Decart, one of the only Israeli companies developing a large language model (LLM), has ignited the enthusiasm of investors and raised $100 million from existing investors, on much better terms than in the past. Less than a year after its last funding round in December 2024, AI company Decart AI is receiving a $3.1 billion valuation, a more than six-fold increase.
This is a rare move for an early stage startup. Companies are usually eager to add new investors with each new funding round in order to attract higher-profile investors and bring a larger number of strategically valuable experts to the investor table. In this case, these are existing investors, including Sequoia, Valley Capital, Benchmark and Zeev Ventures – the venture capital fund of the Israeli investor Zeev Ventures, in which Peter Thiel, the first investor in Facebook and Palantir, is also invested. New investors, such as the Israeli Aleph Fund, joined the financing round – apparently due to the connection of partner Michael Eisenberg to Benchmark, in which he previously served as a partner.
Decart explains that although it has raised $153 million so far – it has actually used less than $10 million of the investors’ money because it has covered the heavy costs for graphics processors, which amounted to tens of millions of dollars, from its revenue.
The Tel Aviv-based company has transformed from a startup that developed a model that reduces the costs of training and operating LLMs, into an AI lab that produces revenue from three main areas, involving AI applications for the individual user: an optimization service that is capable of performing more training and operating operations on the same graphics processor – in other words, reducing the cost of Nvidia processors (optimization); a model development service that interfaces with customers’ existing models (API) and is provided to the end user via their cloud; and a new product – research or model development services in the video field, a niche in which the company has positioned itself through the launch of the Oasis and Mirage video generators. The company also plans to launch language models for the audio field, and does not limit itself to a specific format, but calls Decart an AI lab that studies the behavior of AI models as well as the interface between humans and machines.
The company’s latest video language model, Mirage, has received positive market feedback after demonstrating the ability to generate virtual reality using real-time text queries (prompts) faster than models like OpenAI’s Sura or Google’s Veo3. This feature is advantageous, for example, in creating personalized advertisements, video games, or other commercial experiences based on the user’s characteristics, interests, and query content.
Decart was founded in 2023 by CEO Dr. Dean Leitersdorf, CPO Moshe Shalev, both veterans of the IDF 8200 intelligence unit. Over the past year the company has grown from 15 to over 60 employees most working in the Tel Aviv headquarters as well as in offices in northern Israel, San Francisco and New York.
Decart says it became profitable almost from day one. The company generated millions of dollars in revenue through agreements for use of the GPU optimization technology it developed, which is deployed by prominent cloud providers and AI labs. This technology also forms the basis of the company’s video models, and enables a significant reduction in costs, from tens to thousands of dollars (such as $1,400 per hour for producing a prompt-based video, in advanced models) to less than 25 cents per hour in Decart’s models. This cost reduction opens the door to a new era of real-time applications – from the worlds of gaming, through personalized entertainment experiences to advanced robotics. This has led to strong demand for the model’s API, which is expected to be launched soon and join the company’s optimization activities as another key growth engine.
Investor enthusiasm for the company’s achievements may also spill over to potential buyers. Last year, Nvidia completed a billion-dollar acquisition of two companies in Israel that have some similarities to Decart: Run:AI, which developed technology to streamline calculations on GPUs and extract more training and execution capabilities from the same GPU; and Deci AI, which developed algorithms to “shrink” large language models. Nvidia is increasing its presence in Israel by expanding its software team, and a company like Decart could once again whet its appetite for AI acquisitions in Tel Aviv.
Published by Globes, Israel business news – en.globes.co.il – on August 7, 2025.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2025.
