The $30 billion initial public offering of Sydney-based Firmus Grid Ltd. last week marked the first major reality check for AI data center operators. According to reports from Bloomberg, the company's failed offering occurred despite having a 912 megawatt pipeline, of which only 46MW had been built. This astronomical valuation represented a significant leap from just two months earlier, highlighting investor skepticism about the neocloud business model's scalability. Recent developments show Asian AI stocks tumbled as rising bond yields, higher financing costs, and concerns over excessive infrastructure spending weakened investor confidence, with South Korea's Kospi and Japan's chipmakers suffering sharp losses.
Private equity giant Blackstone Inc. holds a 6.7% equity stake in Firmus Grid, making it one of the company's largest investors. As reported by Bloomberg, the investment firm participated in a $2 billion funding round in August with a valuation exceeding $10.5 billion, following its leadership of a $10 billion debt financing facility. The firm reportedly negotiated extensive clauses to protect its equity interests in case of IPO failure or poor market performance.
A new report reveals that power availability emerges as a major challenge for the expansion of AI data centres, with investors increasingly backing technologies that combine energy supply with computing infrastructure. According to The Economic Times, US startups have significantly increased their funding, raising $33.5 billion for technologies addressing AI data centre challenges, which is nearly double compared to the previous year's total raised in 2025. This funding surge reflects the growing recognition that power constraints represent a fundamental limitation to AI infrastructure expansion.
The neocloud business model faces fundamental questions about profitability and cash flow generation. As reported by Bloomberg, CoreWeave Inc., a pioneer of the neocloud model, is expected to generate $7.6 billion EBITDA this year but continues to bleed free cash flow. The company plans to spend as much as $39 billion in capital expenditure, significantly more than its earnings capacity. Investors are increasingly demanding sustainable revenues and profits, questioning whether massive AI investments will deliver expected returns.
Looking ahead, Blackstone faces challenges in monetizing its AI data center investments, with AirTrunk planning to go public in 2027 and another late-stage unicorn Lambda also in its portfolio. According to Bloomberg, investors are increasingly questioning the mismatch between technology and financing, as processing chips depreciate on a 3-6 year schedule while developers remain liable for debt repayments. The $1.3 trillion asset manager's AI data center deals appear increasingly vulnerable to changing market sentiment and emerging power constraints that threaten the scalability of AI infrastructure investments.
