Product2 publishers3 min readPublished
What Anthropic wanted from Decart was cheaper serving, not a new product line, and the walk-away leaves the buy-versus-license question on compute efficiency sitting exactly where operators keep finding it.
The Product Desk · Product desk

Compiled by The Product DeskSomething wrong?How this is made
An efficiency layer is one of the few acquisitions where the buyer can check the arithmetic before signing. You know your compute bill and your utilisation rate, and a vendor promising to "squeeze every ounce of performance from every chip" [6] is making a claim that shows up in a line of your own accounts. Anthropic ran that check to the end of diligence and passed [1].
What ownership adds, on top of the saving, is exclusivity and control of the roadmap. A supply agreement gets the saving on its own, for a fraction of the money. When the acquired thing never touches a customer, the gap between those two prices is mostly what you are paying to stop someone else buying the same discount. Decart's stack sits under both training and inference and was meant to let Anthropic's existing compute absorb more demand [5], which per Bloomberg's reporting made this a deal about the cost of serving customers rather than entry into a new product category [7].
Anthropic's most notable acquisition to date was $400m for a team of fewer than ten people [8]. Against that, around $6bn [2] is roughly fifteen times its largest deal so far [21], and about $2bn over the valuation Decart's May round set [22], a mark-up of roughly half in four months [13]. A company close to a listing wants a cost base that looks efficient and an acquisition history that looks disciplined, and buying Decart would have improved the first while complicating the second [15]. Reporting does not say which of those considerations ended the talks, or whether the sticking point was price or something diligence turned up [14].
Gizmodo, which read the original Bloomberg story as the kind of anonymously sourced piece that might exist to help one side in a negotiation [20], supplies the right discount on the $6bn figure: someone with a use for it put it in circulation. Both accounts agree on the mechanism. The Decart Optimization Stack tunes the interaction between AI hardware and software to cut compute cost [19], and no user of the product ever sees it.
What separates ownership from a contract in a case like this comes down to whether most of the benefit is available by agreement, and whether a customer ever touches the thing. Decart's own two halves land on opposite sides of that line: Lucy takes live video of a person and generates high-resolution video of them wearing something they are not, with eBay as both investor and customer [17], while the optimisation stack is a cost line [19]. Anthropic wanted the invisible half [5]. The version of this decision that survives a board meeting starts with the licence quote and treats the gap up to the purchase price as the cost of exclusivity. Paying an equity multiple for a discount only makes sense if the buyer can name the rival who would otherwise have licensed the same stack.
Ranked by verification strength, evidence, and original report placement.
Anthropic performed due diligence on Decart but ultimately walked away from a purchase, Bloomberg reported, citing people familiar with the matter who asked not to be identified.
The Anthropic-Decart talks were first reported in August at around $6bn, with the caveat that nothing had been finalised and the negotiations could fall apart.
One person familiar with the matter said Anthropic and Decart may still pursue other opportunities to collaborate.
Decart's optimisation stack is designed to make chips work harder across both training and inference, and the idea, per Bloomberg, was to let Anthropic's existing compute absorb more demand.
Decart's own website promises to "squeeze every ounce of performance from every chip".
The deal was about the cost of serving customers, not about Anthropic entering a new product category.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Single anonymous chain, two retellings
One set of unnamed Bloomberg sources carries the diligence, the walk-away, the roughly $6bn figure and the still-open collaboration. Both companies declined to comment, so nothing is on the record. The only number with an independent origin is the close-to-$4bn May valuation the Wall Street Journal reported, and the mark-up arithmetic The Next Web builds on it is sound arithmetic performed on leaked inputs.
Diligence, then nothing
The transaction stopped at diligence, and no agreement replaced it, so nothing was adopted. The only named usage sits on the other side of Decart's business, with eBay as investor and try-on customer and the chief executive's July account of live-streaming use. For the optimisation stack that reportedly drew Anthropic in, neither outlet names a customer, a deployment or a gain figure.
Settled tone on unnamed sourcing
Both headlines report a finished decision, and the $6bn both hang it on has never been confirmed by anyone speaking on the record. Gizmodo at least turns the incentive problem on its own earlier coverage. The Next Web is careful at the point where it counts, saying outright that the reporting does not establish whether price or diligence killed the deal, which keeps the overreach small rather than absent.
A leak beside a listing
News of a live acquisition talk weeks before Anthropic's expected offering is useful to whoever is still at the table, a point Gizmodo made about the August version and repeats here. Decart's own promotional copy does duty as the evidence for what its stack achieves. With both companies silent, the only voices in the story are people with something riding on how it reads.
Outcome confirmed, reason unresolved
That the deal is off is not seriously in doubt: two outlets, one origin, no denial from either company, and a Gizmodo writer who predicted this outcome a month ago. What remains unresolved is everything an operator or investor would act on, namely why the talks ended, whether $6bn was ever an agreed number, and what the efficiency layer measurably delivers.
invest
Decart's founders traded a higher Nvidia bid for Anthropic paper that never printed1 publisher
product
Nvidia shows up three times in Anthropic's $35bn deal with Lambda4 publishers
product
Cursor's case for being bought is unit economics, and the post-close note promises nothing else4 publishers
build
SpaceX went shopping for a second AI coding company five days after closing Cursor2 publishers
Publishers with included, body-backed reporting in this cluster.
1 article · September 8, 2026
1 article · September 7, 2026