Oracle's Contract Churn Cuts Both Ways as AI Buildout Meets a Balance-Sheet Test

Oracle's pitch to investors rests on a simple proposition: demand for its AI infrastructure is strong enough to keep even aging hardware earning money. The company points to robust uptake and better pricing on contract renewals, an argument that carries weight only if those renewals translate into dependable cash flows. With free cash flow in negative territory and debt piled high, that link is where the real debate begins.

The counterweight arrived this week. Bloomberg reported that Microsoft let its agreement to rent AI computing capacity from Oracle lapse, having scaled back usage over recent months because the capacity relied on older Nvidia chips that no longer met its requirements. Neither Oracle nor Microsoft publicly confirmed the non-renewal. Reuters picked up the report and flagged Oracle share losses in afternoon trading, though the stock finished Friday up 1.5%.

The episode sharpens the question of what happens to capacity once a major tenant walks away. Oracle's own commentary suggests older chips can still find takers, provided a customer's needs match the performance on offer. Economically, lining up a paying replacement matters more than blanket assurances about demand.

A Backlog That Needs a Cash-Flow Translation

At the end of August, Oracle put its order backlog at $664 billion — a formidable contractual base. Yet the figure only gains conviction if execution strengthens free cash flow. That is the central tension in the investment case: Oracle must service its obligations while harvesting the financial fruits of its AI expansion.

The starting numbers counsel caution. Debt stood at roughly $125.3 billion as of August 31, and negative free cash flow for fiscal 2026 came in at $23.7 billion. That combination argues against extending credit on the strength of the backlog alone while cash continues to drain. Balance-sheet resilience deserves more weight than the sheer scale of demand until booked business produces a sturdier financial position. Moody's rates Oracle Baa2 with a negative outlook, underscoring how much hinges on the company's finances while the business absorbs heavy outlays.

Political Hurdles Narrow the Field

Political viability adds another layer of pressure. According to the Financial Times, Oracle abandoned a planned Tencent contract after US officials raised concerns. The deal was valued at about $7 billion over five years and would have provided access to as many as 100,000 AI chips in Southeast Asia. The upshot: economic appeal counts for little if a contract cannot clear regulatory muster, making politically durable customer relationships as important to the valuation as contract length and volume.

Oracle is not standing still on the product front. On October 27 it announced, according to media reports, an expansion of its Nvidia AI server plans with Vera-Rubin systems to a scale of multiple gigawatts. It also intends to offer chips from Google, SambaNova and Cerebras. No contract value was disclosed. Two days earlier, Oracle said it would no longer require confidentiality or non-disclosure agreements with federal, state and local authorities for its AI data center projects.

Growth Targets Set the Bar

At its investor day on October 28, Oracle laid out FY30 targets of at least $225 billion in revenue, an adjusted non-GAAP operating margin of at least 40%, and earnings per share on the same basis of at least $21. Management also set cash-flow goals: positive free cash flow in the second half of FY28, and at least $30 billion in FY29. These are company objectives, not results already achieved.

Oracle also found itself caught up in a broader selloff of AI data center stocks on Thursday. AP attributed the sector's losses to worries about financing costs, lofty AI valuations and possible restrictions on data center construction after the US midterm elections, with potential construction halts in focus.

Taken together, the demand story is persuasive; the financial strain currently weighs heavier. A more constructive read would follow once renewals, backlog conversion and free cash flow show progress in tandem. The yardstick is clear — Oracle has to wring more financial durability out of its AI business.