Decart, an AI research lab building chip-agnostic infrastructure and real-time world models, announced a $300 million funding round on May 18, 2026, led by Radical Ventures. Nvidia joined as a new investor alongside several strategic venture arms and high-profile individuals, while Amazon appeared in the announcement as a strategic customer rather than an equity backer.
The raise illustrates an accelerating flow of capital into technology that promises to make AI inference faster and cheaper across multiple hardware platforms. Decart's core optimization software, DOS, runs on Nvidia GPUs, Google TPUs and Amazon Trainium chips, positioning the company at the intersection of a market where chip manufacturers usually push for tighter ties.
The round's new investors included Nvidia, Atreides Management, Valor Equity, Adobe Ventures, Toyota Ventures and eBay Ventures. Returning backers Sequoia Capital, Benchmark and Zeev Ventures also participated. Individual participants were Andrej Karpathy, co-founder of OpenAI, former Disney chief Michael Eisner, the Nintendo family and Moritz Baier-Lentz. Decart said it has now raised more than $450 million in total, following a $100 million round in August 2025 from a valuation of $3.1 billion, as reported by Fortune at the time.
Decart identified Amazon Web Services as a strategic partner for its Trainium chips, underlining the multi-platform pitch. The company did not list Amazon as an investor, and its announcement stressed the customer relationship. That distinction carries weight: Nvidia's participation as a shareholder, while Decart's software optimizes across rival silicon, signals that the dominant chipmaker is comfortable with an inference layer that works with its own hardware and others.
Decart said its DOS 2.0 stack delivers more than 1,600 tokens per second and up to 100 frames per second for full-HD video and world-model inference. The company described those figures as delivering a 100-fold cost improvement and eightfold speed gain. Independent benchmarks are not available. The lab added that new versions of its Lucy immersive world model and Oasis physical-AI model are expected within weeks.
Media reports, citing people familiar with the matter, pegged the round's valuation at about $4 billion. Decart did not confirm that figure. Some publications earlier labelled the raise as Series C, while Fortune had designated the August 2025 funding as Series B; the company itself assigned no series label. In addition, MarketScreener and S&P Capital IQ described the $300 million as the second and final tranche of a $400 million amended transaction, a structure Decart's announcement did not mention. The company did not immediately respond to a request for clarification.
The investment highlights demand for low-latency AI infrastructure as hyperscalers and labs race to cut inference costs. Decart's bet on world models—applications that render real-time interactive environments for media, robotics and autonomous systems—adds a speculative but potentially significant commercial opportunity. The presence of Adobe Ventures, Toyota Ventures and eBay Ventures among the backers points to possible use cases in content creation, mobility and commerce.
Decart has not disclosed run-rate revenue or named cloud-lab customers beyond Amazon. Until independent benchmarks emerge, its performance claims remain the company's own. The round nonetheless places it at the center of a debate over whether chip neutrality can coexist with direct hardware-maker investments, a question that will gain sharper focus as Nvidia-backed firms multiply.