HomeArticle

11.2 billion. The first chip unicorn in 2026 has arrived.

融资中国2026-03-19 14:00
A unicorn in the heat dissipation field is born, accelerating the development of AI hardware infrastructure.

In 2026, the battlefield of the AI arms race has spread from the model and application layers to the hardware infrastructure.

Recently, a company specializing in chip cooling quietly accomplished a major feat. Frore Systems announced the completion of a $143 million Series D financing round, reaching a valuation of $1.64 billion and becoming one of the earliest new unicorns in the deep - tech field in 2026. This company doesn't manufacture chips; instead, it develops liquid - cooling systems to prevent chips from overheating during operation. Its products cover the three major chip suppliers: NVIDIA, Qualcomm, and AMD.

Frore's story has a landmark turning point. About two years ago, Jensen Huang watched their technology demonstration and gave a suggestion: switch to liquid cooling. Frore actually made the switch. This decision, in hindsight, hit a precise time window - as the density of AI computing power continues to rise, air cooling can no longer keep up, and liquid cooling has become the standard for data centers, and this demand is far from reaching its peak.

During the same period as this financing round, new unicorns in the chip sector emerged in quick succession. Positron reached a valuation of $1 billion, Recursive Intelligence started with a $4 - billion valuation, and Eridu completed a $200 million Series A financing. The investment window for AI hardware infrastructure is fully opening. Among these companies, Frore has the most unique logic - it doesn't bet on the success or failure of a single chip manufacturer but positions itself in a place that everyone needs. The more intense the competition in the chip sector, the larger its customer base becomes.

Eight years, $340 million in total financing, and a $1.64 - billion valuation. Frore's path is one of the clearest examples of the investment logic for hardware infrastructure in the AI era.

Another Chip Unicorn

In 2026, the first batch of AI hardware unicorns have emerged.

Just recently, the chip - cooling company Frore Systems announced the completion of a $143 million Series D financing round, led by MVP Ventures, with participation from Fidelity, Mayfield, Addition, Qualcomm Ventures, Alumni Ventures, and other institutions. After this round of financing, Frore's valuation reached $1.64 billion, equivalent to approximately 11.2 billion yuan. The total cumulative financing has risen to $340 million.

Frore doesn't manufacture chips; instead, it solves the problem that occurs after chips start running: cooling.

More precisely, Frore develops liquid - cooling systems, and its products are compatible with mainstream chips and development boards from NVIDIA, Qualcomm, AMD, etc. The company was founded in 2016 by two former Qualcomm engineers and is headquartered in San Jose, California, USA. It has been in operation for exactly eight years.

In the early days of its establishment, Frore's technical focus was on air - cooling for mobile phones and small fanless electronic devices. The market was small but niche. At that time, large - scale models hadn't yet triggered a huge demand for computing power, and cooling was an unremarkable supporting part of the semiconductor industry chain, and few people were willing to focus on this area. Frore spent several years in this niche, accumulating core technical capabilities in cooling small devices.

The real turning point for the company came from a product demonstration.

About two years ago, NVIDIA CEO Jensen Huang watched Frore's technology demo. After watching, Huang gave a suggestion: switch to liquid cooling. These four words represented a substantial change in direction for Frore. The core customers for liquid - cooling are data centers and high - performance AI computing systems, and the scale of this market is on a completely different level from that of mobile - phone cooling. Frore accepted this suggestion, made a transformation, and re - developed a liquid - cooling product line for AI chips.

This decision was precisely timed. From 2024 to 2025, with the explosion of demand for large - scale AI model training and inference, the power consumption of high - end GPUs such as NVIDIA's H100 and H200 continued to rise. The thermal design power (TDP) of a single card has exceeded 700 watts, and the power consumption of a whole cabinet can easily exceed 100 kilowatts. Cooling is no longer an "ancillary project" for data centers; instead, it has become one of the core bottlenecks restricting the density of AI computing power. Air cooling can no longer keep up, and liquid cooling has become the standard. This is a physical constraint, not a business choice. When data - center operators and cloud providers purchase AI servers, the compatibility of liquid - cooling solutions has become a mandatory consideration.

Frore caught this window. Currently, the company's products are compatible with multiple NVIDIA chips and development boards, and it has also developed cooling solutions for Qualcomm and AMD, covering the three major suppliers in the current AI chip market. This multi - platform compatibility strategy reduces the dependence on a single chip manufacturer and means that as long as the demand for AI computing power continues to grow, Frore's potential customer base will not shrink.

