
Firmus Grid有限公司首次公开募股的搁浅正在亚洲蓬勃发展的股票发售市场引起反响,这使得投资者对依赖未来收益的人工智能基础设施交易变得更加挑剔。
The collapse of Firmus Grid Ltd.’s initial public offering is reverberating through Asia’s booming market for share sales, making investors far pickier about artificial intelligence infrastructure deals that hinge on future earnings.
这家得到英伟达支持的数据中心运营商于周五撤回了一项至多筹资55亿美元的交易,此前投资者对该公司估值飙升至300亿美元望而却步。这家澳大利亚公司目前正在与现有投资者探讨一轮私募。
The Nvidia-backed data center operator on Friday pulled a deal to raise as much as $5.5 billion after investors balked at its steep increase in valuation to $30 billion. The Australian company is now exploring a private round with existing investors.
Firmus的惨败表明,在经历了由寻求为人工智能建设筹集资金的公司推动的、持续一年多的持续交易热潮之后,投资者变得更有鉴别力。根据彭博社汇编的数据,今年亚太地区的首次公开募股、配售和大宗交易已筹集近2500亿美元,加上2025年约2600亿美元的总额。
Firmus’ flop is a sign of investors turning more discerning after more than a year of relentless dealmaking, driven by companies looking to raise funds for the AI buildout. IPOs, placements and block trades have raised nearly $250 billion in the Asia-Pacific region this year, adding to the 2025 tally of about $260 billion, according to data compiled by Bloomberg.
“这是投资者在询问谁来买单,以及资本成本是多少,”悉尼Private Portfolio Managers的投资组合经理马克斯·赫伦-维拉科特(Max Herron-Vellacott)表示,他认为在做出投资决策时,风险与回报的计算正受到更严格的审视。“这抬高了每笔交易的门槛,投资者对乐观假设的容忍度将大大降低。”一些近期的上市表现出了日益不安的迹象。
“It’s investors asking who pays for it, and at what cost of capital,” said Max Herron-Vellacott, a portfolio manager at Private Portfolio Managers in Sydney, who see the risk versus return calculus getting greater scrutiny when making investment decisions. “This raises the bar for every deal, and investors will be far less forgiving of optimistic assumptions.”Some recent debuts have shown signs of the creeping unease.
自7月在香港上市以来,苹果公司供应商立讯精密工业股份有限公司已暴跌20%,人工智能基础设施公司中际旭创股份有限公司下跌了约7%。智谱等中国大模型制造商和MiniMax等公司较历史高点已回调超过70%。
Apple Inc. supplier Luxshare Precision Industry Co. has plunged 20% and AI infrastructure firm Zhongji Innolight Co. is down about 7% since their Hong Kong debuts in July. Chinese model makers like Z.AI Co. and MiniMax Group Inc. have corrected more than 70% from their record highs.
投资者开始划一条硬线,区分那些拥有经证明的业绩记录且能够实现变现的公司,与那些处于生命周期早期、更渴望现金的阶段的公司。
Investors are starting to draw a hard line between companies that have a proven track record and are able to monetize, versus those that are in the earlier, more cash-hungry phase of their life cycle.
纽约大学法学院全球公共投资基金论坛执行主任温斯顿·马(Winston Ma)表示,对于像Firmus这样的数据中心运营商而言,分水岭在于产生即时收益的运营中设施,以及由数十亿美元债务融资启动的多吉瓦新项目。
For data center operators like Firmus, the division is between operational facilities that generate immediate yield and multi-gigawatt new projects initiated by billions in debt financing, said Winston Ma, executive director of the Global Public Investment Funds Forum at NYU School of Law.
