Anthropic has abandoned plans to acquire artificial intelligence (AI) infrastructure startup Decart AI in a deal previously pegged at about $6 billion.
The decision comes after the Claude chatbot creator performed extensive due diligence on the Israeli-founded company. While the acquisition talks have dissolved, the two firms may still pursue alternative commercial partnerships, according to Bloomberg.
“Walking away sets a ceiling on what a frontier lab will pay to own compute efficiency instead of contracting for it. Efficiency gains at the silicon and inference layer move fast enough that acquiring them outright locks in today’s constraint at tomorrow’s price,” said Mitch Ashley, vice president and practice lead for Software Lifecycle Engineering and AI-Native Software Engineering at The Futurum Group.
“Watch whether the two sign a commercial agreement instead. That is the real tell,” Ashley said. “Enterprises signing multi-year AI commitments should treat vendor cost curves as provisional and price the option to move.”
Representatives for Anthropic and Decart declined to issue official statements.
“We don’t know exactly why Anthropic walked away. But because it happened after due diligence, it’s reasonable to assume that something about the tech, economics, team, or fit did not support the price,” said Stephanie Walter, practice leader for AI Stack & Enterprise Application Development at HyperFRAME Research. “Notice that the companies may still work together. This suggests Anthropic may see value in Decart’s technology while deciding it doesn’t need to own the company.”
“To me, the interesting piece is that model companies are looking across the AI stack for ways to improve their economics,” Walter said. “Building a better model is only part of the challenge. Running it efficiently and reliably matters too.”
The multibillion-dollar deal was originally designed to enhance Anthropic’s computing efficiency as the company scales operations to meet skyrocketing enterprise demand.
Decart, which was valued at nearly $4 billion following a $300 million funding round in May backed by NVIDIA Corp., Sequoia Capital, and Benchmark, specializes in chip-optimization software. Its proprietary Decart Optimization Stack (DOS) allows AI models to run more efficiently across varied hardware platforms, including chips from NVIDIA, Amazon.com Inc., and Google.
Decart also develops advanced world models such as Oasis and Lucy, the latter capable of real-time video transformation for e-commerce, gaming, and livestreaming applications on platforms like eBay Inc., TikTok, and YouTube.
Under CEO Dario Amodei, Anthropic has historically shied away from large-scale corporate acquisitions, choosing instead to channel capital directly into raw computing infrastructure.
Ahead of a highly anticipated initial public offering, the San Francisco-based firm has been securing massive multibillion-dollar cloud computing and data-center capacity agreements — including recent commitments with Lambda, Nscale, Fluidstack, and SpaceX — as it seeks to achieve a market valuation on par with top-tier technology giants.
The breakdown of the acquisition underscores broader macroeconomic and regulatory headwinds facing high-stakes AI transactions.
Cross-border acquisitions involving critical semiconductor capabilities, advanced algorithm optimization, and foreign technology transfer face elevated oversight from international regulatory bodies. Decart’s roots in Israel and its focus on underlying chip infrastructure made the prospective transaction a target for potential national security and foreign investment reviews.
Furthermore, the failed transaction unfolds alongside ongoing friction between Anthropic and U.S. defense officials.
Although Commerce Secretary Howard Lutnick recently indicated that Anthropic had resolved previous disputes with the federal administration, military leadership maintains a cautious stance.
Anthropic previously won a court ruling against the military after Defense Secretary Pete Hegseth restricted the firm from select defense contracts over Anthropic’s refusal to allow its Claude models to be deployed for domestic surveillance or fully autonomous weaponry.
Despite the court victory, Under Secretary of Defense for Research and Engineering Emil Michael confirmed on social media that the Defense Department’s designation of Anthropic as a supply-chain risk remains active across the department and the broader defense industrial base.
As Anthropic shifts its focus back toward organic expansion, the company continues its preparations for a public listing targeted ahead of the upcoming U.S. midterm elections, prioritizing massive computational footprint expansions over direct strategic acquisitions.
