Groq is pursuing $650 million in internal funding from existing investors as the AI chip startup shifts focus from hardware to inference computing, according to Axios. The company's inference neocloud business leverages its proprietary AI chips to help developers and enterprises run AI-powered applications.

This fundraising follows Groq's unusual December deal with Nvidia valued at $20 billion, where the chip giant licensed Groq's hardware technology while hiring some senior employees. Though not a full acquisition, existing investors received cash payouts from what would have been Nvidia's largest purchase. Now these same investors face requests to reinvest in Groq's cloud strategy.

The pivot reflects a broader market reality: AI inference—the processing that occurs after prompts—is currently more critical than model training. Groq's interim CEO Adam Winter and CFO Matt Eng are steering this transition, with backing from Disruptive and Infinitium who've committed to fill the round if other investors decline their pro-rata shares.

Groq's fundraising signals continued investor confidence despite the company's unusual path. The inference cloud market represents a significant opportunity as enterprises deploy AI applications at scale, making this pivot strategically sound even after the lucrative Nvidia arrangement.