Groq unveiled TSP technology that outperforms Nvidia GPUs in performance and energy efficiency. California-based startup Groq, founded in 2016 by former Google employer Jonathan Ross, has announced that it has raised a $640,000,000 investment to expand its ability to produce semiconductors designed for artificial intelligence models. The move is aimed at increasing competition in the AI chip sector, which is dominated by Nvidia with a market share estimated by analysts to be as high as 80%. The AI chip market is estimated to be worth $1 trillion by the end of the decade, attracting the attention of not only startups but also giants such as AMD, Intel, Arm, Microsoft, Google, Amazon and Meta. All of them are looking to create competitive solutions to Nvidia's graphics cards (GPUs), which are currently leading the field. According to Groq, their technology, called tensor streaming processor (TSP), achieves higher performance and energy efficiency than existing solutions. The TSP is a single-chip system that can process data in streaming mode, which significantly speeds up tasks for AI-specific tasks. The company claims that its chips can deliver up to 1 petaflops (quadrillion floating point operations per second) of performance while consuming only 100 watts, making them more efficient than Nvidia's GPUs. The investment raised will help Groq expand its team, strengthen its market position and accelerate new product development. The company has already begun shipping its chips to its first customers, including the U.S. Department of Energy, and plans to continue to expand its presence in this promising segment of the semiconductor market. Groq also announced a partnership with GlobalFoundries, one of the largest semiconductor manufacturers in the world, to manufacture its chips. This partnership will enable Groq to increase its production scale and meet the growing demand for its products. In addition, Groq plans to expand its market presence by offering not only chips but also software to utilize them. The company has already released a developer toolkit that allows developers to create AI applications optimized to run on Groq chips. Groq aims to become a serious competitor to Nvidia and other players in the AI model chip market by offering more efficient and specialized solutions for processing data used in AI-based tasks.