# SpaceX Is Closing In On A $6 Billion Deal For Israeli AI Startup Decart > Source: > Published: 2026-08-10 09:30:18+00:00 *SpaceX's reported pursuit of Decart is not just another AI deal. It's a fight over who controls the software that makes expensive chips work harder.* Decart is suddenly sitting in the middle of the AI infrastructure trade. Calcalist reported that the Tel Aviv startup is in advanced talks to sell for $6 billion to $7 billion, with SpaceX widely identified as the likely buyer. Nvidia had already backed Decart in a funding round this year. Now it may have to watch a portfolio company go to Elon Musk instead. That's awkward. It is also useful. You can learn a lot about this market from who wants Decart. The company does not make chips. It builds software and models designed to make AI workloads run faster and more cheaply, including real-time world models that need low-latency inference to work at all. If you are spending billions on GPUs, data centers and power, the software layer that squeezes more output from that hardware stops looking like a tool. It starts looking like the thing that decides who wins: control. Decart was founded in 2023 by Dean Leitersdorf and Moshe Shalev, who have been tied in earlier coverage to Israel's Unit 8200 intelligence unit. TechCrunch reported in December 2024 that Decart had raised $32 million at a valuation of more than $500 million, after a $21 million seed round less than two months earlier. Its first enterprise product focused on optimizing GPU use, and its Oasis demo showed a playable AI-generated world running in real time. That was not a normal consumer launch. It was a public proof point for the engine underneath. ## The price moved first The valuation climb is the part you should not skip over. Decart was valued at more than $500 million in late 2024, according to TechCrunch. CTech reported that it later raised $100 million at a $3.1 billion valuation in August 2025, then raised $300 million at an estimated $4 billion valuation in May 2026. Nvidia took part in that latest round, alongside Radical Ventures, Adobe Ventures, eBay Ventures, Toyota Ventures, Atreides Management and Valor Equity Partners. Now Calcalist says the number being discussed is $6 billion to $7 billion. That is a hard jump in a short time. There is a reason buyers can justify it. Decart's public material describes Oasis 3 as an interactive world model for physical AI, with real-time generated simulation environments for robotics and autonomous systems. It says the system runs on DOS, Decart's inference and training platform, and gives developers API access to generated worlds that respond to control signals. Strip out the marketing language and the point is plain: Decart is trying to make generated environments usable as infrastructure, not just as demos you watch once and share. That fits SpaceX's new shape. Business Insider reported that SpaceX's first earnings report after its June IPO showed $7.8 billion in quarterly revenue, a $541 million net loss and AI capital expenditure of $15.8 billion, above analyst forecasts. The same report said Musk projected a $100 billion annual revenue run rate by year-end. The Wall Street Journal also reported that the company's AI spending helped push the stock lower after earnings, even as revenue beat expectations. Investors like growth. They hate blank cheques. ## Why Nvidia cannot like this Nvidia still wins if Decart sells at a higher valuation than the one it backed: a return is a return. But strategic investors do not only write cheques for markups. They back companies because they want influence, early access and sometimes a path to owning the thing outright. That is where SpaceX creates a problem. If Decart can help workloads move more efficiently across hardware, or make real-time simulation cheaper to run, the company becomes valuable to anyone with a large AI compute bill. Nvidia wants the AI stack to keep pulling customers toward its chips. SpaceX wants to turn its AI infrastructure into a business big enough to justify public-market expectations. Those goals overlap. They are not the same. Look at Cursor. SpaceX agreed in June to acquire the coding startup's parent company, Anysphere, in a $60 billion all-stock deal, according to TechCrunch and CBS News. That deal came only weeks after SpaceX's public-market debut, and it showed how quickly the company is willing to use its stock to buy AI capability rather than build everything from scratch. Decart would be a different kind of purchase. Cursor gives SpaceX a product used by developers. Decart would give it infrastructure closer to the metal - the kind of software that touches model training, inference and simulation costs. That is the more revealing bet. None of this is signed. Calcalist described the talks as advanced, not closed, and a deal at this size would still have to survive legal and regulatory work. The difference between a reported $6 billion discussion and a completed acquisition is not paperwork. It is the deal. Still, the direction is clear. A two-year-old Israeli AI company went from a $500 million valuation to a possible $7 billion sale because buyers are no longer only chasing the biggest models. They are chasing the systems that make those models cheaper to run. If SpaceX closes the deal, Nvidia gets paid and SpaceX gets the asset. You can decide which side of that trade looks happier. **Also read:** [ByteDance's Zhang Yiming Bans AI Model Distillation Even as Rivals Race Ahead](https://startupfortune.com/bytedances-zhang-yiming-bans-ai-model-distillation-even-as-rivals-race-ahead/) • [TSMC's July sales jumped 45% as AI chip demand keeps outrunning tariffs](https://startupfortune.com/tsmcs-july-sales-jumped-45-as-ai-chip-demand-keeps-outrunning-tariffs/) • [DeepSeek Warns Its Ultra-Cheap V4-Flash Pricing Won't Last](https://startupfortune.com/deepseek-warns-its-ultra-cheap-v4-flash-pricing-wont-last/)