Groq raises $650M to rebuild AI inference cloud after Nvidia's $20B licensing deal stripped key talent

AI chipmaker Groq confirms $650M raise, re-staffs after Nvidia's $20B not-acqui-hire deal

What does an AI company do after one of those not-acqui-hire deals, where a rival pays investors a hefty IP "licensing" fee while poaching its critical talent? For AI chipmaker Groq, the answer appears to be raise more money from investors -- who were said to have profited handsomely after a deal with Nvidia in December -- hire more talent, and pivot. On Monday, Groq announced a new $650 million funding round, confirming earlier reports. The raise comes roughly six months after Nvidia signed a non-exclusive licensing agreement for Groq's technology and hired away founder and CEO Jonathan Ross, president Sunny Madra, and other employees. Groq did not disclose its new valuation. It was last valued at $6.9 billion following a $750 million round in September. Ross, who came from Google, was known in the AI chip world for helping create Google's AI chip, the Tensor Processing Unit. He teamed up with another Google engineer, Doug Wightman, to launch Groq a decade ago. Wightman stayed on after the Nvidia deal and became CEO. Groq created a chip it called a language processing unit (LPU), used for inference, and sold it as part of a cloud service or an on-premises hardware cluster. With Nvidia now owning the IP for LPUs, the GPU giant announced its own hardware cluster, the Nvidia Groq 3 LPX inference hardware system, at its GTC event in March. In response, Groq has pivoted to its neocloud business, it said. That business had been run by Madra after Groq acquired his AI data analytics company Definitive Intelligence, in 2024. It has grown to 13 data centers across North America, Europe, the Middle East and APAC and is serving over five million developers and thousands of AI companies, processing trillions of tokens each week, the company says. Groq has also been hiring replacement execs. It added Alan Rice as COO, previously at xAI and Meta, after a career in the U.S. Navy. It also added an entrepreneurial duo, Sinclair Schuller, who joins as CTO, and Rakesh Malhotra as CPO. They previously worked together at Apprenda, an enterprise cloud software company founded by Schuller; they then co-founded Nuvalence, a software-engineering firm acquired by EY in 2024. Malhotra previously spent about a decade working on Microsoft's cloud products. Whether Groq can succeed after almost selling itself depends on how competitive its inference cloud can remain, now that the key hardware IP is shared with Nvidia. Certainly, it has a shot. Inference-related tech is an area experiencing tremendous demand (and VC investment). But it's also seeing increasing innovation and competition. Still, others seem to have survived these sorts of deals. Scale AI's CEO Jason Droege told Forbes that business has rebounded after Meta did a $14.3 billion not-acqui-hire about a year ago, and that the company is on track to do $1 billion in revenue. In the big money game of AI, anything seems possible.