“Firmus的撤回是一个明确的信号,表明公开市场正在对人工智能基础设施重新定价,”兼任Dragon Global Family Office合伙人的马说。“在预计将建设下一波人工智能数据中心大部分产能的亚洲,这抬高了任何首次公开募股的门槛。保荐人将需要更强有力的合同承购、对通电兆瓦更清晰的可见性,以及更保守的杠杆假设。”对于Firmus而言,在其912兆瓦的规划项目中,仅建成了46兆瓦。投资者被要求根据取决于数据中心成功交付的未来收益来对该公司进行估值。
“The Firmus withdrawal is a clear signal that public markets are repricing AI infrastructure,” said Ma, who’s also a partner at Dragon Global Family Office. “In Asia, where much of the next wave of AI data-center capacity is expected to be built, this raises the bar for any IPO. Sponsors will need stronger contracted off-take, clearer visibility on energized megawatts, and more conservative leverage assumptions.”For Firmus, of its 912 megawatt pipeline, just 46MW has been built. Investors were being asked to value the company on future earnings contingent on the data centers being successfully delivered.
投资者将其中许多公司与之进行比较的所谓新云公司,例如CoreWeave公司和Nebius集团公司,是美国杠杆率最高、波动最大的人工智能股票之一。此外,尽管Nscale有限公司和软银集团支持的SB Energy公司上个月申请在美国上市,但它们尚未开始推介各自的交易。
The so-called neoclouds that investors are comparing many of these companies to, such as CoreWeave Inc. and Nebius Group NV, are among the most leveraged and volatile AI stocks in the US. Also, though Nscale Ltd. and SoftBank Group Corp.-backed SB Energy Inc. filed for US listings last month, they have yet to start marketing their respective deals.
虽然纯数据中心开发商面临更多审视,但硬件和芯片名称仍然获得支持。Aequitas Research私人有限公司股票研究主管苏米特·辛格(Sumeet Singh)表示,像已经报告利润的印刷电路板制造商和存储芯片生产商这样与人工智能相关的公司将受到很小的影响。电路板制造商苏州东山精密制造股份有限公司周一开始评估需求,计划在香港上市筹集至多30亿美元。
While pure data center developers face more scrutiny, hardware and chip names still retain support. AI-linked companies like printed circuit-board makers and memory chip producers that are already reporting profits will see little impact, said Sumeet Singh, head of equity research at Aequitas Research Pvt. PCB maker Suzhou Dongshan Precision Manufacturing Co. on Monday started gauging demand to raise as much as $3 billion in a Hong Kong listing.
“Firmus没有产生收入,也缺乏业绩记录,”辛格说。“这类似于美国的SB Energy和Nscale等公司,它们现在正在筹集资金,而实际收入要过几年才会开始流入。”尽管如此,随着较高利率对推动一些收入增长预测的巨额资本需求造成压力,价格已成为资助新交易的投资者的痛点。华尔街对人工智能及其对其他业务潜在影响的焦虑日益加剧,这种恐惧促使投资者抛售了一系列相关股票。
“Firmus wasn’t generating revenue and lacked a track record,” Singh said. “It’s similar to the likes of SB Energy and Nscale in the US, where the money is being raised now while the actual revenue will only start flowing through after a few years.”Still, price has become the pressure point for investors funding new deals as higher interest rates weigh on the hefty capital requirements driving some revenue growth forecasts. Wall Street has grown increasingly anxious about AI and its potential to impact other businesses, a fear that spurred investors to dump a range of related stocks.
PrePublic Equity Partners投资研究主管汉斯·斯泰格(Hans Stege)表示,围绕当地数据中心建设不断增长的负面情绪和舆论增加了项目执行风险,或许调整了应用于这些高耸的头条未来已签约但未建成总额的贴现率。
Growing negative sentiment and public opinion around local data center construction adds project execution risk and perhaps adjusts the discount rate applied to these lofty headline future contracted but unbuilt totals, said Hans Stege, head of investment research at PrePublic Equity Partners.
“目前,我们已经看到私募市场挺身而出,帮助为建设提供资金,”斯泰格说。“考虑到吸引超大规模企业和实验室大笔交易的惊人吸引力,以及来自英伟达等公司的支持,这又有什么奇怪的呢?但你一定认为支持者在某个时候会期待这一群体公开上市。”
“For now, we’ve seen private markets stepping up to help bankroll buildouts,” said Stege. “And why not, given impressive traction attracting large deals from hyperscalers and labs, and support from the likes of Nvidia. But you’ve got to think the backers are expecting a public listing for this cohort at some point.”