Nvidia paid $20 billion for Groq’s technology and poached its CEO. Six months later, Groq raised $650 million, hired new leadership from xAI and Meta, and pivoted to neocloud. The AI chip war’s most dramatic survival story.
What Happened
In December 2025, Nvidia struck a $20 billion deal with Groq — licensing Groq’s LPU technology and hiring away founder/CEO Jonathan Ross and key executives. Nvidia then built its own “Nvidia Groq 3 LPX” inference hardware, announced at GTC in March.
Most companies don’t survive losing their CEO, their core IP, and their primary technology differentiator to the market’s dominant player. Groq just raised $650 million to prove it can.
New leadership: Alan Rice (COO, ex-xAI and Meta), Sinclair Schuller (CTO), Rakesh Malhotra (CPO). The pivot: from chip design to neocloud — running inference infrastructure using the remaining Groq hardware plus Nvidia GPUs.
The key insight: Nvidia’s $20B deal was designed to absorb Groq’s inference advantage. Groq’s $650M raise is the bet that the neocloud market — inference-as-a-service for AI companies that don’t want to build their own data centers — is big enough to build a second act on.
The Structural Read
This connects to the SpaceX Colossus story: the neocloud market is exploding. SpaceX has $27B+ annualized in AI compute contracts. Groq is betting it can capture a slice of the same market — fast inference for AI companies that need speed over cost.
The AI chip competitive moat is shifting. Nvidia absorbed Groq’s hardware advantage. Groq’s response: if you can’t beat Nvidia on chips, beat them on inference economics. The neocloud pivot is the admission that the substrate layer belongs to Nvidia — but the service layer above it is still up for grabs.
Source: TechCrunch









