Decart secures $300 million to develop a real-time world model for Amazon's chips.
Radical Ventures spearheaded the funding round, with support from Nvidia, Sequoia, Benchmark, Adobe, and Toyota. Notable angel investors include Andrej Karpathy, Michael Eisner, and the Nintendo family. The total amount raised now exceeds $450 million.
On Monday, Decart, the AI research lab developing real-time video and world models, revealed that it had secured $300 million in fresh funding led by Radical Ventures. This funding round elevates the two-year-old company's total capital raised to over $450 million, with new investors such as Nvidia, Atreides Management, Valor Equity Partners, Adobe Ventures, Toyota Ventures, and eBay Ventures joining returning supporters Sequoia Capital, Benchmark, and Zeev Ventures.
The angel investor lineup indicates Decart's market positioning. Alongside Andrej Karpathy, the OpenAI co-founder and former head of AI at Tesla, the cap table includes ex-Disney CEO Michael Eisner, the Nintendo family, and gaming investor Moritz Baier-Lentz.
The latest insights from the EU tech ecosystem come from our wise founder Boris, highlighting an interesting mix of EU technology updates, including some questionable AI-generated art. This newsletter is available for free every week.
Decart's offerings diverge from pure software engineering, blending media, gaming, and infrastructure to align with the demographics that are utilizing its products.
The company presents three primary offerings. DOS, or the Decart Optimization Stack, is an inference and training platform that purportedly operates on Nvidia GPUs, Google TPUs, and Amazon Trainium, delivering agentic inference at 1,600 tokens per second, significantly higher than the industry average of around 200, and supporting full-HD video inference at up to 100 frames per second.
Lucy serves as its ‘world model for immersive experiences,’ reacting to user inputs in under 30 milliseconds and currently deployed in virtual try-on, live streaming, and dynamic in-video advertising.
Oasis acts as the parallel product for physical AI, aimed at robotics and autonomous systems since its initial real-time Minecraft-style demonstration went viral in October 2024.
The partnership with Amazon is a significant commercial detail highlighted in the announcement. Decart asserts it is among the first to implement real-time AI models of this kind and scale on AWS Trainium, with its Lucy2 model operating on Trainium3.
Nafea Bshara, vice president of Amazon’s Annapurna Labs, stated that Lucy2 achieves over 80% Model FLOPS Utilization, indicating that a larger portion of the chip's raw power is effectively engaged in productive tasks.
Dean Leitersdorf, Decart’s CEO, described world models as essential for transitioning AI from virtual settings to the real world, contending that language models primarily function in text and lack an understanding of physical-world behavior.
Decart's funding history clarifies its growth trajectory. The company completed a $32 million Series A with a $500 million valuation in December 2024, four months after raising $21 million in seed funding; Fortune reported in August 2025 that Decart had secured $100 million at a $3.1 billion valuation.
The latest funding round brings overall fundraising to over $450 million. Leitersdorf and co-founder Moshe Shalev have been developing the company since 2023; their recent appearance on the Sequoia podcast provides a clear public statement of their belief that vertically integrated optimization, rather than merely larger models, represents the missing component in the real-time AI stack.
The investor composition reflects a familiar Nvidia-equity trend. As noted by TNW, the chip manufacturer has allocated over $40 billion in AI equity in 2026 alone, with most smaller investments following a pattern where Nvidia acquires a stake, the company commits to a long-term GPU arrangement, and some GPU revenue returns to Nvidia as returns on equity.
By asserting that its DOS stack operates across all three major chip families, Decart presents a more stringent evaluation of this pattern: equity-wise, it is considered an Nvidia portfolio asset, while also acting as a public reference for Amazon Trainium and Google TPU in terms of customer engagement.
The identified target markets align with the angel investor lineup. Lucy is operational with retailers and streaming platforms for virtual try-on and dynamic advertising, according to the company; Oasis is aimed at robotics and autonomous vehicle simulation, with gaming and live-experience applications encapsulating both offerings.
In the overarching industry context, as the sector congregates at events like Cannes and discussions around AI-generated media transition from potential to procurement decisions, the current market faces challenges with real-time, low-latency inference at production scale, a problem that existing video models have yet to address effectively.
Decart did not reveal run-rate revenue, the new valuation, or the hyperscaling customers that are licensing DOS. The upcoming public benchmarks include DOS 2.0, introduced alongside the funding, with Lucy 2.5 and Oasis
Other articles
Google has sold so much TPU capacity that its own researchers are now waiting in line for the remaining resources.
Google's TPU stack is backing agreements with Anthropic and Meta. According to Bloomberg, this success has turned internal compute access at Google and DeepMind into a highly sought-after resource.
Perhaps consider using a different option for your AI images instead of Gemini and ChatGPT. I just found an alternative that might interest you.
Ideogram may not replace all AI image generators, but its precision with text, format controls, and remix features make it a worthwhile option for creating posters, thumbnails, social graphics, and other visuals that require strong design elements.
Walmart's new ONN tablets are budget-friendly and provide substantial value.
Walmart is offering two new Onn choices for budget tablet shoppers, featuring a compact 8.1-inch version and a larger 13-inch Pro model that includes a big display, a stylus, and Android 16.
Perhaps consider moving away from Gemini and ChatGPT for your AI images. I recently found an alternative that you might like.
Ideogram may not be a substitute for every AI image generator, but its precision in text, formatting options, and remixing features make it worthwhile for creating posters, thumbnails, social media graphics, and other visually-driven designs.
Dust secures $40 million to advance enterprise AI beyond the single-user phase.
Dust has secured $40 million in a Series B funding round co-led by Abstract and Sequoia, with participation from Snowflake and Datadog, to expand its 'multiplayer' agentic workspace for businesses.
US and Philippines take action
Under Secretary Helberg toured the New Clark City site alongside American companies. The BCDA proposed a two-year lease grace period but declined Washington's request for diplomatic immunity at the facility.
Decart secures $300 million to develop a real-time world model for Amazon's chips.
Decart has secured $300 million in funding, with Radical Ventures leading the round, bringing the total amount raised by the real-time world-model lab to over $450 million. NVIDIA, Adobe, Toyota, and eBay Ventures participated in this funding, alongside angel investors Karpathy and Eisner.