Our Take
Groq survived the not-acqui-hire intact because it owned a second business (neocloud) that Nvidia didn't want—not because chip startups are resilient.
Why it matters
This shows how IP licensing deals between AI infrastructure rivals reshape startup strategy. Groq's pivot to cloud services, not chips, is now the path forward for inference players watching their core tech get absorbed by larger competitors.
Do this week
Inference cloud operators: audit your revenue mix now to confirm you have at least two defensible business lines, so IP licensing alone doesn't collapse your runway.
Groq raises $650M, pivots to inference cloud after Nvidia licensing deal
Groq announced a $650 million funding round on Monday, led by Disruptive (a Dallas-based late-stage firm whose founder Alex Davis chairs Groq's board) and Infinitum, a Fort Lauderdale hedge fund. The company did not disclose its new valuation. Its last valuation was $6.9 billion following a $750 million round in September (company-reported).
The funding comes six months after Nvidia signed a non-exclusive licensing agreement for Groq's language processing unit (LPU) technology in December and hired away Groq founder and CEO Jonathan Ross, president Sunny Madra, and other employees. Nvidia then announced its own inference hardware cluster, the Nvidia Groq 3 LPX system, at its GTC event in March.
In response, Groq has shifted focus to its neocloud business, an inference cloud service that grew from Groq's 2024 acquisition of Definitive Intelligence (previously run by Madra). The neocloud now operates 13 data centers across North America, Europe, the Middle East, and APAC, serving over five million developers and thousands of AI companies and processing trillions of tokens weekly (company-reported).
Groq has also replaced its departed executives. Alan Rice joined as chief operating officer, previously at xAI and Meta. Sinclair Schuller joined as CTO and Rakesh Malhotra as CPO; both previously worked together at Nuvalence (acquired by EY in 2024) and Apprenda. Malhotra spent a decade on Microsoft's cloud products.
The real vulnerability: neocloud must outrun Nvidia's inference offerings
Groq's survival depends entirely on whether its inference cloud can compete now that Nvidia owns the underlying LPU IP. The company has existing scale (13 data centers, five million developers on record), but inference is experiencing both tremendous demand and increasing competition. The question is not whether Groq can survive a licensing deal; it is whether a cloud service built on licensed hardware—which Nvidia can now also offer—can hold margin and customer loyalty.
The precedent cited by the article is Scale AI, whose CEO reported that the company rebounded after Meta's $14.3 billion not-acqui-hire about a year ago and is on track to do $1 billion in revenue. That comparison is incomplete: Scale AI's core business (data labeling and curation for model training) was never threatened by Meta's deal because Meta didn't acquire or license Scale's primary product. Groq's situation is different. Nvidia now owns the IP that made Groq's hardware cluster valuable. Groq must now prove that operational excellence, developer network, and service integration can sustain the business where hardware commodity differentiation cannot.
What this means for inference operators
This deal structure—licensing core IP to a better-capitalized rival while keeping a service business—is becoming a standard exit path for infrastructure startups that lose the hardware innovation race. It works only if the service business has independent defensibility and existing momentum. Groq had both: neocloud was already generating revenue and customer count before the Nvidia deal closed. Startups earlier in that transition, or those relying entirely on proprietary hardware, should not assume they will have the same cushion.