2.1.227 describe a bug in which feature flags were evaluated without checking the user's subscription tier specifically when a session started from an expired login token. The stated practical effect: Max plan subscribers, who already have full access, could be wrongly prompted to enable usage credits for Fable. Anthropic attributes the fault to that specific session-start condition rather than to the entitlement logic generally, and the fix targets exactly that path.

The bug is minor, but the release note incidentally confirms how Fable access is currently gated inside Claude Code — behind a subscription-tier check performed at session start, one that expired-token sessions had been bypassing.

Reports of an infrastructure buildout ahead of a possible IPO

Several outlets reported, on the same day, a cluster of Anthropic infrastructure and pricing moves. finance.biggo.com reported Anthropic locking in Sonnet 5 pricing and enabling Claude Code's Auto Mode by default, explicitly framing both as steps taken ahead of an IPO. Memeburn reported plans for an in-house custom silicon chip, and CoStar (with a French-language report from Le Revenu) described a new data-center development platform launched jointly with Macquarie Asset Management and GIC. Separately, Yahoo Finance France reported a $9.1 billion cloud agreement between Anthropic and Riot.

None of these are Anthropic's own announcements in the provided record — they are press reports, and the IPO framing in particular is the outlets' own characterization rather than a confirmed Anthropic statement. Read together, though, the reported moves — price stability, in-house chip development, dedicated data-center capacity, and a large multi-year cloud commitment — point in one direction: Anthropic reducing its dependence on third-party compute pricing and supply, consistent with the financial groundwork the reports attach to a potential public listing.

Sources: Anthropic Locks in Sonnet 5 Pricing, Enables Claude Code Auto Mode by Default Ahead of IPO — finance.biggo.com · Anthropic Will Launch an In-house Custom Silicon Chip — Memeburn · News | Anthropic, Macquarie and GIC launch data center development platform — CoStar · Anthropic signe un accord cloud de 9,1 milliards $ avec Riot — Yahoo Finance France

Opus 5 under developer scrutiny

SitePoint published a developer-focused benchmark analysis of Claude Opus 5 the same day, examining how Anthropic's current flagship performs on coding-oriented tasks. Set against the pricing and infrastructure reporting above, it places Anthropic's commercial positioning — a locked-in Sonnet 5 price and a default-on Auto Mode in Claude Code — alongside an independent, concrete look at what developers are actually getting for that price.

Sources: Claude Opus 5: Performance Benchmarks for Developers — SitePoint

OpenAI moves on two fronts the same day

Context for Anthropic's positioning: OpenAI made two sets of moves on 2026-08-10. On the model side, it launched GPT-5.6-Cyber, a model tuned for cybersecurity tasks with reduced refusals, which VentureBeat reported completing 95% of advanced cybersecurity tasks in testing; the-decoder.com's coverage frames the launch as aimed at helping defenders find vulnerabilities before attackers do, and OpenAI paired it with an expansion of its Daybreak program into separate Red and Blue access tiers. On the financial side, CNBC and Bloomberg reported OpenAI completed a $7 billion tender offer for employee shares at an implied $852 billion valuation, described in that coverage as a step ahead of a potential IPO, alongside a new premium business pricing tier reported by Yahoo Finance.

The cybersecurity model launch lands in the same conversation Anthropic has engaged with around dual-use AI security tooling, though OpenAI's own framing emphasizes defensive use. Combined with the Anthropic infrastructure reports above, the day's coverage shows both companies making pricing and capacity moves that outlets connect to IPO timing, even though neither company has, per the available record, confirmed an IPO directly.

Sources: As AI-led attacks multiply, OpenAI launches a new cyber model — TechCrunch · OpenAI launches GPT-5.6-Cyber with reduced refusals, 95% completion on advanced cybersecurity tasks — VentureBeat · OpenAI wraps $7 billion share sale ahead of potential IPO — CNBC · OpenAI Buys Back $7 Billion of Employee Shares in Tender Offer — Bloomberg.com