The Idea: Our story begins in the early 2010s at Google where Jonathan Ross, who was working on the ad systems team, grew interested in neural networks. At the time, Google researchers were realizing that if their voice search and image recognition models were rolled out to all users, Google would effectively go bankrupt buying GPUs and building data centers to support the compute load. So Jonathan, working on a "20% project" (Google’s famous policy allowing employees to spend one day a week on side projects), began designing a chip specifically for machine learning. He had no formal background in chip design, which he later cited as an advantage - he didn't know "what couldn't be done.” This was the beginning of groq.

The Execution:

• 2016: Ross founds Groq in Mountain View. The team operates in "deep stealth" for nearly four years, ignoring the hype cycle to rebuild chip architecture from the ground up.
• 2020: They finally ship their first silicon. Unlike competitors who tried to beat Nvidia on raw power, Groq focused on latency - how fast the chip could respond.
• April 2021: They raise a $300M Series C led by Tiger Global and D1 Capital, hitting a $1B valuation. They hit "unicorn status," but still largely unknown outside deep tech circles.
• November 2023: The "Viral Moment." As the world obsesses over ChatGPT, Groq releases a public demo of an open-source model running on their LPU. It generates 500+ tokens per second (vs. Nvidia’s ~40). The internet breaks. It is the first time AI feels "instant," appearing on the screen faster than a human can read.
• August 2024: Riding the wave of "inference is the new training," they raise a $640M Series D led by BlackRock, tripling their valuation to $2.8B.
• September 2025: With AI agents becoming the dominant trend (which require massive inference speeds), Groq raises a $750M Series E at a $6.9B valuation.
• December 2025: In a stunning holiday twist, Nvidia announces a "non-exclusive licensing deal" worth ~$20 Billion. Effectively an "acqui-hire" of the highest order, Jonathan Ross and 90% of his team join Nvidia to integrate Groq's speed into the Nvidia ecosystem.

The lesson? Experts are often blinded by their own expertise. Jonathan cited his lack of formal training as his greatest competitive advantage: because he didn't study chip design, he didn't know "what couldn't be done." While the veterans were busy optimizing the status quo, the outsider was busy inventing the future. Because sometimes, the best way to disrupt an industry is to know absolutely nothing about how it’s "supposed" to work.