Oracle
Recurring license-support + cloud subscription (SaaS) plus consumption/contracted GPU compute (OCI IaaS); pivoting from high-margin software toward capital-intensive AI data-center capacity.
The thesis on this name
State of AI Compute
OCI's GPU-cloud land-grab and the RPO backlog make Oracle a fast-rising hyperscaler-class buyer tied to the OpenAI / Stargate buildout.
State of the AI Cloud
Be selective-to-short where the re-rating IS the risk: Oracle's entire AI-capacity re-rate rests on ONE unprofitable counterparty (~54%/~$300B of RPO is OpenAI/Stargate). The SpaceX (SPCX) leg has largely paid — the stock fell ~28% from ~$185 on 2 Jul 2026 to $133.11 on 7 Aug 2026 and the multiple re-based from ~87x to ~56x annualised revenue — so what remains there is a cash-burn avoid (capex at 235% of revenue, roughly -$25.0B of first-half free cash flow) on leases that are still partly cancellable (Google/Reflection 90-day-terminable after 2026-12-31). Express both as defined-risk shorts/underweights against the long toll-collector book, marked to a stated reference price so they mark-to-market.
State of the AI Cloud
The AI-capacity dark horse whose entire re-rating rests on one unprofitable counterparty — ~54% of a ~$523-553B RPO is OpenAI/Stargate alone.
State of the AI Cloud
The re-rating IS the ~54% OpenAI concentration; expressed as a defined-risk underweight vs the diversified-book anchors, marked to the pinned ref.
State of AI for Healthcare
The only listed way to own an EHR of scale, and it is the one losing — avoid, and specifically not short. Per KLAS's US acute-care EHR share report of 14 May 2026, as relayed by the trade press, Oracle Health holds 21.9% of US acute-care hospitals against Epic's 43.7%, and posted a third consecutive year as the largest net share loser (−56 hospitals, −14,676 beds in 2025); Millennium scored lowest of the acute EHRs across large, midsize and small organisations in 2026 Best in KLAS. None of that is expressible in the equity. Oracle Health is not a reported segment; FY2026 revenue was $67.4B (+17%) with $17.0B of net income; and at $129.87 / $374.09B (31 Jul 2026), on 16.1x forward earnings (a consensus estimate) against a 52-week range of $114.50–$345.72, the stock is an AI-capex story with no health content in it. Shorting ORCL to express an EHR view is a bet on OCI, on AI capex and on financing risk — three exposures this board has no view on. This entry exists to rule the name out, which is the honest answer to 'is there a listed EHR': the #1 has no shares at all and the #2 is a line item inside an AI-infrastructure stock, so there is no way to be long or short the US electronic health record.
State of AI for Healthcare
The only listed way to own an EHR of scale — 21.9% of US acute-care hospitals per KLAS against Epic's 43.7% — and it is losing at it: −56 hospitals and −14,676 beds in 2025, a third consecutive year as the largest net share loser, inside a $67.4B revenue base that trades entirely on cloud infrastructure.
State of Enterprise AI SaaS
Reference-only arm: OCI +84% ($4.9B) is a capacity-utilization AI-infrastructure story riding the consumption wave, backed by a sticky database/apps installed base. Lower conviction as an arm given the ~$50B AI-capex intensity and 30,000-job restructuring signal margin/execution risk alongside the growth.
State of Enterprise AI SaaS
No longer a SaaS story but a leveraged AI-infra capex bet — $523B RPO concentrated in OpenAI, FCF at -$24.7B, debt ballooning to $125B; the seat debate is irrelevant, the financing risk isn't.
State of Enterprise AI SaaS
OCI +84% consumption story; capped by $50B capex intensity. Small, low-conviction arm.
Earnings, margins, COGS & capex
FY26 was a record top line ($67.4B, +17%) but the story is the pivot to AI infrastructure: OCI/IaaS revenue $18.1B (+77%) and RPO exploding to $638B (+363% YoY). The catch is capital intensity — capex $55.7B (+162%), FCF -$23.7B, and $134.6B total debt funding the buildout. Earnings beat estimates yet the stock fell ~10% on the print and is down ~57% from its 52-week high on AI-capex/funding fears.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~95¢ is cost of goods and ~0¢ operating expense, leaving ~5¢ of operating profit (~25¢ net).
