Oracle stock slides as AI spending outpaces returns and backlog credit evaporates
Dec 11, 2025
Key Points
- Oracle's stock falls as investors realize the company's multi-billion-dollar AI infrastructure deals face a months-long revenue lag between securing power, procuring chips, and generating returns.
- The market has withdrawn credit from Oracle's $523 billion backlog, now assigning the partnership negative $70 million in value as token prices commoditize and unit economics hinge on sustained customer spending.
- Enterprise AI adoption has flatlined at 45% of businesses, with OpenAI adoption declining month-over-month as the initial capability spike fades absent new breakthroughs.
Summary
Oracle's stock has tumbled as investors lose faith in the company's ability to turn massive AI infrastructure spending into near-term profits. The company's recent earnings fell slightly short of expectations while raising its spending forecast, signaling a lengthening runway before returns materialize.
Building AI infrastructure requires months of sequential steps: securing power, procuring chips, waiting for shipments, racking equipment, testing systems, and only then generating revenue. Investors did not anticipate this lag when Oracle announced its multi-billion-dollar AI deals.
Oracle's remaining performance obligations stand at $523 billion in contracted revenue not yet recognized. The stock trades at only $568 billion market cap. The market initially credited Oracle for the backlog when it was announced but has since withdrawn that credit steadily. According to one Financial Times analysis, the market is now assigning the partnership negative $70 million in value.
Did Larry Ellison misread the economics? Microsoft's Satya Nadella, who owns OpenAI's IP and runs one of the most scaled AI platforms globally, notably did not chase similar deals despite having more information than anyone in the room. The market appears to be catching up to Nadella's restraint. Token prices are commoditizing. If competitors serve compute at 50% lower cost, unit economics collapse. Jensen Huang's model assumes a rack costs $4 million and generates $20 million in revenue, netting $16 million. That math only holds if customers keep paying and capital keeps flowing despite their own losses.
Enterprise AI adoption has flatlined. Ramp's AI index shows adoption holding at 45% of businesses in November, with OpenAI adoption down 1% month-over-month while Anthropic gained 0.8% and Google 0.7%. OpenAI spiked early in 2025 but has been flat since. The slowdown follows a rapid run-up in adoption that coincided with a step-change in model capabilities. Without new breakthroughs or implementation gains, adoption won't accelerate again.
Ramp analyst Ara Carazian cautions against overreading the results. The market changes rapidly, and it's not unreasonable that OpenAI's growth rate would normalize as competitors find their footing. Many businesses, especially small operators and those with simple workflows like a laundromat or small restaurant, have no practical need for $20-per-month AI subscriptions. Others are getting AI value embedded in SaaS tools they already use rather than paying directly for chat interfaces.