Nvidia Corp. is reportedly in advanced talks to acquire Artificial Intelligence21 Labs Ltd., a Tel Aviv-based startup focused on large language models and agent development tools. The report, originating from Calcalist, said that a potential deal could be worth between 2 billion and 3 billion. At the top end of the range, that’s more than double the valuation Artificial Intelligence21 received after its last publicly disclosed funding round in 2023, which included backing from Nvidia, Google LLC, Samsung Electronics Co. and other prominent investors.
Calcalist reported that Artificial Intelligence21 quietly raised another 300 million from investors earlier this year, and it’s believed the round closed at a valuation similar to the 2023 raise. The report also revealed that Artificial Intelligence21 is generating about 50 million in annualized revenue. The company offers a series of open-source large language models called Jamba, which it says can process long prompts up to 2.5 times faster than competing algorithms thanks to a hybrid architecture that blends transformer components with a neural network design called Mamba, which replaces transformers’ attention mechanism with a state space model to reduce memory usage and improve inference speed.
Artificial Intelligence21 monetizes its technology through a paid software platform called Maestro, which provides tools to help developers build and manage artificial intelligence agents. Maestro is designed to harmonize documents across different file formats, remove unnecessary information that might confuse an artificial intelligence agent, and apply other optimizations before data is processed. After an artificial intelligence agent generates a prompt response, Maestro validates the accuracy of the output and can present results in specified formats such as reports, while also allowing customization of infrastructure usage for responses. Nvidia might be planning to integrate Maestro into Nvidia Artificial Intelligence Enterprise, its software suite that ships with its graphics cards and includes pre-packaged large language models and development tools. News of the talks follows Nvidia’s recent move to ink a 20 billion deal to license the technology of Groq Inc., a startup that sells processors optimized for inference workloads and whose founding chief executive and other key employees will join Nvidia.