Revenue trend
Margins
roughly steady YoY despite infra build
steady
down ~5pts as low-margin OCI infra scales (was 72.5% in early FY24)
far below AWS/Azure 30-40%; mgmt guides +30-40% improvement over contract life as utilization ramps
up +36% YoY
COGS structure
COGS rising faster than revenue as OCI hardware depreciation and data-center operating costs (power, GPU racks) scale — the structural driver of the ~5pt gross-margin step-down. AI hardware depreciation is the single largest swing factor.
Capex
$55.7B in FY26 (+162% YoY, overshooting ~$50B guide); FY27 net capex guided to ~$70B. Funded by $43B debt + $5B equity raised in FY26 (incl. a record $25B bond) plus ~$40B more financing planned for FY27 (incl. a ~$20B share sale).
Latest earnings
Beat — revenue $19.2B and non-GAAP EPS $2.11 (+24%) topped estimates; stock still fell ~10% on capex/financing concerns
FY2027 total revenue raised to ~$90B; FY27 net capex ~$70B; ~$40B additional financing planned. Management reiterated multi-year OCI/RPO ramp toward $100B+ cloud revenue.
- Q4 revenue
- $19.2B (+21%)
- Q4 non-GAAP EPS
- $2.11 (+24%)
- Q4 GAAP EPS
- $1.45 (+21%)
- RPO
- $638B (+363% YoY, +$85B QoQ)
- Q4 AI infra contracts signed
- ~$67B (mostly prepaid or bring-your-own-hardware)
- FY26 FCF
- -$23.7B
- Total debt
- ~$134.6B
Growth drivers
- AI training/inference GPU compute demand (OCI superclusters, Nvidia GPUs)
- ~$300B OpenAI/Stargate 5-yr compute contract starting CY2027
- RPO conversion: $638B backlog with $75B prepaid/customer-supplied-hardware portion de-risking funding
- Multicloud database (Oracle DB on AWS/Azure/Google) + autonomous DB attach
- Fusion/NetSuite SaaS cross-sell into the installed enterprise base
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-06-22. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’26
Bull & bear
Oracle is the most contracted AI-infrastructure play in the market — a $638B backlog (≈9x FY26 revenue) and a raised FY27 guide to ~$90B mean the growth is largely pre-sold, not speculative. If utilization ramps and OCI margins normalize toward management's +30-40% target, today's ~$149 price (down 57% from the high, ~22x forward earnings) is a re-rating opportunity.
- RPO $638B gives multi-year revenue visibility few companies can match; $75B is prepaid or customer-supplied hardware, materially de-risking the funding
- OCI +93% in Q4 — Oracle is genuinely winning AI infrastructure share as the 'fourth hyperscaler'
- FY27 revenue guide raised to ~$90B (+~34%); the inflection from backlog to recognized revenue is just beginning
- Margin recovery optionality: OCI gross margin is depressed by upfront depreciation, not by structurally bad economics — utilization and contract maturation should lift it
- Valuation reset: ~57% off the high and forward P/E ~22x (between cheaper MSFT ~19x and pricier AMZN ~28x) prices in a lot of the funding fear
- Sticky, profitable legacy database/ERP base funds operations and dividend ($2.00/yr) while the cloud build matures
Oracle is funding hyperscaler-scale capex without a hyperscaler balance sheet: $55.7B capex, -$23.7B FCF, and $134.6B debt (+45% in 9 months), with ~$40B more financing coming. The backlog is real but concentrated in OpenAI (~$300B) whose ability to pay is unproven, OCI gross margins are ~14%, and the stock has already de-rated ~57% — the market is pricing the risk that the AI demand doesn't earn its cost of capital.
