When it comes to venture capital, one truth keeps proving itself over and over again: it’s not dozens of portfolio investments that define a fund’s outcome, but one or two outsized wins. This phenomenon is known in the VC world as the power law.
Toward the end of 2025, the buzz around the deal between NVIDIA and Groq brought this principle back into focus — this time highlighting the sharp contrast between hard facts and social-media hype.
NVIDIA and Groq: not an exit, but a strategic shift
According to TechCrunch, NVIDIA entered into a non-exclusive licensing agreement with AI chip startup Groq. As part of the deal, NVIDIA is hiring Groq founder Jonathan Ross along with several key engineers from the company.
Crucially, NVIDIA made an official clarification: this is not an acquisition of Groq. At the same time, CNBC reported that certain Groq assets involved in the deal could be valued at around $20 billion. NVIDIA neither confirmed nor denied this figure.
Why this is not a “big exit”
On social media, the deal was quickly framed as:
- “one of the largest VC exits in history,”
- “a 50–60x return,”
- “billions returned to LPs.”
However, the facts tell a different story:
- Groq remains an independent company
- the company itself was not sold
- no confirmed cash returns to LPs have been disclosed
In other words, this is not a realized exit, but rather a case of strategic monetization.
What power law really means in VC
In venture capital, power law means:
- most investments deliver average or below-average outcomes
- one or two companies grow so dramatically that they determine the fund’s overall returns
The Groq story represents a process that is still unfolding. In September 2025, the company raised:
- $750 million
- at a $6.9 billion valuation
These numbers are not yet realized outcomes — but they clearly signal how large the upside potential can be.
Why NVIDIA chose Groq
Groq develops LPU (Language Processing Unit) chips as an alternative to traditional GPUs. The company claims that its LPUs can:
- run large language models 10× faster,
- while consuming 10× less energy.
In today’s AI race, this directly addresses one of the most critical bottlenecks: inference efficiency.
For NVIDIA, the deal means:
- bringing a competing technology under strategic control
- absorbing top-tier engineering talent into its ecosystem
- reinforcing its leadership in the AI hardware market
The key lesson for VCs
The Groq hype reinforces a fundamental takeaway:
Venture capital does not reward stability — it rewards asymmetry.
Fund performance does not come from:
- twenty “good” startups,
but from - one investment made correctly, and at the right time.
And these wins:
- do not happen quickly
- do not always look like flashy exits
- often take years to fully materialize
