OpenAI won’t make money by 2030 and still needs to come up with another $207 billion to power its growth plans, HSBC estimates

HSBC's updated forecast warns that OpenAI, still private and unprofitable, is projected to remain loss-making through 2030 despite rapid revenue growth to over $213 billion by 2030; the bank calculates a $207 billion funding shortfall through 2030 and models $792 billion of cloud and AI infrastructure costs from late-2025 to 2030. HSBC assumes OpenAI will target 36 GW of AI compute by decade-end (part of a $1.4 trillion compute plan through 2033) and cites multiyear cloud commitments including a $250 billion Microsoft deal and $38 billion with Amazon that carry no new capital injections. The scale of data-center and electricity needs (a $620 billion rental bill alone) creates material financing and operational risk for OpenAI and its ecosystem — pressuring cloud providers, chipmakers and credit markets and implying further equity/debt raises or aggressive monetization to bridge the gap.

Analysis

Market structure: The HSBC workforces the math — 36GW by 2030 and $1.4T compute by 2033 — which centralizes demand for GPUs, cloud capacity and power. Clear winners are Nvidia (NVDA) for silicon scarcity pricing, and hyperscalers Microsoft (MSFT) and Amazon (AMZN) for preferential host deals; losers include credit-sensitive capex players (Oracle/ORCL) and ad-dependent platforms (META) facing margin squeeze. Macro cross‑assets: expect higher corporate issuance (higher IG supply), wider tech CDS, upward pressure on power/gas and copper, and local USD funding demand for large capex programs.

Risk assessment: Tail risks include a forced OpenAI capital raise or distressed asset sales (high‑impact, low‑probability within 12–36 months), regulatory intervention that limits monetization, and GPU supply shocks which could spike compute costs 20–50% in months. Near term (days–weeks) watch CDS and capital‑markets windows; medium term (3–12 months) funding rounds and hyperscaler guidance; long term (to 2030) hinges on sustained ARPU conversion (paid base rising from 10% to 20% would add ~$194bn revenue per HSBC). Hidden dependency: OpenAI’s viability is levered to MSFT/AMZN commercial incentives and grid capacity in key regions.

Trade implications: Tactical approach: overweight NVDA (2–3% of portfolio, 6–12 month horizon via shares or 9–12 month call spreads) to play pricing power; overweight MSFT/AMZN (1–2% each) to capture cloud economics; short ORCL (1–2% short equity or buy 12‑month put spread) and consider buying ORCL CDS if spreads widen >30bp. Pair: long NVDA / short ORCL (capture hardware scarcity vs credit risk). Use options to express skew: buy NVDA 6–9 month call spreads and buy ORCL 12 month puts to limit capital and exploit asymmetric risk.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA (equity or 9–12 month call spreads) with a 6–12 month horizon to capture continued GPU pricing power; add on pullbacks >15% and trim on a 30% rally.
• Allocate 1–2% long to MSFT and 1% to AMZN (shares) to play cloud hosting capture; hold 12–24 months, increase position if either reports >5% sequential cloud-margin improvement tied to AI deals.
• Initiate a 1–2% short in ORCL (equity short or buy 12‑month put spread); if access to credit, buy ORCL CDS if spreads widen >30–50bp. Target profit if CDS normalizes or ORCL equity drops 20%.
• Pair trade: long NVDA / short ORCL sized dollar‑neutral (e.g., $1m long NVDA vs $1m short ORCL) to isolate hardware scarcity vs credit/capex risk; reprice after quarterly results or any OpenAI capital announcement within 90 days.
• Use options for asymmetric risk: buy NVDA 6–9 month call spreads (limit cost) and buy ORCL 12 month put spreads; size combined options exposure to <=1–2% portfolio risk to limit vega exposure.