A perception had gained ground (and perhaps still holds) that chip giant and the world's most valuable company by market cap, Nvidia (NVDA), would see its dominance in the AI world diminish due to its perceived weakness in inference. Inference is basically the real-world test for models, which have been trained on large datasets. The trained models apply what they “learned” to answer questions, classify images, translate text, or generate content.
To address this, late last year, Nvidia made its largest acquisition ever: buying AI startup Groq for $20 billion. Except this was not a buyout; it was a licensing agreement. Essentially, the contours of this agreement involved Nvidia gaining access to Groq's low-latency inference technology and Groq remaining an independent company. However, the optics were that of a purchase, as the then-CEO and COO of Groq, Jonathan Ross and Sunny Madra, respectively, joined Nvidia.