Nvidia Pays $20B to License Groq AI Inference Tech, Eliminates Rival
Nvidia has agreed to pay approximately $20 billion to license Groq’s AI inference technology under a non-exclusive deal that includes hiring Groq founder Jonathan Ross, president Sunny Madra and key engineering staff. This initiative eliminates a potential competitor and establishes Nvidia’s entry into the non-GPU AI inference chip market.
1. Nvidia Secures Groq’s AI Inference Technology Through Strategic Licensing
In late December, Nvidia entered into a non-exclusive licensing agreement with startup Groq for its AI inference architecture. The deal grants Nvidia rights to Groq’s inference IP — including its proprietary instruction set and compiler optimizations — enabling Nvidia to integrate specialized inference accelerators alongside its existing GPU lineup. By licensing rather than fully acquiring Groq, Nvidia preserves flexibility to collaborate with other ecosystem partners while fortifying its inference roadmap.
2. Reported $20 B Valuation Reflects Significant Premium Over Prior Funding
Although neither party disclosed official financial terms, multiple industry sources peg the transaction at roughly $20 billion. That figure represents nearly three times Groq’s post-money valuation of $6.9 billion following its September funding round. For Nvidia, a $20 billion outlay — its largest-ever deal compared to the $6.9 billion Mellanox acquisition in 2020 — underscores the strategic importance it places on diversifying beyond GPUs into custom inference silicon.
3. Key Groq Leadership Joins Nvidia to Drive Next-Gen Chips
As part of the agreement, Groq’s founder and CEO Jonathan Ross, president Sunny Madra and approximately 100 core engineers will join Nvidia. Ross, credited as a driving force behind Google’s original tensor processing unit, is expected to lead a dedicated Nvidia inference group. Groq’s remaining operations, including GroqCloud services, will continue under new leadership. This talent infusion accelerates Nvidia’s internal roadmap for ASIC-class inference solutions and bolsters its engineering bench strength.
4. Strengthening Nvidia’s Position in the AI Inference Market
With this licensing and acqui-hire, Nvidia both neutralizes a rising competitor and expands its product arsenal for cloud and edge inference workloads. Analysts estimate the global AI inference market will exceed $50 billion by 2028. Nvidia’s move positions it to offer differentiated inference accelerators optimized for low-latency, per-query pricing models that complement its high-throughput GPUs. The transaction also pre-empts major hyperscale cloud providers from adopting Groq technology independently, reinforcing Nvidia’s leadership across AI training and inference segments.