Groq raised $350M at a $3.5B valuation to complete its pivot from chipmaker to neocloud

Groq raised $350 million in a Series A round at a $3.5 billion valuation, the company said Sunday. Investment firm Disruptive led the round, with planned participation from NVIDIA Corporation.

The valuation is half of what Groq commanded eleven months ago. In September 2025, the company was worth $6.9 billion.

Two months later, NVIDIA hired Groq founder Jonathan Ross. The deal included a $20 billion licensing agreement that paid out to investors.

Groq does not call this a down round. A company spokesperson said the new price establishes a valuation for the "post-NVIDIA-licensing-deal version of Groq."

The money trail

This is Groq's second major raise in eight weeks. In June 2026, the company brought in $650 million to begin its pivot. Combined with Sunday's round, Groq has secured $1 billion in fresh capital since early summer.

The company spent its first life building custom chips called LPUs (language processing units). Those chips competed with NVIDIA on inference. Inference is the compute that runs AI models in real time.

After NVIDIA absorbed Groq's founding team, the company rebuilt itself around NVIDIA's GPUs (graphics processing units) instead.

Alex Davis is Groq's executive chairman and the CEO of Disruptive. He framed the bet in a statement: "Inference will without a doubt become the largest and most critical layer of AI infrastructure."

Groq now operates 13 data centers across North America, Europe, the Middle East, and Asia Pacific. More than six million developers use the platform. The company plans to scale from 54 megawatts (MW) to more than 200 MW in 2027.

The neocloud economics question

Groq now competes directly in the neocloud market. Neoclouds rent out GPU-powered infrastructure for AI workloads. The field includes CoreWeave, Lambda, and Nebius.

All of them depend on NVIDIA hardware.

Several have received billions in NVIDIA investment.

The unit economics remain unproven. CoreWeave, the largest neocloud, reported strong second-quarter revenue growth. It landed contracts with Meta Platforms and Anthropic.

Investors still worry about high capital spending, heavy debt, and fast-depreciating hardware.

Groq's financials are private. The company has not disclosed revenue, burn rate, or a path to profitability.

One billion dollars buys runway. But running 13 data centers on NVIDIA hardware at 54 MW burns cash every day. The jump to 200 MW will require more.

What NVIDIA gets

NVIDIA's planned participation is strategic, not charitable. Every neocloud that runs NVIDIA hardware extends NVIDIA's grip on the inference layer.

Groq, CoreWeave, Lambda, Nebius: they all compete with each other. None of them compete with NVIDIA.

The licensing deal that absorbed Groq's founder gave NVIDIA both the talent and the intellectual property. What remains is a customer. Groq brings a brand, six million developers, and a data center footprint. NVIDIA would rather have that inside its ecosystem than outside it.

I find it telling. The same company that licensed away Groq's core technology is now investing in the shell that remains.

The competitive field

The neocloud market is not a winner-take-all race. Enterprises spread workloads across providers to avoid lock-in. But scale matters: the operator with the most megawatts and the widest geographic reach wins the largest contracts.

Groq's 13 data centers across four regions give it a head start over most competitors. The 200 MW target for 2027 would put it in the same capacity range as CoreWeave's planned expansion. The question is whether Groq can fill that capacity with paying customers before the cash runs out.

The broader signal: a company that lost its founder and its chip team just raised $1 billion on a pivot. That tells you how much capital chases inference infrastructure. It does not tell you whether the returns will justify it.

Source