- FCF -$23.7B and total debt $134.6B — every dollar of growth currently consumes capital; the model depends on continuous debt/equity issuance into a possibly tightening credit market
- ~$300B single-customer (OpenAI) concentration: cancellation, deferral, or counterparty credit failure would gut the backlog narrative
- OCI gross margin ~14% vs AWS/Azure 30-40% — if utilization or pricing disappoints, the unit economics never work and GPUs impair
- Capex ~83% of revenue and rising to ~$70B (FY27): Oracle is taking hyperscaler-level balance-sheet risk from a weaker starting position than MSFT/AMZN/GOOGL
- AI-bubble macro: Burry and others flag dot-com-style stretch; an AI capex pause would strand assets and break the demand assumption
- Stock fell ~10% on a beat — the market is now skeptical of the financing story even when results clear estimates
What it is worth
Forward P/E comps + reverse-DCF sanity on the RPO-conversion thesis
~$90-120
AI capex demand softens or OpenAI defers/renegotiates, OCI utilization disappoints, GPUs impair, and credit-market access tightens on a $134.6B (rising) debt load; multiple de-rates further toward the legacy-software base.
~$150-185
backlog converts but margins stay thin near-term and leverage keeps a valuation discount vs cash-rich peers; stock compounds with revenue, not multiple expansion.
~$250-300
RPO converts on schedule, OCI margins normalize and FCF inflects positive by ~FY28; re-rates back toward prior highs as the 'fourth hyperscaler' thesis is validated (sell-side avg target ~$253).
At ~$149 (mkt cap ~$430B), ORCL trades ~22x forward EPS — above MSFT (~19x) and below AMZN (~28x). On a reverse-DCF, the price implies the market is NOT fully crediting the $638B backlog: it bakes in mid-teens revenue CAGR with structurally thin OCI margins and continued negative FCF, i.e. heavy skepticism that the capex earns its cost of capital. The bull case needs OCI gross margins to ramp toward 25-30%+ and FCF to inflect positive ~FY28-29 as utilization matures; the bear case is margin/utilization disappointment + OpenAI counterparty risk + a tighter credit market for the $40B+ of planned financing.
SWOT
Strengths
- $638B RPO backlog (+363% YoY) — multi-year contracted visibility unmatched among peers relative to revenue base
- OCI/IaaS growing +77% FY26 / +93% Q4, fastest among the major clouds
- Entrenched mission-critical database + ERP installed base (Fusion, NetSuite, autonomous DB) with high switching costs and recurring support
- Multicloud strategy (Oracle DB running inside AWS/Azure/Google) turns rivals into distribution
- Profitable and cash-generative at the operating level — $32.0B operating cash flow, +54% YoY
Weaknesses
- OCI gross margin est. ~14%, far below AWS/Azure 30-40% — AI hardware depreciation crushes near-term cloud profitability
- FCF deeply negative (-$23.7B) and total debt $134.6B (+45% in 9 months) — funding the build from a weaker balance sheet than Microsoft/Amazon/Alphabet
- Extreme customer concentration: ~$300B of RPO is one customer (OpenAI/Stargate)
- Capex (~$55.7B, 83% of revenue) requires sustained, on-time, high-utilization demand to ever earn its cost of capital
- SaaS growth only +10% — the legacy software engine is mature/slow vs the AI narrative
Opportunities
- Convert $638B RPO into revenue at improving utilization, lifting OCI margins +30-40% over contract life as mgmt guides
- Sovereign/enterprise AI clouds across multiple countries (GPU-of-choice for parties locked out of Big Three capacity)
- Inference (not just training) demand broadening the customer base (Nvidia, Meta and others added to RPO)
- Database-on-GPU and AI-data-platform attach to the existing enterprise estate
- Stargate/SoftBank partnership scaling national-scale AI infrastructure
Threats
- AI-capex bubble risk — if AI demand disappoints, contracted backlog could be renegotiated/cancelled and stranded GPU assets impair
- OpenAI's own ability to pay ~$300B is unproven — counterparty/credit risk to the headline contract
- Hyperscaler competition (AWS, Azure, Google) with deeper pockets, custom silicon (Trainium/TPU) and lower cost of capital
- Rising rates / credit-spread widening on a rapidly levering balance sheet — debt-funded model is rate-sensitive
- Nvidia GPU supply, pricing and roadmap dependency; power/grid constraints on data-center delivery
Moats, dependencies & bottlenecks
Moats
Oracle DB + Fusion/NetSuite ERP are deeply embedded; migrations are costly and risky, sustaining license-support recurring revenue.
$638B of signed commitments is a near-term moat but concentration + counterparty risk make it less durable than installed-base lock-in.
Turns rivals into channels and keeps Oracle data workloads sticky regardless of where compute runs.
