Groq raises $650M to scale AI inference cloud to 200 MW by 2027

Groq, the AI inference hardware and cloud company known for its Language Processing Unit (LPU) technology, announced on June 22, 2026 that it has raised $650 million in new funding led by Disruptive and Infinitum. The raise is aimed at scaling Groq's global AI inference cloud toward 200 megawatts of compute capacity by the end of 2027.

What's new

The $650 million round was co-led by Disruptive, whose CEO Alex Davis serves as Groq's Chairman, and Infinitum, led by Founder and CIO John Yetimoglu. Both firms are reinvesting, reflecting continued confidence in Groq's commercial trajectory. No post-money valuation was disclosed.

Key targets from the announcement:

• Compute target: Scale to 200 MW of AI inference capacity by end of 2027
• Scope: Global expansion of Groq's AI inference cloud infrastructure
• Use of capital: Accelerating operational scale, not just hardware development

Davis framed the investment thesis in strategic terms: "Groq has spent years building the technology, infrastructure, and operational expertise required for the next phase of AI. We believe that combination positions Groq to become a foundational layer of the AI economy."

Context

Groq built its reputation on purpose-built inference hardware — the Language Processing Unit — designed to run large language models at high throughput and low latency. This architecture is optimized specifically for token generation rather than training, which is a fundamentally different workload profile from GPU-based infrastructure.

The company has expanded from serving developers to serving enterprise and government customers, running its GroqCloud inference platform as a managed service. The inference market has outpaced early projections as AI adoption has scaled: instead of training being the dominant compute workload, inference — serving real-time requests from users and AI agents — has grown to match or exceed it.

Groq competes directly with hyperscale cloud providers (AWS, Google Cloud, Microsoft Azure) and other inference-focused services built on NVIDIA's GPU ecosystem.

Why it matters

The investment thesis laid out by Yetimoglu is blunt: "We believe inference will become the largest infrastructure market in technology." If that proves correct, a company with purpose-built hardware, existing operational scale, and enterprise relationships has a structural advantage over general-purpose cloud providers adapting GPU fleets to inference workloads.

The 200 MW target by 2027 is a meaningful capacity commitment. At frontier model serving scale, that represents the ability to handle millions of concurrent inference requests. For Groq, demonstrating it can build and operate infrastructure at that scale — not just ship benchmark results — is the proof point investors and customers are watching.