Groq Raises $350M From Nvidia After Selling Its Chip Soul

Groq just closed a $350 million Series A led by Dallas investment firm Disruptive, with planned participation from Nvidia, capping a strange twelve months in which the company went from Nvidia challenger to Nvidia customer. The new capital values the company at $3.5 billion, down from the $6.9 billion Groq was valued at last September, just a few months before Nvidia hired the startup's founder and CEO, Jonathan Ross, and other top talent as part of a $20 billion licensing deal that the company paid out to investors.

Combined with the $650 million raised in June 2026, this brings recent funding to $1 billion in roughly two months. The official announcement frames the round as fuel for a build-out that will more than triple the company's compute footprint next year.

From LPU insurgent to Nvidia neocloud

To understand why a well-known chip startup is suddenly raising a Series A, you have to rewind. Groq was focused on building its own chips, dubbed LPUs (language processing units), to compete with Nvidia on inference, the type of compute needed to run AI workloads in real time. The LPU is a deterministic, SRAM-heavy accelerator built around streaming tensors through on-chip memory, which is what gave Groq its reputation for very low-latency token generation.

Then came the December 2025 deal that reshaped the company. Nvidia licensed all of the company's technology on Christmas Eve last year, in a $20 billion deal that also saw the GPU giant hire founder and then-CEO Jonathan Ross, along with president Sunny Madra and other members of the Groq team. Nvidia plans to launch its own Groq-based hardware later this year, and then develop more advanced chips based on Groq, but with an expanded team and utilizing Nvidia IP, including NV Link.

What was left of Groq had cash, data centers, customers, and a brand, but no chip roadmap. So it rebuilt as something else. After it lost its star team, Groq shifted from being a pure AI chipmaker into a cloud and data center provider that operates Nvidia systems. Groq is now an Nvidia Cloud Partner (NCP), a certification covering the design, deployment and operation of Nvidia accelerated computing in line with Nvidia's reference architecture and operational standards.

The numbers that matter

• Round size: $350M Series A
• Valuation: $3.5B, roughly half the $6.9B September 2025 peak
• Lead investor: Disruptive, whose founder Alex Davis is now Groq's executive chairman
• Strategic investor: Nvidia, planned participation
• Two-month total raised: $1B ($650M in June plus this $350M)
• Footprint: 13 data centers across North America, Europe, the Middle East, and Asia Pacific, serving more than 6 million developers, enterprises, and AI-native companies
• Capacity plan: expand capacity from 54 megawatts to more than 200 megawatts by 2027

Groq is careful about the label. A spokesperson told TechCrunch that despite the difference in valuation, the company doesn't see it as a down round, but rather as establishing a new valuation for the post-Nvidia-licensing-deal version of Groq. Fair enough on paper, since the entity investors are buying into today is materially different from the one that raised at $6.9B. But cutting a valuation in half inside a year still tells you something about how the market prices founders, engineers, and IP.

Why this happened now

Inference demand is the tailwind everyone is chasing. Executive chairman Alex Davis put the thesis bluntly: "Inference will without a doubt become the largest and most critical layer of AI infrastructure." Every major model provider needs somewhere to serve tokens, and hyperscaler capacity is stretched thin. That is the opening Groq is running at.

The Nvidia angle is the twist. Nvidia now owns the technology Groq built, employs the people who built it, and is also writing a check to the shell that remains. That gives Groq preferred access to Nvidia hardware allocations, and it gives Nvidia another distribution channel for GPUs outside the hyperscalers. Groq's pivot places it alongside other Nvidia-backed neoclouds, including CoreWeave, Lambda, and Nebius, as investors continue evaluating the long-term profitability of infrastructure-heavy AI cloud businesses.

The real story behind the headline

Two facts sit in tension. Groq's pitch now rests on a single argument: that running AI models in production will eventually dwarf the scale and spending of training them. It is not alone in making that case, and it is no longer the best-funded name making it. CoreWeave is public. Lambda and Nebius are scaling fast. And on the pure-silicon side, new entrants are picking up the anti-Nvidia banner Groq set down. Rival inference-chip startups such as Fractile have raised at buoyant valuations, and one London challenger recently tripled its worth to $3.3bn while betting openly against Nvidia.

What Groq keeps is real: a working developer platform with millions of users, trillions of tokens per week, global data center presence, and a partnership with the vendor that now sets the pace of the industry. What it gave up is the differentiated silicon story that let it charge a premium in the first place.

Winners, losers, and second-order effects

1. Winners: Nvidia, which extracted the IP and talent it wanted, is now backing a distribution partner, and effectively neutralized a chip competitor. Disruptive doubled down and holds the executive chairmanship. Developers using Groq's API get more capacity and continuity.
2. Losers on paper: Employees and earlier investors who held equity at the $6.9B mark, though the licensing payout softened the blow. The independent inference-silicon thesis loses one of its most visible flag-bearers.
3. Second-order: Every remaining inference-chip startup now has to answer whether their exit is a real IPO or a Groq-style acqui-license. That will affect how they price rounds and structure IP.
4. For customers: The neocloud tier is consolidating around a handful of Nvidia-preferred operators. Expect tighter GPU supply for smaller shops and more aggressive pricing wars among the top five.

What it means if you build on Groq

If you already use the Groq API for fast inference on open models, the immediate signal is stability: more money, a locked-in Nvidia relationship, and a stated plan to nearly quadruple power capacity. Latency-sensitive workloads that depended on LPU behavior may look different over time as the platform migrates toward Nvidia accelerated computing for training and inference workloads, but the developer-facing endpoint is meant to stay put. If you were evaluating Groq specifically because it was not Nvidia, that reason no longer holds.

The financing is not fully closed yet. Completion of the Series A remains subject to customary closing conditions. Assuming it lands, the question hanging over the sector is the one TechCrunch flagged directly: while inference is in high demand as enterprises scale AI workloads, it's an open question whether neoclouds will be a profitable enough business to provide returns on their considerable investment in the long term. Groq now has a billion dollars and a year or two to answer it.