Nvidia buys synthetic data startup Gretel for $320 million
About 80 employees of the startup will join Nvidia, and its technologies will be integrated into the company's cloud services working with generative artificial intelligence.
About 80 employees of the startup will join Nvidia, and its technologies will be integrated into the company's cloud services working with generative artificial intelligence.
About 80 employees of the startup will join Nvidia, and its technologies will be integrated into the company's cloud services working with generative artificial intelligence.
By the time of the acquisition, the startup had attracted over $67 million in venture capital investment, Wired reports .
Gretel was founded in 2019 by Alex Watson, John Myers, and Ali Golshan (current CEO). The company has developed a synthetic data generation platform and a set of APIs for developers who want to build AI models but don’t have access to large amounts of confidential training data. Gretel doesn’t build its own AI models, but adapts existing open-source solutions, adding differential privacy and security mechanisms.
The acquisition comes amid Nvidia's aggressive development of synthetic data, which it uses to train AI models. Synthetic data allows for the creation of unlimited training sets to address the problem of a lack of real-world data.
Synthetic data, which is generated by computers, mimics real-world data and allows AI model training to scale, making it particularly attractive to financial institutions, government agencies, and healthcare companies where privacy is important.
Nvidia is actively developing this direction. In 2022, the company introduced Omniverse Replicator, a tool for generating 3D data for training neural networks. Last year, Nvidia launched Nemotron-4 340B, a series of open AI models that help generate synthetic datasets for various industries.



