4004 News
Episode

Firmus IPO Collapse Exposes AI Data Center Bubble

An analysis of the Firmus IPO failure and its implications for the AI infrastructure sector. The episode highlights the disconnect between projected valuations and actual revenue, the role of media scrutiny in preventing investor harm, and the systemic risks facing neocloud companies backed by major financial institutions.

The Firmus IPO: A Case Study in AI Infrastructure Risk

The attempted IPO of Firmus, an NVIDIA-backed AI data center company in Australia, has become a pivotal moment for the AI infrastructure sector. Initially targeting a $50 billion valuation, Firmus faced a rapid collapse in pricing, with shares dropping from an initial $11 offer to potentially $6 or $7. This drastic devaluation underscores the growing skepticism among investors regarding the financial viability of neocloud companies that promise massive capacity but lack proven execution.

Revenue Concentration and Execution Gaps

A critical flaw in Firmus’s model is its extreme revenue concentration, with 90% of its backlog tied to just two customers: Meta and an unnamed global tech giant. Furthermore, the company’s plans to build 10 gigawatts of capacity in five years are underpinned by deals for data centers that do not yet exist. This disconnect between projected revenue and physical infrastructure creates a fragile foundation for valuation. As noted by hedge fund CEO David Allen, valuing such companies based on EBITDA multiples is fundamentally flawed, comparing it to valuing an airline based on the assumption that planes are free and last forever.

The Role of Media Accountability

The Australian financial press played a crucial role in exposing these risks, contrasting sharply with the more favorable coverage often given to similar US companies like CoreWeave. Australian journalists rigorously scrutinized Firmus’s metrics, debt loads, and delivery timelines, effectively warning investors of the potential bubble. This media accountability served as a protective mechanism for retail and institutional investors, preventing a potentially larger market correction. The episode argues that the US financial media’s failure to apply similar scrutiny to American neoclouds has allowed hype to outpace reality, increasing systemic risk.

Institutional Exposure and Systemic Implications

The involvement of major financial institutions, including Blackstone’s credit and insurance funds, in backing Firmus highlights the broader exposure of pension and insurance portfolios to speculative AI bets. This raises serious concerns about the stability of these long-term funds if the AI infrastructure bubble bursts. The Firmus case serves as a warning that without rigorous due diligence and critical media oversight, investors remain vulnerable to the risks of overvalued, execution-dependent tech ventures. The episode concludes that a fundamental shift in how financial media reports on AI infrastructure is necessary to protect markets and investors from future collapses.

Key insights

Firmus’s IPO pricing collapse from $11 to potentially $6 per share reveals a significant disconnect between AI infrastructure valuations and actual revenue. This indicates that investors are increasingly wary of companies with high projected growth but low current earnings.
Impact: This trend may lead to a broader re-evaluation of AI data center stocks, causing a correction in the sector and reducing capital inflows into unproven neocloud models.
Market Valuation →
The concentration of 90% of Firmus’s revenue backlog in just two customers creates a high-risk business model. This dependency on a few large clients makes the company vulnerable to contract changes or cancellations.
Impact: Companies with similar revenue concentration face heightened operational risk, potentially leading to liquidity crises if key clients reduce their commitments.
Business Strategy →
Australian financial media’s rigorous scrutiny of Firmus served as a critical check on hype, contrasting with the more favorable coverage of US neoclouds. This difference in media approach highlights the importance of independent journalism in protecting investors.
Impact: The lack of similar scrutiny in the US may allow overvalued AI companies to continue raising capital, increasing the risk of a future market crash.
Media Accountability →
The involvement of Blackstone’s credit and insurance funds in backing Firmus exposes long-term institutional capital to speculative AI bets. This raises concerns about the stability of pension and insurance portfolios linked to volatile tech sectors.
Impact: If AI infrastructure companies fail, pension and insurance funds may suffer significant losses, potentially impacting retirement security and insurance payouts.
Institutional Risk →
Valuing AI data centers based on future EBITDA without proven execution is fundamentally flawed. The Firmus case demonstrates that projected revenue is insufficient to support high valuations without tangible infrastructure.
Impact: Investors may need to adopt more conservative valuation models for AI infrastructure companies, focusing on actual capacity and revenue rather than projections.
Investment Analysis →

Action items

Conduct a thorough review of revenue concentration in AI infrastructure portfolios. Identify companies with high dependency on a few customers and assess the risk of contract cancellations.
Impact: This will help investors mitigate the risk of liquidity crises and portfolio losses due to customer concentration.
Implement stricter due diligence processes for AI data center investments. Focus on verifying physical infrastructure and delivery timelines rather than relying on projected revenue.
Impact: This will reduce the risk of investing in companies with unproven execution capabilities and overvalued projections.
Encourage financial media to adopt a more critical approach to AI infrastructure reporting. Highlight the importance of independent journalism in protecting investors from hype.
Impact: This will improve market transparency and help prevent future bubbles in the AI sector.
Assess the exposure of pension and insurance funds to speculative AI bets. Review the involvement of institutional investors in high-risk AI infrastructure projects.
Impact: This will help institutional investors manage their risk and protect long-term capital from volatile tech sectors.
Develop more conservative valuation models for AI infrastructure companies. Focus on actual capacity and revenue rather than projected EBITDA.
Impact: This will provide a more accurate assessment of company value and reduce the risk of overvaluation.

Quotes

“if the IPO was so well covered at $11 a share, why was it cut? Well, I think the answer is obvious.”
“valuing Fermus based on its EBITDA multiple, as in working out how valuable it is based on the cash that it might bring in, is kind of like valuing an airline based on planes being free and lasting forever.”
“The Australian coverage of firmas should be an embarrassment to the American finance and tech press, because it shows what actual journalism looks like.”