A reality check on AI boom: Eight risks to look out for

AI is an exceptional technology, but its amplifiers need to sober up in the face of visible headwinds of regulation, an unconvinced workforce, and costlier financing.

OpenAI and Anthropic postponing their estimated trillion-dollar IPOs is further evidence that capital markets are realising the distance between promises and prices, encapsulated in eight risks.

3 min readOct 16, 2026 06:20 AM IST First published on: Oct 16, 2026 at 06:20 AM IST

AI models are growing much faster than the financing, governance and social acceptance of the revolution. A sobering influence has emerged. On October 8-9, Nvidia-backed Firmus withdrew its nearly $5bn IPO after investors resisted a valuation of about $31bn, three times its August valuation. Almost simultaneously, OpenAI indicated that its September annualised revenue was nearer $50bn, not the nearly $70bn previously signalled. OpenAI and Anthropic postponing their estimated trillion-dollar IPOs is further evidence that capital markets are realising the distance between promises and prices, encapsulated in eight risks.

First, the economy. AI stocks are at near-record valuations. The Iran war, high US gas and diesel prices and the yawning US fiscal deficit have spooked the bond markets into trading US treasuries, with yields at multi-decadal highs. Higher interest rates are hitting homebuyers and are expected to adversely affect the growth of hyperscaler data centres, which AI cannot do without. Second, AI revenue growth trailing capex requirements. With OpenAI indicating lower-than-expected annualised revenue in CY26, Anthropic at an operating loss of 8bn in CY25 and Meta’s free cash flow falling to ~800mn from $8.5bn a year earlier, room to raise expectations appears limited.

Third, costlier capex financing. Of the expected ~$725 bn capex in CY26 by leading US hyperscalers on new data centres, the H1 assessment was ~40 per cent by end-June. Initially cash-funded, capex graduated to debt in formats such as private credit and bespoke financing templates, most of which are opaque. Goldman Sachs reported that most hyperscaler bonds issued in 2026 had fallen below issue price by end-August. Such unseen risk could impact banks, insurers and pension funds in the event of an AI selloff. Fourth, sobering valuations. In mid-2026, SpaceX raised $75bn, listing at $1.77tn, and OpenAI and Anthropic were reportedly targeting $60bn each, targeting trillion-dollar IPOs. But sobering assessments have forced both Sam Altman and Dario Amodei to delay their listings. Fifth, regulatory risk. OpenAI and Anthropic were jolted when three of their frontier models were subject to sudden regulatory slowdowns, highlighting the risk of government control by diktat. Donald Trump also made AI CEOs accept responsibility for safety and agree to establish internal controls and external audits. Sixth, the risk of rogue AI is resulting in multimillion-dollar claims covering cyber attacks, copyright and privacy against directors, officers and entities, with hyperscaler potential in the same league as landmark litigation against tobacco, glyphosate, pharma and environmental liabilities.

Seventh, AI in war: Normalised, yet an amoral revenue generator. Use of autonomous AI is documented in Ukraine, Gaza and Iran, without due consideration for human scrutiny, civilian harm, proportionality and accountability. Some multilateral sense is bound to prevail sooner rather than later. Finally, rising global opposition to data centres. Communities cite reduced drinking water, expected electricity outages, noise pollution, land loss, and limited permanent employment. Protests have forced some rollbacks. But there are national-security concerns because China’s largely unrestricted buildout of another 50 GW beyond its current 24 GW capacity (US capacity is 56 GW) has gathered pace. AI is an exceptional technology, but its amplifiers need to sober up in the face of visible headwinds of regulation, an unconvinced workforce, and costlier financing.