After Nvidia poached its talent, here's what became of the remaining Groq team

Once a rising star in the bid to challenge Nvidia’s AI dominance, Groq is now charting a new course after a dramatic unbundling of its core chip technology and leadership. Last December, Nvidia paid roughly $20 billion to license Groq’s signature chip architecture and absorb its top executives, halting short of a full acquisition but fundamentally transforming the startup. Nvidia’s move, one of its largest deals ever, strengthened its position in the lucrative AI inference market - the segment focused on running AI models - where competitors have been making inroads primarily in speed and cost.

What’s left of Groq is a $3.5 billion AI-focused cloud infrastructure business that now operates with twelve data centers and a user base of over 6 million developers and thousands of companies. While the company was once known for its proprietary hardware, it is instead shifting to rely on Nvidia systems - the very technology Groq formerly sought to outpace. This pivot follows the departure of most of its engineering talent to Nvidia, along with the bulk of the team that built GroqCloud, its in-house cloud platform. Data center operations and core business functions remained with Groq, and the company has since recruited new leadership with expertise in cloud and infrastructure management. Adam Winter was appointed CEO in March, succeeding brief interim chief Simon Edwards, as the firm announced an aggressive refocusing on cloud services.

Despite these structural changes and a reduced valuation - from $6.9 billion before the Nvidia deal to $3.5 billion in its recent funding round - Groq is backed by $1 billion in fresh capital and has ambitions to quadruple its data center capacity over the next year. The company's leadership is betting on continued surging demand for AI compute as inference becomes an increasingly commoditized, scale-driven business. "We think that the needs for inference are great. It’s sort of like the restaurant business. People just need food," a source close to Groq said.

Transition hasn’t been seamless, according to former employees and customers. Many clients reported that Groq’s once “very, very hands-on” technical support deteriorated after the Nvidia deal, and several noted delays in AI model updates and reduced choice in available models. Paul Richards, CEO of AI startup Recall, said his company scaled back its reliance on Groq, citing the sudden departure of all their previous contacts and discontinued models. Lawrence Liu, cofounder of Willow, had to initiate a transition with new support staff, though service remained stable. Andre Dean Smith, CEO of ScreenApp, said Groq lagged on model updates and did not deliver an expected feature, causing his company to temporarily shift elsewhere before resuming expanded use of Groq services. Not all users reported issues; Eoin McMillan, founder of Sourcetable, said operations continued unaffected from his perspective.

The terms and structure of the Nvidia-Groq deal have attracted scrutiny from regulators, including a reported Justice Department investigation into whether the transaction was designed to avoid antitrust review. Nvidia stated that the deal reflected "the American system working as designed," while emphasizing Groq’s independence as a separate business entity. Company representatives state that departing support contacts were replaced and that no formal complaints reached management.

In a rapidly crowding inference market, Groq now faces competition from specialist service providers like Fireworks and Together AI, cloud natives such as CoreWeave, and tech giants including Amazon and Google. Groq asserts that its operational experience and early access to next-generation Nvidia inference systems position it for continued relevance. Investors say its billion-dollar-financed AI cloud should still be a significant player, but others note that differentiators like speed are narrowing, and developer tools such as OpenRouter make it easier for customers to move workloads between providers. With the economics of speed and price now paramount, Groq’s bet is simply on meeting insatiable AI compute demand. As one source close to the company observed, "You don't necessarily need a moat."