The Hottest Problem in AI — Literally

While the AI infrastructure race has fixated on chips, power, and networking, a quieter but equally critical bottleneck has emerged: heat. As NVIDIA's next-generation GPU racks demand upwards of 120 kW of cooling capacity — a 7-9x increase over traditional configurations — the physics of thermal management have become an existential constraint on AI performance.

Enter Frore Systems, an eight-year-old deep tech startup out of Silicon Valley that just raised $143 million in Series D funding at a $1.64 billion valuation, officially crossing into unicorn territory. The round was led by MVP Ventures, with participation from Fidelity Management & Research, Qualcomm Ventures, Mayfield Fund, Top Tier, Addition, Clear Ventures, StepStone Group, and Alumni Ventures. Total capital raised now stands at $340 million.

The message from investors is clear: cooling is no longer a commodity — it's foundational infrastructure for the AI era.

From Qualcomm Labs to Cooling Revolution

Frore Systems was born from a "lightbulb moment" in 2018, when co-founders Dr. Seshu Madhavapeddy (CEO) and Dr. Surya Ganti (CTO) were working together on Qualcomm's ultrasonic fingerprint sensor business. Both brought formidable technical pedigrees — Madhavapeddy as a serial entrepreneur with two previous successful startups, and Ganti with deep R&D experience spanning Qualcomm, GE Research, and MIT.

Their initial breakthrough was AirJet®, the world's first solid-state active cooling chip. Unlike traditional fans, AirJet uses micro-scale vibrating membranes to create airflow, delivering 2x processor performance improvement in fanless devices — all while being silent, dustproof, and water-resistant. The product was a technical marvel for smartphones, tablets, and ultra-thin laptops.

But the company's trajectory shifted dramatically after a fateful encounter with NVIDIA CEO Jensen Huang.

The Jensen Huang Pivot

Approximately two years ago, Huang personally received a demonstration of Frore's AirJet technology. Impressed by the underlying physics, he made a suggestion that would reshape the company's future: develop liquid-cooling solutions for AI chips and data center systems.

The timing was prescient. AI data centers were beginning to hit thermal walls that traditional air cooling simply couldn't breach. Frore's team took Huang's advice and executed a bold pivot, adapting the core principles of their solid-state cooling technology into LiquidJet™ — a direct liquid cooling coldplate platform — and later LiquidJet™ Nexus, an integrated cooling system designed for next-generation AI server architectures.

This strategic reorientation from edge devices to data center infrastructure proved to be the catalyst for the company's unicorn moment.

The Technology: Why Frore's Approach Is Different

Frore's LiquidJet platform represents a fundamentally different approach to liquid cooling. Rather than using conventional flat coldplates with simple channel designs, Frore applies semiconductor manufacturing processes to metal wafers, creating 3D short-loop jet channels positioned precisely where GPU hotspots occur.

The performance numbers tell a compelling story:

• 75% higher heat transfer efficiency compared to conventional coldplates
• 8°C cooler GPU operation, directly enabling higher sustained clock speeds
• 4% increase in AI tokens per second — a meaningful throughput gain at scale
• 10% PUE (Power Usage Effectiveness) reduction, translating to significant energy savings
• 55% weight reduction in the thermal stack

The flagship LiquidJet Nexus goes further, integrating multiple LiquidJet coldplates into a unified system designed for NVIDIA's Kyber ½U tray architecture. It delivers 2x compute density per rack while supporting 53°C inlet temperatures — a threshold that eliminates the need for mechanical chillers entirely. For hyperscale operators spending billions on cooling infrastructure, this is a game-changing proposition.

Critically, Frore has designed its platform to scale with the AI compute roadmap. The company says LiquidJet is engineered to support NVIDIA's upcoming Rubin, Rubin Ultra, and Feynman architectures — the latter expected to exceed 4,000W per chip. This forward-looking compatibility positions Frore not as a point solution but as a platform play that grows with the industry.

The Market: A $128 Billion Opportunity

Frore's timing could hardly be better. The global data center cooling market is projected to grow from $31.4 billion in 2026 to $128.3 billion by 2033, representing a 22.3% CAGR. Within this, the thermal management segment alone is expected to reach $33.7 billion by 2034, growing at 9.4% annually.

