top of page
Oracle Cloud Infrastructure
Technology
Saas Platforms
Cloud Infrastructure / IaaS
Won a specific slice of enterprise workloads not by outspending AWS and Azure on breadth, but by rebuilding its cloud from scratch as the one place Oracle databases and enterprise applications run cheapest and fastest.
1
MODEL
BUSINESS MODEL
Infrastructure Platform
model bm
HOW THEY BUILT IT
Oracle scrapped its first-generation cloud and rebuilt OCI as a 'Gen 2' cloud engineered specifically for Oracle database workloads with predictable performance and lower egress costs than AWS/Azure; pursued multicloud interconnect partnerships (Azure, Google Cloud) letting customers run Oracle databases co-located with other clouds' compute.
HOW TO ARCHITECT IT
1) Don't try to out-build hyperscalers' full breadth — rebuild around the workload you already own the software layer for. 2) Compete explicitly on price against AWS/Azure since enterprise buyers already running Oracle software are price-sensitive about migration costs. 3) Build multicloud interconnects with competitors rather than insisting on an all-or-nothing migration.
DISTRIBUTION MODEL
Enterprise Sales
dm
HOW THEY OPERATIONALIZED
Direct enterprise sales leveraging Oracle's existing database/applications customer base; custom-quoted contracts bundling OCI with existing licensing; multicloud interconnect agreements extend distribution through competitors' own channels.
HOW TO REPLICATE WHAT WORKED
Lead with existing Oracle customers facing a cloud-migration decision, price aggressively on egress/database-workload cost, and use multicloud interconnects to win workloads even inside accounts primarily committed to a competing hyperscaler.
| PATTERNS OF THIS MODEL
PATTERNS IN SECOND-WAVE CLOUD BUILT AROUND AN OWNED WORKLOAD:
1. DO NOT FIGHT A HYPERSCALER ON BREADTH — FIGHT ON THE WORKLOAD YOU ALREADY OWN THE SOFTWARE FOR. OCI exists because Oracle owns the database; the cloud is a defensive wrapper around an installed base, not a general-purpose bet.
2. A FAILED FIRST ATTEMPT IS SURVIVABLE IF YOU SCRAP IT COMPLETELY. Oracle abandoned its first-generation cloud and rebuilt "Gen 2." Incumbents that patch their first cloud attempt stay permanently behind.
3. PRICE AND EGRESS ARE THE ONLY LEVERS A CHALLENGER CONTROLS. Buyers already running your software are price-sensitive about migration, not about features. Competing on cheaper egress is a structural attack on the hyperscaler business model, not a discount.
4. MULTICLOUD INTERCONNECT BEATS ALL-OR-NOTHING MIGRATION. Letting customers run your database next to AWS/Azure/Google compute concedes the compute war to win the data war — and data is the sticky layer.
5. THE INSTALLED BASE IS THE MOAT AND THE CEILING. Growth is bounded by the number of Oracle software customers; net-new greenfield workloads default elsewhere.
FOR FOUNDERS: the transferable move is defending an owned workload with infrastructure, not building infrastructure and hoping for workloads.
What companies with this model reveal
| OPPORTUNITY INTELLIGENCE
GOLDMINE 1 — REBUILD AROUND THE WORKLOAD YOU ALREADY OWN.
Standard: you cannot out-breadth a hyperscaler, but you can be structurally better at the one workload whose software layer you control. Oracle scrapped its first-generation cloud and re-engineered OCI specifically for Oracle database performance. The rule: do not fight on the incumbent's axis; fight where you own the layer above or below them.
GOLDMINE 2 — INTERCONNECT WITH YOUR COMPETITORS RATHER THAN DEMANDING EXCLUSIVITY.
Standard: multicloud interconnects with Azure and Google Cloud let customers keep their existing compute while moving the database workload to you. All-or-nothing migration demands lose deals that partial migration wins. This is the single most copyable move here for any infrastructure challenger.
GOLDMINE 3 — PRICE ON THE LINE THE BUYER ALREADY RESENTS.
Standard: egress fees are the most disliked charge in cloud. Competing explicitly and loudly on that specific line converts a technical comparison into a budget argument the CFO understands.
THE PIT — TRAILING POSITIONS IN CAPITAL-INTENSIVE CATEGORIES REQUIRE PERMANENT SUBSIDY.
Cloud infrastructure is a capex race. A challenger holds share by out-spending or out-discounting, and both compress margin indefinitely. Verify current OCI capacity, capital commitments and AI-related backlog directly before relying on any figure — this area has moved fast and I have not confirmed current numbers.
