Hyperscalers & Cloud
Sells enterprise databases and business software and runs a fast-growing cloud (OCI) increasingly rented for AI training.
{'verdict': '4 signals sit in the elevated band: operating leverage, the AI-monetization gap, the valuation premium vs fundamentals and organic-demand sustainability. This trips the convergence flag. Ranks 3rd of 68 on composite fragility (F\xa069.25), below TSLA and xAI.', 'as_of': '2026-07-11', 'source': 'engine-restatement (T1)', 'snapshot': {'composite_f': 69.25, 'n_elevated': 4, 'convergence': 'active', 'rank': 3, 'elevated': ['operating leverage', 'the AI-monetization gap', 'the valuation premium vs fundamentals', 'organic-demand sustainability']}}
Why is Oracle extending server useful lives to 6 years when Amazon just shortened to 5, citing AI obsolescence?
The FY2025 10-K confirms the 5→6 year extension but gives no rationale beyond completing "an assessment." The contrast with Amazon's 6→5 shortening (which explicitly cited AI/ML obsolescence) is unexplained in the filing. Oracle's direction may reflect that its newer purpose-built AI data center equipment is more standardized and durable — or it may reflect earnings management at a time when the company was issuing $43B in debt. The 10-K does not quantify the dollar impact of the change. NOT SOURCED: earnings impact.
Who are the customers behind the $317.5B Q1 FY2026 RPO spike?
The RPO footnote in the 10-Q/10-K does not attribute RPO to specific customers. But Oracle names its AI customers in its own furnished 8-K earnings exhibit: the Q3 FY2025 EX-99.1 (2025-03-10) states Oracle "signed cloud agreements with several world leading technology companies including: OpenAI, xAI, Meta, NVIDIA and AMD" and references "our first Stargate contract." So the customer set (incl. OpenAI/Stargate) is filed — you can pull the 8-K — even though the per-customer RPO split is not broken out. The Stargate JV (SoftBank + OpenAI + Oracle, announced Jan 2026) is the most likely driver of the spike; the per-customer attribution is inference, but the named relationships are primary-filed.
Can Oracle service $134.6B in debt + $248B in off-balance-sheet lease commitments if OCI growth decelerates?
At 9-month FY2026 revenues of $48.2B (annualizing to ~$64B), Oracle's total on+off-balance-sheet obligations (~$383B) represent ~6x annual revenue. The $43B in new debt carries rates of 4.45-6.25% (annual interest ~$2B+ on the new tranche alone). If OCI growth slows from 70%+ to 20%, Oracle's revenue path becomes insufficient to cover lease + debt service without further borrowing. The coverage math is NOT SOURCED in forward projections.
What happens to the RPO if the AI companies in the Stargate ecosystem face funding pressure?
The RPO represents contracted obligations, but "variable consideration" for usage-based portions is excluded from the disclosed figure. If customers consume less than contracted, the revenue recognized from the RPO will undershoot. Oracle's recourse for non-payment or contract renegotiation is not disclosed. NOT SOURCED: penalty terms, non-consumption clauses.
Oracle's Ampere stake exit ($4.3B, Nov 2025) — what does this say about Oracle's AI-chip strategy?
Oracle received $4.3B when SoftBank acquired Ampere (Arm-based chip company). Oracle had been a major Ampere customer and equity holder. The exit is disclosed in the Q3 FY2026 10-Q as a $2.7B gain. It suggests Oracle no longer needs to hedge into chip manufacturing stakes — it has locked in multi-decade GPU supply via OCI compute contracts. The strategic retreat from chip equity is a clean signal.