Groq Secures $650 Million to Rebuild Inference Cloud After Nvidia Deal

Key Points

• Groq is raising $650 million from existing investors to fund its inference‑cloud business.
• The financing round is effectively guaranteed, with Disruptive and Infinitium pledging to back it if needed.
• Nvidia’s $20 billion not‑acqui‑hire paid cash to Groq’s investors, hired senior engineers and licensed the chip technology.
• Interim CEO Adam Winter and CFO Matt Eng lead the company as it rebuilds around its LPU hardware.
• Groq’s LPU chips claim higher token‑per‑second performance than comparable Nvidia GPUs.
• The inference market is heating up, with competitors like Cerebras, Fractile and Google launching specialized hardware.
• Model providers are cutting API prices, intensifying cost pressure on inference cloud services.
• Nvidia’s upcoming Blackwell and Vera Rubin platforms aim to close the performance gap that Groq exploits.