Bloomberg reports Anthropic completed due diligence on Israeli AI startup Decart and chose not to proceed — a capital-allocation decision that says more about the frontier lab’s pre-IPO priorities than any deal it has signed.
Decart Valuation Ladder · Key Numbers
~$6B
Discussed acquisition size (not signed); ~1.5× premium to last private mark
According to Bloomberg’s September 8 report — sourced to people familiar with the matter, with both Anthropic and Decart declining to comment, and the full piece paywalled — Anthropic completed due diligence on Israeli AI startup Decart and chose not to proceed with an acquisition. No agreement was signed. The discussed size, carried over from the original mid-August report that the two were in talks, was approximately $6 billion. Bloomberg notes the two companies could still pursue other forms of collaboration, leaving the relationship explicitly open.
The walk-away, not the original talks, is the news. The mid-August report that Anthropic and Decart were in discussions was a preliminary signal; this reverses it. A company that completes due diligence has seen the books, the technology, the team, and the cap table — and then decided the price of entry at roughly 1.5× Decart’s most recent private mark was not the right use of its capital. Because both parties declined to comment, everything here rests on Bloomberg’s sourced reporting and should be attributed accordingly throughout.
Decart is worth understanding beyond Bloomberg’s compressed description of “software that makes chips run more efficiently.” Founded in 2023, the company is also a real-time world-model developer — its products Lucy and Oasis represent a distinct capability tier above pure inference-optimization software. Its valuation trajectory has been steep: $500 million in late 2024, $3.1 billion by August 2025, and $4 billion on a $300 million round led by Radical Ventures with NVIDIA participating. The ~$6 billion figure Anthropic reportedly walked away from represents a meaningful premium to that already-elevated mark.
Deal Timeline
Late 2024
Decart valued at $500M — early-stage AI startup with real-time world-model ambitions
August 2025
Decart reaches $3.1B valuation; momentum building in inference-efficiency and world-model space
Bloomberg first reports Anthropic and Decart in acquisition talks; ~$6B discussed size
8 September 2026
Bloomberg reports Anthropic completed due diligence and walked away — no agreement signed; collaboration door left open
The key insight: Anthropic didn’t walk away before looking — it walked away after. Completing due diligence on a ~$6B asset and declining to proceed is a deliberate capital-allocation judgment, not an inconclusive conversation. At a moment when Anthropic is spending heavily on compute and weeks from a listing reportedly marketed toward a multi-trillion-dollar valuation, that judgment tells you exactly what the lab believes its marginal dollar should buy.
The Structural Read
Start with what this is not. Bloomberg draws no connection between the walk-away and any IPO trouble. Manufacturing a distress narrative here would be editorializing past the reporting. The honest frame is simpler and more interesting: a frontier lab with a rare M&A appetite looked hard at a multibillion-dollar acquisition and decided its capital was better deployed elsewhere. That is capital discipline, and it is a concrete signal in an environment where AI acquisition prices have compressed rationalanalysis out of most deals.
The first structural lens is compute-versus-acquisition as a capital-allocation choice. Anthropic is in the middle of its most aggressive infrastructure investment cycle — GPU clusters, data-center commitments, the raw compute on which its model roadmap depends. Against that backdrop, ~$6 billion directed at an inference-efficiency and world-model vendor is ~$6 billion not spent on owned inference capacity. The company chose the latter. That preference ordering is the data point: when forced to choose between buying an efficiency layer and owning more of the infrastructure that layer would optimize, Anthropic kept the cash for the infrastructure.
The second lens is build-versus-buy on the efficiency layer itself. Decart’s pitch — software that extracts more performance from the same silicon, plus real-time world models — is precisely the capability a frontier lab might want to internalize. Choosing not to, after a full diligence process, is a read on how Anthropic values that layer at current prices: either the integration complexity outweighed the synergies, the technology was replicable internally, or the company prefers to partner rather than absorb. None of those readings implies the technology is weak; all of them imply the price was not compelling enough to justify the organizational absorption cost at this particular moment.
