Amazon is in talks to buy the Israeli-founded AI video startup Decart for about $7 billion, the Wall Street Journal reported on Sunday 11 October, citing people familiar with the matter. Reuters relayed the report and noted that Amazon declined to comment outside regular business hours and Decart did not respond to requests for comment. The WSJ said a deal could come together soon if talks do not fall apart or another suitor emerges. That is the whole of what is known. Because the facts are thin, the more useful question is what the reported price implies, what it cannot imply, and why this particular company has attracted so many buyers in five months.
What is reported, and what is not
Three separate reports make up the story. The first is the WSJ report of the Amazon talks at about $7 billion. The second is a Bloomberg report, relayed by Reuters, that Anthropic explored a deal for Decart at around $6 billion in August and decided not to pursue it last month after due diligence. An 8 September article by The Deep Dive, summarising Bloomberg's account, said no reason had been confirmed and that neither company had explained what changed. The third is the Decart funding history: a $300 million round announced on 18 May 2026, led by Radical Ventures with Nvidia joining as a new investor, which the WSJ, citing PitchBook data, said valued the company at nearly $4 billion. Decart's total funding is reported as above $450 million.
It is worth recording that the reporting on this company has already been inconsistent once. On 18 August, an aggregator alert headlined that Anthropic was nearing a $7 billion deal for Decart, mostly paid in Anthropic shares, citing Israeli outlets including Calcalist and saying Nvidia had offered more and been passed over. Bloomberg's later reporting, relayed on 8 September, had Anthropic dropping the idea at $6 billion. The earlier claim did not come true. That is not evidence the Amazon report is wrong, since it comes from the WSJ and has been relayed by Reuters, but it is a reminder that, for private-company deal talks, 'in talks' and 'nearing' are descriptions from sources, not outcomes.
What Decart builds
Decart was founded in 2023 by Israeli engineers and builds real-time generative video models. Its best-known product, Lucy, edits live video as it streams, according to Decart's website as cited by Reuters. A second product, Oasis, generates simulated environments to train and test robotics and autonomous-driving systems. A claim reported in the August coverage is that Decart's optimisation software makes models run 'up to eight times faster on the same hardware'; that is the company's marketing language as relayed by an aggregator, and DrafterDaily has seen no independent measurement.
The distinction from clip-generation video models matters for the strategic logic. A clip generator takes a prompt and spends seconds or minutes producing a finished video. A real-time model must generate each frame fast enough to respond to input as it arrives, which makes inference speed and cost per frame the core engineering problem. That is why the same capability gets described both as a video product and as an efficiency technology. Bloomberg's August report, according to The Deep Dive, said Anthropic was interested in using Decart's technology to handle more demand on its existing infrastructure, which suggests a buyer might value the inference work as much as the video product. That is an interpretation of a secondary summary, not a confirmed motive, and the reason Anthropic walked away is not public.
The price path, in arithmetic
Set the reported numbers side by side and a pattern appears. The May round valued Decart at nearly $4 billion. The $6 billion explored by Anthropic in August would have been a premium of about 50% to that figure, as The Deep Dive also calculated. The reported $7 billion from Amazon is about 1.75 times the May valuation, reached in under five months from 18 May to 11 October, and about 17% above the $6 billion that Anthropic reportedly dropped. The $300 million raised in May equals about 7.5% of a $4 billion valuation, so an investor who bought in that round at the nearly-$4-billion level would, if the reported sale price held, be sitting on a paper gain of roughly 75% in five months. All of this is our arithmetic on reported, rounded figures, and it assumes the nearly-$4-billion figure is the post-money valuation, which the coverage does not specify.
The comparison the coverage did not draw is between the two bids. If the Anthropic approach is read as a floor and the Amazon talks as a ceiling, the range of reported buyer interest spans $6 billion to $7 billion, a spread of about $1 billion. A bid is not a valuation, though: the $6 billion figure was an exploratory deal that did not proceed and the $7 billion is a reported talk figure that may not be the final price, and any price in a negotiated acquisition can include stock, earn-outs or retention payments that make the headline number differ from what sellers receive. None of the sources describe the structure.
What we cannot calculate
No source gives Decart's revenue, users, customers or cash burn. That means no price multiple can be computed: a $7 billion price on, say, $50 million of revenue is a very different proposition from the same price on $500 million, and the coverage gives no basis to choose between them. It also means that readers who describe $7 billion as cheap or expensive are describing their own assumptions. If the buyer is paying mainly for engineering talent, model weights and compute efficiency rather than for a revenue stream, comparing the price to revenue would be the wrong test anyway, but the sources do not tell us whether that is the case here.
The other missing fact is what changed between August and September. Due diligence can end a deal for many reasons, including price, technology claims that do not verify, legal exposure, integration cost or a change in the buyer's priorities. The reporting names none of them, and a reader should treat any account that picks one as speculation. What can be said is that a second large buyer reportedly entering at a higher price within weeks suggests that whatever caused the first buyer to withdraw was not, on its face, an industry-wide view that the asset was worth less. If the WSJ report is accurate, that is.
What a closed deal, or a collapse, would change
If the deal closes at about $7 billion, Amazon would own a real-time video-model team that, on the reported figures, was valued at nearly $4 billion five months earlier. The coverage does not say what Amazon intends to do with it, and DrafterDaily will not guess. A closed deal at that price would also give investors in other generative-media companies a fresh reference point: a buyer paying roughly three-quarters more than a private round within months. Amazon's next quarterly results, expected in late October according to Vantage Markets' relay, could be the first venue for any comment, though the date is not confirmed.
If talks collapse, the effect on Decart is harder to read. A second failed process, after the Anthropic one, would raise questions about what diligence is finding, though failed talks are common and the sources say nothing about diligence on Amazon's side. Decart would still hold the money from its May round and the reported interest of at least two large buyers. For readers following this sector, the actionable point is narrow: treat the $7 billion as a reported figure in talks, wait for confirmation from either company, and distrust any commentary that treats the price as established or that offers a revenue multiple without a revenue figure.