THE SECOND PIT — CONCENTRATION IN A SMALL NUMBER OF ENORMOUS AI CONTRACTS.
Large committed-capacity deals transform a reported backlog and concentrate customer risk into a handful of counterparties. Read backlog growth and customer concentration together, never separately.
MOVE WITH CAUTION — YOUR MOAT IS THE DATABASE, NOT THE CLOUD.
If the workload you were built for migrates to open-source or cloud-native alternatives, the differentiated cloud underneath it becomes an undifferentiated one.
Untapped Business Model / Gaps / Goldmines / Pits
Patterns & Insights
2
MARKET
mkt mt es
MARKET TYPE
Consolidated Market
WHY THEY WON
Cloud infrastructure has consolidated around AWS, Azure, and Google Cloud. OCI won a defensible share by owning the niche of enterprise customers already running Oracle Database and applications, facing real switching costs elsewhere.
ENTRY STRATEGY
Greenfield Entry
EXECUTION
Oracle entered directly with its own rebuilt Gen 2 architecture rather than acquiring a cloud provider, using its existing enterprise software base as the go-to-market lever.
FOOTHOLD STRATEGY
fs
Beachhead Strategy
The beachhead was Oracle's own existing database/applications customers facing an inevitable cloud-migration decision, a captive audience with switching costs favoring OCI.
GROWTH CAMPAIGN
CAMPAIGNS THAT WORKED
Public, aggressive price-comparison marketing against AWS/Azure on egress fees, paired with multicloud interconnect announcements as high-visibility partnership news.
KEY LEARNING
If you're a smaller player in a hyperscaler-dominated market, identify the specific workload where existing software ownership creates real switching-cost friction, and win that slice on price and coexistence rather than replacement.
gc
Market Context
| MARKET INTELLIGENCE
THE STANDARD: In a consolidated market you entered late, you do not win share — you win the SUBSET OF WORKLOADS whose migration cost to a rival is prohibitive because you already own the layer beneath them.
RULE 1 — LATE ENTRY INTO A CAPITAL-INTENSIVE CONSOLIDATED MARKET REQUIRES AN EXISTING CAPTIVE BASE, NOT A BETTER PRODUCT.
AWS, Azure and Google Cloud hold the large majority of public cloud infrastructure. A fourth entrant with no structural advantage loses on catalogue breadth, region count and developer mindshare. OCI's viable claim was never "better cloud"; it was "the only place your Oracle Database licence, support terms and performance characteristics carry across unchanged."
RULE 2 — LICENSING TERMS ARE THE MOAT THAT LOOKS LIKE A PRODUCT FEATURE.
When the vendor controls how its own database is licensed on competitors' infrastructure, the migration maths is set by contract, not by engineering. Any founder whose product depends on someone else's licensing should read this as the risk; any founder who owns a licence should read it as the asset.
RULE 3 — CONSOLIDATED INFRASTRUCTURE MARKETS ARE WON ON EGRESS, PRICING PREDICTABILITY AND PROCUREMENT, NOT ON FEATURES.
The credible late-entrant differentiators in this category have been aggressive egress pricing, simpler price lists and bundled enterprise agreements. These are commercial-model attacks on incumbents who cannot match them without repricing enormous existing revenue.
RULE 4 — MULTI-CLOUD INTERCONNECT TURNS A CONSOLIDATED MARKET FROM ZERO-SUM INTO SHARED.
Direct interconnect arrangements with Microsoft, and Oracle Database services running inside AWS, Azure and Google Cloud regions, mean the database franchise travels even where the infrastructure does not. In a market you cannot win outright, monetising your layer inside rivals' estates is the rational play.
RULE 5 — THE AI CAPACITY CYCLE HAS RE-OPENED THIS CONSOLIDATED MARKET, AND THAT CUTS BOTH WAYS.
Large AI training and inference commitments have driven very large reported backlog figures across the hyperscalers, Oracle included, alongside heavy capital expenditure and debt. Contracted backlog is not revenue, and concentration in a small number of AI counterparties is a genuine credit and disclosure risk. Figures in this area move quarterly and public estimates disagree — verify against the latest filings rather than any secondary summary.