Business Engineer · Build-vs-Buy Read
The Efficiency Layer Is Valued — Just Not at 1.5× Premium
Frontier labs have every incentive to want inference-efficiency software inside the tent. Anthropic’s pass is not a verdict on the category — it is a price signal. At $4B in private markets with NVIDIA already in the cap table, Decart has substantial institutional validation. At ~$6B to a buyer weeks from an IPO, the acquisition economics require a very specific bet on irreplaceable technology or defensible moat. Diligence apparently produced enough uncertainty on one or both counts to make keeping the cash the better trade.
The third lens is what this does to Decart’s position. A public walk-away by a would-be acquirer after completed diligence is a durable market signal. Future counterparties — strategic buyers, late-stage investors repricing their positions, potential partners — now hold this data point. That does not mean Decart is in difficulty; NVIDIA’s participation in its last round and the collaboration door Bloomberg reports as still open are meaningful indicators of standing. But private valuations are partly constructed from the credibility of acquisition interest, and a documented declined bid at 1.5× the last mark is a reset input, not a neutral one.
Three Implications
IMPLICATION 1 · Anthropic’s M&A Posture Is a Pre-IPO Capital Signal
Frontier labs rarely do large M&A. When one that is weeks from a public listing completes diligence on a ~$6B asset and declines, it is communicating a preference ordering to future investors: owned compute over acquired capabilities. That ordering will be read by bankers, analysts, and institutional allocators as they price the listing. It is not a conservative signal — it is a specific one about where management believes value accrues.
IMPLICATION 2 · The Efficiency Layer’s Buy Price Has a Ceiling
Inference-efficiency software and real-time world models are genuinely valuable — NVIDIA’s participation in Decart’s last round confirms that. But Anthropic’s walk-away suggests that ceiling is below the 1.5× private-mark premium being discussed. For the broader category of AI infrastructure vendors pitching acquisition value to frontier labs, this sets a pricing reference point: strategic value is acknowledged, but integration risk and build-or-partner alternatives compress the multiple a buyer will actually pay.
IMPLICATION 3 · Decart’s Private Valuation Carries a New Reference Point
A completed diligence process that ends in a walk-away is public information now. Decart has NVIDIA on its cap table, a collaboration door still open with Anthropic per Bloomberg, and real products in market — none of that disappears. But every future conversation about Decart’s valuation will be conducted with this data point in the room. Private marks are partly social constructions, and a documented declined bid at ~$6B from a strategic buyer is a permanent recalibration input for subsequent discussions.
Business Engineer Framework
The Map of AI Redrawn — Where Efficiency Layers Sit in the Stack
The Anthropic–Decart episode maps directly onto the AI stack’s infrastructure layer: the zone between raw silicon (NVIDIA, AMD) and frontier model developers where inference-efficiency software and world models compete for acquisition value. The Map of AI framework traces exactly where those battles are being won, lost, and repriced — and why labs keep choosing compute ownership over capability acquisition at this stage of the cycle.
According to Bloomberg’s September 8 report — single-source, paywalled, both companies silent — Anthropic looked at everything Decart had to offer at a ~$6 billion price and decided its capital was more valuable pointed at compute than at an inference-efficiency and world-model acquisition. That is not a collapse, not a distress signal, and not an indictment of Decart’s technology. It is a precise, documented preference ordering from a lab at the exact moment its spending, its balance sheet, and its approaching public listing all converge — and preference orderings, especially ones revealed under diligence pressure, are among the rarest and most useful reads in this market. This article is not investment advice.
This is businessanalysis, not investment advice. The report is Bloomberg’s, sourced to people familiar with both Anthropic and Decart declining to comment; the full article is paywalled. The ~$6 billion is a discussed size carried from August talks, not a signed price, and Bloomberg draws no connection between the decision and Anthropic’s IPO — framing it as distress would go beyond the reporting. Decart is a real-time world-model developer as well as a chip-efficiency software company.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
