According to media reports citing informed sources, the “AI Chip giant”$NVIDIA (NVDA.US)$has successfully acquired the company Gretel, which focuses on synthetic data, for a total value in the nine figures. Informed sources added that the specific value is still unknown, but the overall value of this trade clearly exceeds the recent valuation of Gretel, headquartered in San Diego, at up to 0.32 billion dollars, while other specific terms related to the acquisition agreement remain unclear.

In the view of some professional AI developers, NVIDIA's significant investment in acquiring the synthetic data company Gretel is an important component of their strategic planning to achieve a "full-stack AI" goal. By introducing synthetic data technology, NVIDIA can provide high-quality AI training data for global enterprises or government institutions to develop/fine-tune generative AI large models or AI agents, thus enhancing a complete ecosystem that encompasses high-performance hardware like the H100/H200/Blackwell series GPUs to AI software, and then to data systems and AI application development.

According to available information, Gretel offers a comprehensive platform for acquiring synthetic data to better train AI large models, and provides application programming interfaces (APIs) for professional AI developers who wish to build generative AI large models but lack sufficient training data or have privacy and copyright concerns regarding the use of real-world data. Gretel's platform can also fine-tune existing open-source large models to add differentiated privacy and secure access features.

Synthetic data refers to data that is not created by humans, simulating real-world data. It is generated using extremely complex statistical methods or technologies such as deep learning and generative AI large models, and can be presented in multimedia, tabular, or textual form. Various organizations across different industries can use it for research, testing, new development, and machine learning studies.

The most important point is that synthetic data is expected to address the explosive demand for higher-quality real-world training data from major developers in the AI field, such as OpenAI, amidst the rapid iteration of AI large models, while the supply of such data has been severely lacking. The synthetic data provided by Gretel can largely meet the substantial data required for training by these leading developers of large models, with the overall quality of this synthetic data being comparable to that of real-world data.

Informed sources pointed out that Gretel and its team of about 80 employees will be integrated into NVIDIA's business division, and its synthetic data technology will be an important part of NVIDIA's growing generative AI developer service ecosystem based on Cloud Computing Service, which also means NVIDIA is getting closer to realizing its so-called ambition of a "full-stack AI".

NVIDIA has declined to comment on media reports related to the acquisition of Gretel.

According to media reports, this acquisition coincides with NVIDIA launching its own synthetic data-assisted generation tools, allowing developers to iteratively train their exclusive AI large models and tailor them specifically for targeted applications. The successful acquisition of Gretel undoubtedly significantly enhances and perfects NVIDIA's ecosystem in AI application development.

Media reports citing Pitchbook data indicate that, prior to being acquired by NVIDIA, Gretel had raised over 67 million dollars in venture capital funding.

In December 2023, Gretel announced a partnership with$Microsoft (MSFT.US)$its ultra-large-scale Cloud Computing Service platform Azure and joined the Microsoft-led startup$Pegasus (6262.JP)$Program (Microsoft for Startups Pegasus Program). Last November, Microsoft announced the collaboration of Gretel with Azure's exclusive AI Cloud Service — Azure OpenAI Service. The Azure OpenAI Cloud Service is an enhanced Cloud Computing service provided by Microsoft that integrates OpenAI's proud AI large model, allowing users to invoke OpenAI's incredibly powerful latest version AI large language model via REST API.

NVIDIA's "AI Full Stack" grand vision receives strong support from Gretel.

Acquiring the synthetic data-focused company Gretel is an extremely important part of NVIDIA achieving its "AI Full Stack" grand vision, thus significantly enhancing and perfecting NVIDIA's complete ecological chain from Hardware to Software, and then to Data and AI application developers' exclusive platform. This move fills a crucial gap in data supply within NVIDIA's existing software and hardware ecology, making its "AI Full Stack" strategic layout more complete.

Synthetic data can significantly alleviate the issue of insufficient supply of real data while protecting data privacy. Gretel's platform generates synthetic data highly similar to real-world data through complex simulations using Statistics and deep learning methods, allowing developers to obtain a large amount of high-quality training data, thereby enhancing the overall performance and generalization ability of AI large models. By acquiring Gretel, NVIDIA is able to integrate this capability into its cloud-based NVIDIA DGX AI developer ecosystem, helping AI large model developers train or fine-tune generative AI large models more efficiently.

Currently, NVIDIA is not only a leader in the AI Hardware field, but its strategic layout has also covered AI Software tools and service platforms. In recent years, NVIDIA has successively launched several software products, including generative AI tools, aimed at providing developers with end-to-end AI technology innovation or AI-native application development platforms. Gretel's technology can seamlessly integrate into NVIDIA's existing AI developer ecosystem, further perfecting its data generation and training processes, thereby enhancing the competitiveness and attractiveness of the entire Cloud Computing-based Software service platform.

Editor/danial