$BTC $NVDAB $MSFTB #Bitcoin #AI #挖礦
NVIDIA-backed Australian AI data center operator Firmus Technologies was originally set to list on the Australian Securities Exchange (ASX) on October 23, but has now announced it is withdrawing its IPO plans due to insufficient investor demand. Firmus had hoped to raise around $5 billion through the listing and secure a valuation of about $30.6 billion. However, growing caution in the market over the high valuations, massive capital expenditures, and financing risks associated with AI infrastructure ultimately forced the company to turn to private markets for funding.
Notably, Firmus is not a typical AI startup. As previously reported, Nico Caprez, NVIDIA’s Vice President of AI Infrastructure Growth and the son-in-law of CEO Jensen Huang, was a key figure in a major GPU financing deal between NVIDIA and Firmus. The deal involves plans to purchase more than 170,000 NVIDIA GPUs and CPUs, as well as a special arrangement under which NVIDIA provides credit support in exchange for a share of cloud service revenue.
Firmus’s failed IPO has prompted the market to reassess financing models in the AI infrastructure industry: is backing from NVIDIA and having major tech companies as customers enough to support a valuation of tens of billions of dollars?
Valuation Nearly Triples in Three Months, but Firmus Cancels IPO Due to Insufficient Demand
The Financial Times reported that Firmus had planned to list at A$11 per share, implying a market capitalization of about A$44 billion, or roughly US$31 billion, and aimed to raise about US$5 billion. If successful, it would have been one of Australia’s largest IPOs since telecommunications company Telstra went public in 1997.
However, Firmus’s valuation rose rapidly in a short period, raising investor concerns. Bloomberg reported that the company was valued at about $10.5 billion less than three months ago, yet its proposed IPO valuation had risen to about $30 billion.
In response to weak investor demand, Firmus tried lowering its offering price, but ultimately failed to secure sufficient support and formally canceled its listing plans on October 9. In a statement, Firmus said recent market volatility and broader market conditions had made it impossible to raise funds on terms that reflected the strength of its business and its long-term growth prospects. Co-founders Oliver Curtis and Tim Rosenfield also said in a letter to shareholders that they would seek private-market funding and assess other international public-market financing opportunities.
Reuters cited people familiar with the matter as saying that Firmus may first raise funds privately before considering a Nasdaq listing in the United States, but the company has not confirmed these plans.
From Bitcoin Miner to AI Data Center Gamble: Firmus Bets Big on AI Infrastructure
Founded in 2019, Firmus initially operated in Bitcoin mining. As demand for AI computing power grew, it shifted to data centers and AI cloud computing services, becoming what is known as a Neocloud provider.
Neocloud generally refers to a new type of cloud company specializing in GPU computing power and AI model training and inference services. Compared with major cloud platforms such as Amazon AWS, Microsoft Azure, and Google Cloud, these companies are smaller but can rapidly build dedicated infrastructure for AI workloads.
Firmus has expanded aggressively in recent years and established partnerships with technology companies such as Meta and OpenAI. It hopes to meet the continued global growth in demand for AI computing power by building large-scale data centers.
However, this business model requires extremely high upfront capital expenditure. Significant investment is needed to buy GPUs, build data centers, secure power supplies, and install cooling and networking infrastructure.
The Financial Times reported that although Firmus has secured contracts for more than 900 MW of power and data center capacity, only about 46 MW is currently operational. Investors are not simply buying an established, operational AI cloud business; they are betting on whether the company can complete its massive construction plans on schedule. Firmus’s revenue is also heavily reliant on projects that are not yet complete, raising concerns about its future profitability, construction timelines, and funding needs.
NVIDIA Does More Than Sell GPUs—it Also Provides Financing and Credit Support
Another reason Firmus’s failed listing has attracted attention is its close capital relationship with NVIDIA. As previously reported, Nico Caprez, NVIDIA’s Vice President of AI Infrastructure Growth, helped broker a major partnership between NVIDIA and Firmus in June this year.
(Who Is Nico Caprez? Jensen Huang’s Son-in-Law Leads Neocloud Business as NVIDIA Discusses Succession Internally)
The deal involves Firmus’s plan to use financing to purchase more than 170,000 current- and next-generation NVIDIA GPUs and CPUs. However, it is not easy for a rapidly expanding Neocloud company to borrow enough from banks to buy large quantities of AI chips.
Financial institutions need to assess not only the value of GPU assets themselves, but also future market demand, the rate at which chips depreciate, and whether the company can secure enough customers to cover the cost of computing leases.
NVIDIA provided what is known as credit support in the deal, helping Firmus secure financing. NVIDIA could also benefit from the partnership in two ways: direct sales of hardware such as GPUs and CPUs, and, through a special transaction structure, a share of the revenue Firmus generates from selling AI cloud computing capacity in the future.
This article, “NVIDIA’s Support Wasn’t Enough? Australian Neocloud Startup Firmus Pulls the Plug on Its $5 Billion IPO,” originally appeared on .
【Further Analysis】
【My Further Perspective】
Firmus’s failed listing is more than an isolated IPO setback; it also marks a more rigorous public-market examination of the cash-generating ability and high valuations of AI infrastructure (Neocloud) companies. The risks of a closed-loop model that relies on chip giants to provide “vendor financing” to support downstream purchases are becoming clear. An operational footprint of just 46 MW is difficult to reconcile with a forward-looking valuation of $30 billion. This suggests investors are growing weary of narratives built solely on “piling up GPU computing capacity” and unfulfilled long-term expectations, with investment priorities shifting from “computing-power euphoria” to scrutiny of “real returns on investment (ROI).”
【Market Impact】
1. **U.S. Stock Market (NVIDIA, Bearish to Neutral)**:
NVIDIA’s model of providing credit support to facilitate purchases could raise concerns about the quality of its financials. If Neocloud providers are unable to raise funds through public markets, the pace at which NVIDIA can recognize potential hardware sales could slow, while its implicit credit risk increases. This could also put pressure on the valuations of independent cloud service providers preparing to go public, such as CoreWeave.
2. **Cryptocurrency Market (Neutral)**:
For listed companies that have tried to follow the path of shifting from “Bitcoin mining to AI computing centers” (such as Core Scientific and IREN), valuation premiums could narrow. The short-term boost this theme provides to related tokens or mining stocks may also weaken.
3. **Taiwan Stock Market (AI Supply Chain, Bearish to Neutral)**:
Taiwanese AI server manufacturers (such as Quanta, Foxconn, and Wistron) and liquid-cooling suppliers may have orders from the four major CSPs to support them in the short term. However, capital spending on incremental orders from second-tier Neocloud companies could be delayed by financing difficulties, putting temporary valuation pressure on high-valuation supply-chain stocks.
【What to Watch Next】
If Firmus has to raise funds in the private market at a discount (a down round), this could establish a trend of Neocloud valuations deflating. Investors should also monitor disclosures in NVIDIA’s financial statements regarding the structure of its accounts receivable and guarantees for customer financing, as well as signs of cash flow stress among U.S.-listed mining companies shifting to AI infrastructure.
【Risk Reminder】
High leverage and chip depreciation risks are accumulating under the capital-intensive AI infrastructure model. Investors should be alert to the risk of supply-chain valuation declines if the capital expenditure cycle slows.