Firmus IPO falters, raising the bar for Asian AI listings.

Firmus Grid Ltd.'s initial public offering (IPO) collapse is triggering ripple effects across Asia's red-hot equity issuance market, making investors more selective about AI infrastructure deals that rely on future earnings. The Nvidia-backed data center operator withdrew a listing on Friday that would have raised up to $5.5 billion, after investors balked at its valuation, which had surged to $30 billion. The Australian company is now in talks with existing investors about a private funding round. Firmus's chilly reception shows investors are becoming more cautious after more than a year of dealmaking frenzy, during which companies rushed to market to finance AI buildouts. Data show that IPOs, placements and block trades in Asia-Pacific have raised nearly $250 billion so far this year, on top of about $260 billion in 2025. "Investors are asking who pays, and at what cost of capital," said Max Herron-Vellacott, a portfolio manager at Sydney-based Private Portfolio Managers. He said risk-reward calculations are being scrutinized more closely when making investment decisions. "It raises the bar for every deal, and investors will be far less tolerant of optimistic assumptions than before." Recent performances of some new listings have also revealed creeping unease. And investors are starting to draw clear lines between companies: those with reliable track records that can monetize, and those earlier in their life cycle that are more cash-hungry. For data center operators like Firmus, the dividing line is between operating facilities that generate immediate revenue and multi-gigawatt new projects launched with billions of dollars in debt financing, said Winston Ma, executive director of NYU School of Law's Global Public Investment Funds Forum. Ma is also a partner at Dragon Global Family Office. "Firmus's pulled offering is a clear signal that public markets are repricing AI infrastructure," Ma said. "In Asia, the next wave of AI data center capacity is expected to be built in large volumes, which will raise the bar for any IPO. Sponsors will need stronger contracted offtake, clearer visibility on energized megawatts and more conservative leverage assumptions." For Firmus, only 46 megawatts of its 912-megawatt project pipeline has been built. Investors were asked to value the company on future earnings, all contingent on whether the data centers can be delivered successfully. Many of these companies are being compared by investors to so-called neoclouds such as CoreWeave and Nebius Group, among the most leveraged and volatile AI stocks in the U.S. Meanwhile, although Nscale Ltd. and SoftBank Group-backed SB Energy filed for U.S. listings last month, they have yet to launch marketing for their respective deals. Pure-play data center developers face stricter scrutiny, but hardware and chip stocks still enjoy support. Sumeet Singh, head of equity research at Aequitas Research, said AI-related companies that are already profitable, such as printed circuit board manufacturers and memory chip producers, are less affected. PCB maker Suzhou Dongshan Precision Manufacturing began assessing demand Monday for a Hong Kong listing that could raise up to $3 billion. "Firmus has no revenue generation and lacks a track record," Singh said. "This is similar to SB Energy and Nscale in the U.S., where money is raised now and actual revenue only starts to show up years later." Still, with rising interest rates weighing on massive capital needs, price has become a pressure point for investors funding new deals, and some revenue growth forecasts are being driven by those capital needs. Growing Wall Street anxiety about AI and its potential to disrupt other businesses has prompted investors to sell off a range of related stocks. Hans Stege, head of investment research at PrePublic Equity Partners, said growing negative sentiment and public opinion around local data center construction increases project execution risk and may adjust the discount rates the market applies to these high-profile contracted-but-not-yet-built totals. "Right now, we're seeing private markets step up more aggressively to fund the buildout," Stege said. "Given their astonishing ability to attract large orders from hyperscalers and labs, and backing from the likes of Nvidia, that's not surprising. But you have to think the backers expect this cohort of companies to go public at some point."