Nvidia has entered into a non-exclusive licensing agreement with Saudi-backed AI start-up company Groq, as part of their efforts to enhance their AI inference capabilities.
The move has the potential to completely reshape the AI hardware landscape.
Reports had initially claimed that AI semiconductor behemoth Nvidia had acquired the privately-held Groq for $20bn.
However, it has now emerged that Nvidia has paid $20bn to enter into a non-exclusive licensing agreement with Groq, where Groq essentially become Nvidia employees.
The agreement includes an acqui-hire arrangement that will bring Groq’s founder Jonathan Ross, President Sunny Madra, and other employees to Nvidia, where they will focus on scaling high-performance, low-cost AI inferencing solutions.
As a prominent player in the Semiconductors industry with substantial liquidity (current ratio of 4.47), Nvidia is well-positioned to integrate Groq’s talent and technology.
The deal brokered provides Nvidia with access to Groq’s novel hardware architecture and software compiler capabilities, potentially complementing Nvidia’s existing portfolio, including its NVLink C2C interconnect technology.
Groq’s technology, which uses on-chip SRAM rather than High Bandwidth Memory (HBM), could provide Nvidia with performance advantages in specific inference workloads, potentially relating to Nvidia’s Rubin CPX processor announced in early September that focuses on large context inference optimization.