The structure of the investors in this financing round is worth noting. Fidelity represents the judgment of large - scale financial institutions on the commercial viability of this sector. Mayfield and Addition are active technology venture funds in Silicon Valley, and Qualcomm Ventures represents the direct entry of an industry player. Qualcomm is both a product - compatibility target for Frore and a participating investor in this round. This "customer - as - shareholder" structure is usually a sign of relatively solid commercial progress in the hardware sector.

The simultaneous entry of "industry + finance" usually means that a company has reached the critical point of moving from the technology - validation phase to the business - expansion phase. Institutional investors see a clear path to financial returns, and industry players see the value of strategic synergy. Their joint action indicates that Frore's current product and customer status can support its valuation logic.

Eight years, $340 million in total financing, and a $1.64 - billion valuation. Frore used a hardware story that few others were willing to tell to break into the ranks of unicorns during the most crowded period in the AI sector.

Doesn't Make Chips, but Everyone Needs It

Frore's financing round is not an isolated case in the 2026 deep - tech investment landscape.

During the same period when Frore announced its Series D financing, new unicorns in the chip sector were emerging in quick succession. Positron, a competitor to NVIDIA, completed a financing round in February, reaching a valuation of $1 billion. Recursive Intelligence started with a $4 - billion valuation. Eridu, which focuses on AI network chips, just completed a $200 million Series A financing. Within just a few months, several high - valuation companies have emerged in the deep - tech chip sector, a frequency that has been rare in the past few years.

Why is capital flooding in at this time? The answer is simple: the AI computing - power arms race has moved from the software layer to the hardware layer, and the investment window for the hardware layer is at a critical turning point from early - stage validation to large - scale implementation. In the past few years, AI investment in the software layer has produced a number of high - valuation companies, but opportunities in the hardware layer have lagged behind because hardware has a longer R & D cycle and higher validation thresholds. Now, as the demand for computing power from AI applications moves from the laboratory to large - scale deployment, the commercial viability of the hardware layer is becoming clearer, and naturally, investors' interest is flowing in.

In this wave, most people's attention is focused on chip design itself. The question is who can make a GPU more powerful than NVIDIA's and who can achieve a higher energy - efficiency ratio in specific inference scenarios. This is a correct but crowded path, with high entry barriers, long R & D cycles, and obvious risks of competing directly with NVIDIA. Frore's chosen path is completely different. It doesn't participate in the competition of chip design but stands outside the chips, solving the physical problems that chips inevitably face during operation. This choice has a key advantage in logic: there are only a few winners in the chip - design sector, but the customers in the cooling sector are everyone.

This logic is very straightforward in business. NVIDIA sells chips, Qualcomm sells chips, and AMD sells chips, but all three of their chips need cooling. By providing products to all three, Frore has positioned itself in a place where "everyone needs me regardless of who wins." In the highly uncertain competitive landscape of the AI chip sector, this cross - platform neutral position is actually a more solid moat. The more competitors there are and the more vibrant the market is, the larger Frore's potential customer base becomes.

From an investment perspective, it's still uncertain who the final winner in the AI chip sector will be, but it's much more certain that the density of AI computing power will continue to increase and the demand for cooling will continue to grow. Investing in a company that doesn't bet on the success or failure of a single chip manufacturer but benefits from the growth of the entire sector naturally has a better risk profile. This is why this financing round was able to attract both industrial capital and financial capital - the former looks at strategic synergy, and the latter looks at financial returns, and both logics hold true for Frore. Industrial players invest because Frore's products directly enhance the competitiveness of their own chips, and financial players invest because the growth curve of this demand is clear enough. Although the two types of capital have different demands, they find a common ground in Frore.

It's worth noting that there hasn't been a large - scale iconic company in the cooling sector yet. Frore becoming a unicorn with a $1.64 - billion valuation is one of the most important financing events in this niche sector to date. This means that the sector is still in its relatively early stages, the landscape is not yet set, and the first - mover advantage is extremely valuable. Being the first to complete product compatibility with the three major chip manufacturers, NVIDIA, Qualcomm, and AMD, means that Frore has established a presence on the procurement lists of potential customers. It would take later entrants more time and resources to replicate this position.

