Chip cost reduction starts with acquisitions.
Compile | ZER0
Advertisement
Edit | desert shadow
Nvidia is becoming one of the most active corporate investors in the AI boom. After investing in more than 30 startups last year, it has now added two Israeli AI startups to its investment list –Run:ai and Deci.
NVIDIA announced today that it has entered into a definitive agreement to acquire Run:ai.According to CTech reports, the transaction price is expected to be approximately $700 million.
According to LinkedIn, Run:ai has about 150 employees and has raised a total of US$118 million; Deci has about 100 employees and has raised a total of US$55 million. Nvidia's transaction with Deci was not publicly disclosed, and the transaction price is unknown.
Advertisement
According to The Information, the two startups recently acquired by Nvidia aim to reduce the cost of developing or running generative AI models and help customers make more efficient use of AI computing resources, thereby boosting demand for Nvidia's AI chips.
Over the past year and a half, Nvidia's revenue has skyrocketed as demand for its chips has surged along with the AI craze. However, more and more companies are beginning to feel burdened by the high operating costs and difficult-to-balance returns of AI models. If Nvidia can help reduce the cost of running models, more companies may be willing to place orders with Nvidia.
01.Run:ai: Improve AI chip efficiency,Achieve efficient cluster resource utilization
Founded in 2018, Run:ai is a kubernetes-based workload management and orchestration software provider designed to achieve efficient GPU cluster resource utilization.
It enables developers to run multiple AI workloads in parallel rather than sequentially, making AI chips more efficient, which in turn helps reduce the number of Nvidia GPUs needed to complete a task.
The acquisition of Run:ai is Nvidia's largest acquisition in Israel since its $6.9 billion acquisition of Mellanox in March 2019.
NVIDIA's Israel R&D center, first established in 2016, has 4,000 employees in Israel and is led by Amit Krig of NVIDIA's SVP software and NIC product lines. In February 2022, Nvidia also announced the acquisition of Excelero, an Israeli high-performance software-defined storage startup.
02.Deci: Self-developed smaller model,The open source field versus Llama and Mistral
Deci, founded in 2019, has a different focus, adjusting AI models so they can run more cheaply on AI chips and powering machine learning applications developed using Nvidia's CUDA software – for example by making the models smaller.
According to its website, Deci serves clients including Adobe and Applied Materials. Its approach is similar to OmniML, another startup Nvidia quietly acquired last year, which aims to shrink machine learning models.
The startup initially helped enable relatively simple AI applications on end-side devices such as mobile phones and cars, and then turned to self-developed large models to compete in the open source field. Its self-developed model has been uploaded to its website and Hugging Face community. Microsoft is also a partner of Deci, and Deci's technology is accessible through Microsoft Azure AI Studio.
In December last year, Deci's model DeciLM ranked first in Hugging Face's model rankings containing 7 billion parameters, surpassing opponents such as Llama and Mistral in the open source field. But after Google launched its new Gemma model at the end of February, Deci lost its leading position.
Deci's annual recurring revenue (ARR) is estimated to be in the millions of dollars and it already has dozens of customers.
03.The business model remains unchanged after the acquisition.Attract more customers with efficiency-improving AI chips
NVIDIA's current market capitalization hovers around US$2 trillion, and its valuation has increased from US$1 trillion to more than US$2 trillion in just 9 months, making it one of the top three global market capitalization companies.
NVIDIA announced that it will continue to offer Run:ai products under the same business model and will continue to invest in the Run:ai product roadmap as part of the NVIDIA DGX Cloud. NVIDIA DGX Cloud is an AI platform co-engineered with leading enterprise development clouds, providing integrated, full-stack services optimized for generative AI.
NVIDIA DGX and DGX Cloud customers will have access to Run:ai’s AI workload capabilities, specifically large language model deployment. Run:ai's solutions are integrated with products such as NVIDIA DGX, NVIDIA DGX SuperPOD, NVIDIA Base Command, NGC containers and NVIDIA AI Enterprise software.
Nvidia said that after acquiring Run:ai, customers can expect to benefit from better GPU utilization, improved GPU infrastructure management and greater open architecture flexibility.
According to The Information, two people with knowledge of the deal said Run:ai's ability to improve the efficiency of AI chips may attract more customers to NVIDIA DGX Cloud. Nvidia may offer Deci and OmniML technology along with CUDA software to make it cheaper for developers to build AI-driven applications.
04.Conclusion: possible transactions in the near futurehelpfulStrengthen control of the AI chip market
Nvidia has been leveraging its investor relations as it pursues mergers and acquisitions. In the months leading up to the two deals, Nvidia asked at least one major venture capital firm to present nearly a dozen of its most promising AI companies to executives, including Nvidia founder and CEO Jensen Huang, a person familiar with the matter said. Startups.
Nvidia's recent deal could also help it tighten its grip on the AI chip market, making its GPUs more attractive to companies that initially thought they were too expensive.
Source: CTech, The Information, Globes