[TechCrunch] After Nvidia's $20B Deal, AI Chip Startup Groq Reportedly Raising $650M

**Groq** — the AI chip startup known for its ultra-fast LPU (Language Processing Unit) inference hardware — is looking to raise **$650 million** in new funding from existing investors, sources tell TechCrunch.

What happened

The funding round comes on the heels of a major **December 2025 deal with Nvidia** worth an estimated $20 billion. That deal — described by some as a "not-an-acquisition" — involved:

• The licensing of Groq's hardware technology to Nvidia
• The departure of some top-level senior Groq employees to the chip giant
• A cash payout to Groq's investors that Axios reported would have been Nvidia's largest acquisition, had it been structured as a full buyout

Following that restructuring, Groq is now pivoting its focus to an **inference neocloud business** — offering cloud-based inference services built on its own homegrown AI chips and systems. The company is currently led by interim CEO and CFO Adam Winter, and Matt Eng as interim CFO.

The funding details

According to TechCrunch's sources, the $650 million round may be effectively guaranteed: Groq's backers **Disruptive** and **Infinitium** have agreed to fill the round should other existing investors not take their pro-rata shares.

Why this matters

The Groq-Nvidia deal and subsequent fundraise signal several important trends for AI hardware and self-hosting:

**Infrastructure consolidation.** Nvidia's appetite for acquiring — or quasi-acquiring — AI chip IP suggests the hardware layer is consolidating rapidly. For the self-hosted community, this could mean fewer options for alternative inference hardware over time.

**Neoclouds are real.** Groq's pivot to an inference neocloud validates the market for specialised AI inference-as-a-service, competing with the likes of Together AI, Fireworks, and other inference providers. This ecosystem benefits self-hosters by driving down inference costs and improving tooling.

**The LPU architecture lives on.** Despite the talent movement, Groq's hardware continues to power fast inference, and the new funding suggests confidence in the neocloud direction. For anyone running local models, the existence of competitive inference hardware keeps downward pressure on GPU pricing.

For more on the hardware landscape, see our Best Hardware for Self-Hosted AI guide, or explore how DeepSeek R1 and other reasoning models perform on different hardware configurations.

**Source:** TechCrunch — After Nvidia's $20B not-acqui-hire, AI chip startup Groq reportedly raising $650M