Jensen Huang's words "switch to liquid cooling" two years ago now seem to be more than just a product suggestion; they're like an entry ticket. They helped Frore complete its technical reserves and product development before the real explosion of demand for liquid cooling. When the market window opened, the company already had products to sell. This rhythm of "preparing two years in advance" is the most difficult - to - replicate and most valuable competitive advantage in hardware startups. Timing is even more difficult to grasp in the hardware field than in the software field. If you're too early, you'll burn through money; if you're too late, you'll miss the opportunity. Frore hit the right note.

The Boom in the Cooling Sector Is Just Beginning

Frore becoming a unicorn is a result, but what's more worth paying attention to is the direction it points to. The investment window for AI hardware infrastructure is fully opening.

In the past few years, the main battlefield of AI investment has been in the model and application layers. Large - scale model companies have received the largest share of financing, and application - layer startups have shown the fastest growth curves. Opportunities in the hardware layer have always existed, but it's more difficult to tell a good story. Hardware has a long cycle, slow validation, and a complex customer - decision - making process, and its data isn't as impressive as that of software companies. This has led to a structural lag: the AI bubble in the software layer has started to deflate, while the investment enthusiasm in the hardware layer is just picking up.

This shift in rhythm is essentially driven by the demand side. The parameter scale of AI models continues to grow, and the scale of inference deployment is also expanding rapidly. As a result, the computing - power density of data centers is constantly increasing. The physical consequence of the increasing computing - power density is that the heat generated per unit space is getting higher and higher. This problem can't be solved by software optimization; an answer must be found at the hardware layer. Cooling is the most certain part of this answer.

From a longer - term perspective, the penetration rate of liquid cooling is still in its relatively early stages. Traditional data centers mainly use air cooling, and the large - scale deployment of liquid cooling has only really accelerated in the past two or three years. As the demand for the construction and renovation of AI data centers continues to be released, there is still a large growth space for the procurement scale of liquid - cooling solutions. The financing that Frore has just received will most likely be used to accelerate product - line coverage and expand the scale of customer deployment, so as to quickly establish a deeper market share during the window period of the accelerating penetration of liquid cooling.

Another characteristic of this sector is strong customer stickiness. Once a cooling system is compatible with and deployed for a certain chip or server architecture, the replacement cost is relatively high. Data - center operators won't frequently change their cooling solutions because this involves the re - design and validation cycle of the entire cabinet. This means that once Frore enters the procurement system of a customer, the customer churn rate will be relatively low, and its revenue predictability is stronger than that of many hardware companies. This customer structure is one of the important foundations for supporting its high valuation.

Looking at the entire AI hardware ecosystem, the competitive landscape of the cooling sector is still relatively fragmented at present. Large - scale server manufacturers have their own cooling solutions, but there aren't many independent companies focusing on liquid cooling for AI chips. Frore is the one with the highest level of completion and the largest financing scale. This landscape will most likely change in the next two or three years - as the market scale expands, more players will enter, and the competition will intensify. However, the first - mover advantage in this sector is real. Establishing product - compatibility relationships with NVIDIA, Qualcomm, and AMD simultaneously is not something that can be easily replicated. It requires time, technical accumulation, and mutual trust between the two parties. Frore has already achieved this, and it won't be easy for later entrants to catch up.

From a more macroscopic perspective, Frore's story reveals an underestimated logic in AI hardware investment: you don't have to compete in the hottest sectors. Finding a position where "everyone needs me regardless of who wins" may actually be a more stable choice. The beneficiaries of the computing - power arms race are not only those making guns but also those making bullets, gunstocks, and heat sinks. In the more competitive main sectors, the demand for supporting sectors is more certain, and the valuation logic is clearer.

For investors, this is also a signal worth paying attention to. The investment opportunities in AI hardware infrastructure are moving from the "story - telling" stage to the "implementation - based" stage. Companies that can present real customers, real compatibility, and real deployment data will have stronger bargaining power in the upcoming financing environment. Frore took eight years to reach this point, not by telling a good story but by having a set of products in operation. This difference is more important than ever in the 2026 hardware investment market.

Cooling used to be one of the least glamorous parts of the semiconductor industry chain. Now, it's becoming one of the most certain businesses in the AI era. Frore is not the end but a sign that this sector is truly starting.

This article is from the WeChat official account "Rongzhong Finance" (ID: thecapital), written by Lv Jingzhi and edited by Wu Ren. It is published by 36Kr with authorization.