Overview
- Nvidia said it signed a non‑exclusive licensing agreement with Groq and hired founder Jonathan Ross and other engineering leaders, with CNBC reporting an unconfirmed $20 billion value that the company declined to comment on.
- The move targets AI inference workloads where custom chips have gained traction, with some analysts calling the arrangement an acquisition in all but name and noting growing scrutiny of licensing‑plus‑hiring deal structures.
- Groq’s LPUs focus on low‑latency inference using on‑chip SRAM, contrasting with Nvidia GPUs that pair with high‑bandwidth memory from suppliers such as Micron and Samsung.
- Nvidia previously told investors it is sold out of cloud GPUs for its fiscal Q3 ended Oct. 26, reinforcing capacity constraints that have pushed some customers to evaluate alternatives.
- Upstream, Micron posted record fiscal Q1 revenue of $13.6 billion and adjusted EPS of $4.78 on soaring HBM/DRAM demand, while Bernstein reiterated an Outperform on Nvidia with a $275 target, citing attractive valuation and durable AI capex tailwinds.