A $6 billion acquisition that would have been one of the biggest in Anthropic’s history is off the table. According to Bloomberg, Anthropic completed due diligence on AI startup Decart AI and then walked away. No deal was signed, no terms were finalized, and both companies are declining to comment publicly.
Decart is not a household name, but its focus is strategically interesting. The startup builds software that makes AI chips run more efficiently, which directly reduces the cost of training and running large models. For a company like Anthropic, which burns through serious compute to develop and serve Claude, that kind of technology has obvious appeal. So why pass on it?
The most likely explanation is valuation. Six billion dollars is a steep price for an infrastructure-layer startup, especially one that has not yet proven itself at scale across multiple customers or model architectures. Anthropic may have decided that the efficiency gains Decart offers are either replicable internally or not worth the price tag given where the company is putting its capital right now. The two sides may still collaborate in some form, according to people familiar with the situation, which suggests the relationship is not entirely closed off.
The timing matters here. Anthropic is preparing for what is shaping up to be one of the most closely watched IPOs in recent memory. Reports indicate the company is targeting a valuation that rivals or exceeds SpaceX’s, which puts it in a very different financial position than it was even a year ago. Pre-IPO, every major capital decision gets scrutinized. A $6 billion acquisition of an unproven startup would raise questions from prospective public investors about capital discipline. Walking away may actually be the smarter move for a company trying to present a clean story to Wall Street.
It’s also worth placing this in the context of how Anthropic operates. Unlike OpenAI, which has been more aggressive about partnerships and adjacent bets, or Google DeepMind, which has the parent company’s resources behind it, Anthropic has historically been cautious about large acquisitions. Its spending has gone toward compute infrastructure and internal research, not roll-up deals. This episode fits that pattern.
For Decart, the situation is more complicated. Being valued at $6 billion in acquisition talks and then not getting acquired is a double-edged outcome. It validates the technology but leaves the company in an awkward spot with investors and potential acquirers who will now wonder what Anthropic found during due diligence that made them step back. That question will follow the startup into any future fundraising or acquisition conversations.
The broader trend here is that compute efficiency is becoming a serious competitive axis in AI. As model training costs climb and inference at scale gets expensive, startups that can make chips do more with less are genuinely valuable. Microsoft, Google, and Amazon all have internal efforts in this direction, and several well-funded startups are competing in the same space. Decart will need to find another path to scale, whether that is a different acquirer, a strategic partnership, or going it alone with additional venture backing.
For Anthropic, the more pressing story remains the IPO. If that goes as planned, the company will have far more flexibility to make acquisitions or infrastructure bets on its own terms. The Decart deal may simply have been the wrong size at the wrong time.