Groq's $650M raise rebuilds it after Nvidia

Groq has confirmed a $650mn raise to rebuild itself as an AI inference cloud, six months after Nvidia paid out its investors and hired away its founder. The Groq $650M round is a bet that purpose-built chips still beat GPUs. Groq spent years as one of the loudest challengers to Nvidia. Then, last December, Nvidia all but took it apart. Now the company has confirmed how it plans to rebuild. Groq raised $650mn in a round led by Disruptive and Infinitum, it said, with existing backers reinvesting alongside them. The raise was first reported in May. The company has now closed it, and named the team that will spend it. What Nvidia left behind In December, Nvidia struck a non-exclusive licensing deal for Groq's chip technology, in an arrangement reported to be worth around $20bn. It hired away founder and chief executive Jonathan Ross, president Sunny Madra and other engineers. Nvidia has since folded the technology into its own line-up. At its GTC event in March, it launched the Groq 3 LPX inference system, built on the licensed designs. That left Groq with its data centres, its software and an awkward question: what is it now? A pivot to the cloud The answer is inference, the business of running trained AI models rather than building them. Groq is leaning hard into what it calls its "neocloud" arm. It now operates 13 data centres across North America, Europe, the Middle East and Asia-Pacific. It says it serves more than five million developers and processes trillions of tokens each week across the sites that run those models. The new money will fit out that footprint with its latest hardware, including Nvidia's LPX system. Groq plans to quadruple total capacity to 200 megawatts by the end of 2027. New names at the top Adam Winter has taken over as chief executive, with Matt Eng as finance chief and Disruptive founder Alex Davis as chairman. Alan Rice joins as chief operating officer, after stints on xAI's Colossus project and at Meta's data centres, and an earlier career in US Navy submarines. From July, Sinclair Schuller becomes chief technology officer and Rakesh Malhotra chief product officer. The pair previously built and sold the enterprise software firms Apprenda and Nuvalence. Malhotra spent roughly a decade on Microsoft's cloud products. The same investors, again There is an unusual twist to the round. Many of the backers now writing cheques are the same ones Nvidia cashed out in December. Disruptive and Infinitum, which both hold board seats, led the deal. Earlier supporters have included Samsung, Cisco and BlackRock. Groq did not disclose a new valuation. It was last worth $6.9bn, after a $750mn round in September. A crowded, expensive bet Groq is wading into a brutal market. Demand for inference is climbing fast, and money is pouring in around it. AI-infrastructure startup Baseten recently raised $1.5bn at a valuation of up to $13bn. Frontier labs OpenAI and Anthropic keep pushing the cost of compute higher, which only sharpens the appetite for cheaper ways to run their models. The bottom line Groq's pitch is simple. As AI shifts from training models to running them, purpose-built inference chips should beat general-purpose GPUs on speed and cost. Inference, it argues, will eventually need 15 to 20 times more compute than training. The catch is just as simple. Groq now shares its core chip IP with Nvidia, the very giant it set out to beat. Whether a leaner, re-staffed Groq can still win the inference race is the $650mn question.

Inference chip startup Groq raises $650M to grow its cloud platform

Inference chip startup Groq raises $650M to grow its cloud platform Seven months after inking a $20 billion chip licensing deal with Nvidia Corp., Groq Inc. today announced that it has raised $650 million in funding. Growth investment firm Disruptive and hedge fund Infinitum led the round. Groq has developed a chip design called the LPU that's specifically optimized for artificial intelligence inference workloads. In December, Nvidia agreed to license the technologies that underpin the processor. It also hired several key Groq employees, including its founding chief executive. The transaction produced the Nvidia Grok LPU 3, an inference processor that the chip giant debuted in March. It ships as part of a rack-size, liquid-cooled appliance called the LPQ. The system includes 32 trays that each host three Groq LPU 3 units, one central processing unit and network equipment. The accelerators in an inference cluster each include a quartz crystal called a clock that regulates processing speeds. Clocks also play an important role in coordinating the flow of data between chips. When accelerators' clocks move out of sync with each other, data traffic slows down, which negatively impacts AI model response times. The LPU 3 includes a feature that automatically fixes clock drift to avoid data traffic bottlenecks. According to Nvidia, the chip includes 92 lanes that can each move data to other processors at a speed of 112 gigabits per second. That translates to 2.5 terabits per second of bidirectional bandwidth. Accelerating the flow of data between chips is not the only way the LPU 3 speeds up inference workloads. The processor ships with 500 megabytes of onboard SRAM, a high-speed memory variety. SRAM is more performant than the off-chip RAM that other AI accelerators use to store data, which translates into faster inference. Groq operates an LPU-powered cloud platform that companies can use to run inference workloads. The company disclosed today that the platform is processing trillions of tokens per week for five million developers. Groq's cloud runs across 13 data centers spanning multiple continents. The company will use the proceeds from its funding round to grow its inference capacity with the goal of reaching 200 megawatts by 2027. According to Grow, some of the new processing power will be provided by the LPX, the liquid-cooled LPU 3 appliance that Nvidia debuted in March. Other cloud operators can theoretically build LPQ-powered inference services of their own. One way Groq could set itself apart from such potential rivals is by extending its platform with new services such as managed databases. Other AI-focused cloud providers, notably CoreWeave Holdings Inc., have also broadened their focus beyond infrastructure to higher-level services.