Reputation as the cloud that can stand up large GPU superclusters fast; erodes if peers match delivery and price.
Dependencies
Customer (revenue concentration) ~$300B of RPO; OpenAI's ability to fund the contract is the single largest swing variable for the AI thesis.
OCI AI capacity depends on Nvidia GPU supply, price and roadmap; a $40B Nvidia GPU order in May 2026 underscores the dependency.
Buildout is debt-funded ($134.6B); ~$40B more financing planned. Rate/spread moves directly hit the model's viability.
On-time, high-utilization capacity delivery (power, land, construction) is required to convert RPO to revenue at target margins.
Partner + competitor Oracle DB distribution rides on partners who are also OCI's largest competitors.
Advantages
- Largest contracted AI-infra backlog relative to revenue ($638B RPO ≈ 9x sales)
- Fastest-growing major cloud (OCI +77% FY26 / +93% Q4)
- Profitable legacy database/ERP cash engine to partially self-fund
- Multicloud reach that neutralizes the rivals' distribution edge
- Brand/trust with regulated enterprise + sovereign buyers
Weaknesses
- OCI gross margin ~14% vs 30-40% at AWS/Azure
- Negative FCF (-$23.7B) and rapidly rising leverage ($134.6B debt)
- ~$300B single-customer concentration (OpenAI)
- Weaker balance sheet than MSFT/AMZN/GOOGL funding similar-scale capex
- Mature, single-digit SaaS growth outside the AI story
Bottlenecks
- Power and data-center construction capacity to deliver contracted GPU clusters on schedule
- Nvidia GPU allocation and lead times
- Cost and availability of debt to fund ~$70B FY27 net capex
- OCI utilization ramp — idle capacity destroys the already-thin ~14% gross margin
- Skilled data-center / AI-infra labor for multi-country buildout
Top signals & trends
Top signals
Demand is being signed, not just talked about; $75B prepaid/BYO-hardware de-risks funding.
Capital consumption + leverage are the market's central worry — drove the post-earnings drop.
Sentiment shifted from AI-darling to funding-skepticism; valuation has compressed materially.
Management confidence the backlog converts; sets a high but visible bar.
Bull if utilization ramps as promised; bear if it stalls — the crux of the debate.
Trends
Primary tailwind behind OCI's +93% and the RPO surge; also the source of bubble risk if it reverses.
Validates the demand but raises competition and the bar Oracle must clear from a weaker balance sheet.
A debt-funded buildout is acutely exposed to tighter credit; spreads on Oracle's record bond issuance are a watch item.
Multi-country OCI regions and regulated-buyer trust open a differentiated lane vs Big Three.
Macro de-rating of stretched AI names; Oracle is among the most capex-levered, so it is hit hardest on risk-off.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
GPUs for OCI superclusters; ~$40B order in May 2026 — the dominant hardware dependency.
MI-series GPUs + EPYC CPUs as a second AI-accelerator/compute source.
Networking/custom-silicon and data-center connectivity for large GPU clusters.
High-speed data-center switching/fabric for AI clusters.
Power and thermal/cooling infrastructure for dense GPU data centers.
Server/rack hardware integration for AI infrastructure deployments.
~$300B 5-yr compute contract (from CY2027) — the single largest RPO concentration.
Named among new large-contract customers contributing to RPO growth.
Both supplier and a cloud customer for GPU capacity, adding to RPO diversification.
SpaceX (xAI) / enterprise + sovereign AI buyers AI labs and governments seeking GPU capacity outside Big Three allocation — OCI's 'GPU of last resort' niche.
Dominant #2 cloud (~25% share), OpenAI's primary partner, vast balance sheet and custom silicon; forward P/E ~19x.
#1 cloud (~31% share), Trainium/Inferentia custom AI chips, deepest IaaS moat and FCF to fund capex.
#3 (~11% share), TPU custom silicon and Gemini/Vertex AI ecosystem; fastest-improving margins.
Key supplier but also competes via DGX Cloud and direct GPU-cloud relationships with neoclouds.
Pure-play GPU neocloud chasing the same AI-infra contracts; similar debt-funded model.
Primary rival in enterprise ERP/applications where Oracle's legacy moat sits (less relevant to OCI).