Anthropic Walks Away From Its $6 Billion Decart Acquisition After Due Diligence
Anthropic has abandoned its planned acquisition of Israeli AI startup Decart, valued at roughly $6 billion, after due diligence, per Bloomberg and Calcalist, in what would have been its largest known acquisition.
Anthropic has abandoned its planned acquisition of Israeli AI startup Decart, per Bloomberg and Calcalist, walking away after conducting due diligence on the company. The deal, first reported on August 13 at roughly $6 billion, would have been Anthropic’s largest known acquisition by a wide margin. No official reason was given beyond the completion of due diligence.
What Decart Would Have Brought, and Why It Matters That It Did Not
Decart is not a typical acqui-hire target. The company built real-time generative world models (its Oasis driving simulator ran endless generated worlds at 22 FPS) plus GPU-efficiency work, which maps directly onto two frontiers Anthropic cares about: world models for agents and inference cost. Walking away after diligence on a $6 billion target is a notable signal in both directions, and the honest read is limited: we know the process ran and the deal died, not which side got cold feet or what diligence found.
The Pre-IPO Discipline Reading
The timing gives the walk-away its most plausible frame. As reported last week, Anthropic is targeting an IPO listing days before the November midterms, with a valuation case resting on a $190-200 billion 2028 revenue forecast. Companies in that position re-scrutinize every large acquisition: an unpopular or dilutive $6 billion deal becomes a prospectus footnote that underwriters and short-sellers will interrogate, and any diligence ambiguity gets repriced by the public market. Anthropic’s last big M&A-adjacent story was hiring UK AI strategist Matt Clifford amid conflict-of-interest criticism, so the pattern this month reads as a company cleaning its story before listing. For founders, the lesson is that pre-IPO acquirers are the least reliable closers in the market, however serious the talks; the diligence bar is set by the S-1, not the term sheet.
Get Insanely Good at AI
The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.
Keep Reading
How to Fine-Tune Cosmos Predict 2.5 for Robotics With LoRA
Learn how to adapt NVIDIA's 2B and 14B Cosmos Predict 2.5 world foundation models using parameter-efficient fine-tuning methods like LoRA and DoRA.
Anthropic's IPO Targets Mid-October Marketing and a Pre-Midterm Listing
Reuters reports Anthropic is expected to begin marketing its IPO in mid-October at the earliest and list days before the November midterms, with its valuation hinging on a $190-200 billion 2028 revenue forecast.
Claude Formalized Fermat's Last Theorem in Lean After 11 Days of Autonomous Work
Anthropic announced Claude produced a complete machine-checked Lean formalization of Fermat's Last Theorem after working largely autonomously for 11 days, generating verified proofs of 30,300 intermediate theorems.
Anthropic Launches Enterprise Frontier Safeguards After Data Retention Pushback
Anthropic's new Enterprise Frontier Safeguards keep Claude activity logs in the customer's own cloud under customer encryption keys, effectively zero data retention, after business customers including Microsoft pushed back.
Anthropic Launches Fable 5.1 With 75% Cheaper Cache Reads and EU Watermarks
Anthropic released Claude Fable 5.1 and restricted Mythos 5.1 on September 1, keeping base pricing at $10/$50 but cutting cache reads 75% to $0.25 per million tokens, with invisible watermarks on every output.