The demand driver is straightforward: AI chips are getting dramatically more power-hungry with each generation, and air cooling has hit a physical ceiling. NVIDIA's GB200 NVL72 rack configuration requires 120-132 kW of cooling capacity — making liquid cooling not optional but mandatory. As Madhavapeddy put it: "Cooling has become the single greatest limiter of AI performance."

The customer base spans the largest buyers in tech: hyperscale cloud providers (AWS, Azure, GCP), sovereign AI initiatives building national computing infrastructure, and OEMs manufacturing AI server hardware. While Frore hasn't disclosed specific customer names, the breadth of use cases — from NVIDIA GPU cooling to custom hyperscaler ASIC designs — suggests significant pipeline development.

Competitive Landscape: Standing Out in a Crowded Field

Frore enters a competitive market with both established players and well-funded startups:

• Vertiv Group — the current market leader by revenue in data center thermal management
• CoolIT Systems and Asetek — established liquid cooling providers with strong OEM relationships
• Accelsius — raised $65M Series B in January 2026 for two-phase liquid cooling
• Eaton — acquired Boyd Thermal for $9.5 billion, signaling massive corporate interest in the space
• Phononic and xMEMS — competing in solid-state cooling with thermoelectric and piezo approaches

Frore's differentiation lies in its semiconductor-grade manufacturing approach to cooling hardware. By treating coldplates with the same precision as chip fabrication, the company achieves thermal performance that conventional machined or stamped metal solutions cannot match. The 3D jet channel architecture, customized to specific chip layouts, provides a level of thermal optimization that generic solutions miss.

The company also benefits from a unique full-stack thermal portfolio — AirJet for edge computing, LiquidJet for data center GPUs, and LiquidJet Nexus for integrated server systems — making it a one-stop thermal partner across the entire AI compute spectrum.

The Investors: Why They Bet on Frore

Lead investor MVP Ventures, a Menlo Park-based firm founded in 2020, has built a strong track record in deep tech with nine unicorns in its portfolio, including Anduril and Magic Eden. The firm's focus on "differentiated and defensible technology" aligns perfectly with Frore's semiconductor-derived cooling IP.

Fidelity Management & Research's participation is particularly noteworthy. When one of the world's largest asset managers backs a hardware startup, it signals institutional-grade conviction in the market opportunity. Qualcomm Ventures' continued investment reflects both the founders' Qualcomm lineage and the strategic importance of thermal management across the semiconductor value chain.

The investor consortium's diversity — spanning early-stage VCs (MVP, Mayfield), growth equity (Addition, StepStone), strategic corporates (Qualcomm Ventures), and institutional capital (Fidelity) — suggests broad consensus that thermal infrastructure has reached an inflection point.

What's Next: Manufacturing Scale and the Road Ahead

Frore plans to deploy the $143 million across three strategic priorities:

1. Manufacturing expansion beyond its current Taiwan-only production base, likely targeting locations that serve both geographic diversification and supply chain resilience goals
2. Go-to-market acceleration targeting hyperscale cloud operators, sovereign AI programs, and server OEMs
3. Next-generation R&D to stay ahead of the compute curve, particularly for the 4,000W+ chip designs expected in the Feynman generation

The manufacturing expansion is critical. As AI infrastructure buildouts accelerate globally — with projections of 3x compute demand growth through 2030 — Frore's ability to scale production will determine whether it captures a meaningful share of the cooling opportunity or remains a niche player.

The Bigger Picture

Frore Systems' unicorn moment represents more than a single company's success. It marks the emergence of thermal management as the third pillar of AI infrastructure, alongside chips and power. For years, cooling was treated as a solved problem — commodity fans and heat sinks that came as an afterthought. The AI era has shattered that assumption.

The key variables to watch going forward: the pace of Frore's manufacturing scale-up, integration progress with NVIDIA's next-generation architectures, and whether the company can secure anchor contracts with hyperscale customers. In a market where the laws of physics are as important as the laws of software, Frore Systems has positioned itself at a critical intersection — and investors are betting $340 million that cooling is, indeed, the new compute.