OpenAI won't make money by 2030 and needs another $207B, HSBC estimates (fortune.com)

HSBC’s updated forecast argues that OpenAI will still be unprofitable by 2030 and faces a roughly $207 billion funding shortfall to deliver on its growth plans. The bank models steep infrastructure needs — including commitments to some $1.4 trillion of compute by 2033 and 36 gigawatts of AI power by 2030 — driving estimated cloud and AI infrastructure costs of $792 billion from late‑2025 to 2030 and a $620 billion data‑center rental bill. Even with revenues ramping to about $213 billion by 2030 and consumer penetration rising to an estimated 44% of adults, HSBC says those receipts won’t cover the cash burn unless OpenAI dramatically ups paid conversions, captures more ad spend, or squeezes extraordinary compute efficiencies. Recent multiyear cloud deals (roughly $250B with Microsoft and $38B with Amazon) expanded capacity without new equity, intensifying the funding gap. This matters because it reframes AI’s “megacycle” as intensely capital‑hungry: success depends not just on models and market share but on financing massive, energy‑intensive infrastructure. The forecast spotlights systemic risks for investors, hyperscalers and chipmakers (Nvidia, AMD, Oracle), raises questions about debt markets’ appetite for AI capex, and underscores monetization and regulatory uncertainties. In short, HSBC sees OpenAI’s technical ambitions as plausible but financially perilous — a reminder that transformational AI scale may require sustained, large‑scale funding or a materially different revenue model.