Eseandre Mordi
⬤ Nvidia moved higher in pre-market trading after announcing a major shift in its AI strategy. Groq's CEO revealed a non-exclusive licensing deal where Nvidia will pay $20 billion in cash to access, develop and deploy Groq's AI inferencing technology. GroqCloud will keep operating independently, while Groq CEO Jonathan Ross—a former Google engineer who helped build the Tensor Processing Unit project—will join Nvidia. NVDA closed at $188.61 on December 24 and was trading near $189.84, up about 0.65% in the pre-market session.
⬤ The deal aligns with Nvidia's push into AI inferencing while the company reportedly scales back its SuperCloud (DGX Cloud) initiative. That project aimed to provide GPU-based cloud services directly to enterprise customers but risked competing with major hyperscalers developing their own custom ASIC hardware. Nvidia has also been expanding internal engineering resources tied to a custom ASIC program.
⬤ Nvidia's established CUDA software ecosystem remains a key advantage. The agreement is non-exclusive, meaning Groq can still partner with other firms while cooperating with Nvidia. This flexibility may help both companies navigate the evolving AI hardware landscape.
⬤ The announcement highlights Nvidia's expansion beyond AI model training into large-scale inferencing, strengthening its position in the global AI supply chain. The initial positive market reaction suggests investors see the Groq deal as strategically supportive of Nvidia's long-term AI roadmap, though competition between GPUs and custom ASIC solutions continues to intensify.
Eseandre Mordi