MARKET TYPE: Consolidated Market (public cloud infrastructure), defended via database-licence captivity and re-opened by AI capacity demand.
| MARKET ENTRY PLAYBOOK
THE STANDARD: A LATE ENTRANT WITH AN EXISTING INSTALLED BASE SHOULD NOT COMPETE ON THE CATEGORY'S GENERAL PROMISE. It should compete on the one workload its installed base cannot move anywhere else.
RULE 1 — REBUILD RATHER THAN ACQUIRE WHEN THE CATEGORY'S ECONOMICS LIVE IN THE ARCHITECTURE.
Buying a second-tier cloud provider would have bought customers and a cost structure that could never match a hyperscaler's. A second-generation rebuild was slower and produced a network and isolation model that could be positioned as materially different.
RULE 2 — LEAD WITH THE WORKLOAD YOU ALREADY OWN, NOT WITH THE PLATFORM.
The entry wedge for a late cloud entrant is its own database and application estate. Every incumbent-adjacent entrant has an equivalent: the workload where switching to you is the path of least resistance.
RULE 3 — PRICE ON THE DIMENSION THE LEADERS MONETISE HARDEST.
Egress and network charges are where hyperscaler bills surprise customers. Attacking a competitor's most resented line item is a cleaner entry than attacking its headline rate.
RULE 4 — CO-LOCATION AND MULTICLOUD PARTNERSHIPS TURN A DISADVANTAGE INTO DISTRIBUTION.
A late entrant that runs its estate inside the leaders' data centres converts "customers will not leave AWS or Azure" from an objection into an entry route.
RULE 5 — AN INSTALLED-BASE ENTRY INHERITS THE INSTALLED BASE'S OPINION OF YOU.
Existing commercial relationships supply access and also supply scepticism about licensing practices. Access is not goodwill.
EVIDENCE: Oracle rebuilt OCI as a second-generation architecture rather than acquiring a cloud provider, and went to market through its own database and applications base, aggressive egress pricing, and multicloud arrangements placing Oracle database services inside competitors' data centres. Oracle's cloud infrastructure revenue and its very large reported AI-related contracted backlog have been growing quickly but were not re-verified in this pass — take current figures from Oracle's own quarterly filings, and note that backlog is not revenue.
How to enter
| FOOTHOLD STRATEGY PLAYBOOK
THE STANDARD: A CAPTIVE INSTALLED BASE IS A BEACHHEAD ONLY WHERE THE CUSTOMER FACES A FORCED DECISION AND YOU CONTROL THE THING THEY CANNOT MOVE. Owning an account is not the same as being chosen by it.
RULE 1 — Enter where you own the DATA GRAVITY, not the relationship.
The workload that cannot easily leave is the one whose licensing, performance and support terms you set. Migrating around it is expensive and risky; migrating with you is the cheap path. Relationships get re-tendered; gravity does not.
RULE 2 — PRICE THE MIGRATION, NOT THE PRODUCT.
When you enter years behind on features, the credible offer is that running your existing estate costs materially less on your cloud than on a rival's. This is a discount strategy dressed as a strategy — accept it consciously.
RULE 3 — A LATE ENTRANT MUST FIND THE WORKLOAD THE LEADERS SERVE WORST, NOT COMPETE ON BREADTH.
Bare-metal, high-I/O database workloads and, later, dense GPU capacity are winnable because they are capacity-constrained rather than feature-constrained. General compute against an eight-year head start is not.
RULE 4 — CAPTIVE BASES PRODUCE MIGRATION REVENUE, NOT NET-NEW DEVELOPERS.
The greenfield developer chooses on documentation, community and default. Winning migrations while losing new formation means your growth stops the day the estate finishes moving.
RULE 5 — THE MULTI-CLOUD PARTNERSHIP IS THE ADMISSION AND THE ANSWER.
Placing your irreplaceable workload inside a rival's data centre concedes the destination and keeps the gravity. It is a rational move for a late entrant and a warning sign for anyone whose only wedge is lock-in.
EVIDENCE: OCI entered years after AWS and Azure and built its position on Oracle's own database and applications customers facing an unavoidable cloud decision. Oracle has since disclosed very large remaining performance obligations driven by AI infrastructure contracts — a second, capacity-led beachhead that has little to do with the original installed-base thesis. Figures here move every quarter; check the current 10-Q rather than any summary.
CHECKLIST: (a) Name the asset the customer cannot move. (b) Price the migration. (c) Pick the capacity-constrained workload. (d) Track net-new formation separately from migration. (e) Know when the estate runs out.
How to get the first strong position
MARKET PATTERNS & PLAYBOOK
3
MONEY
money rev pri
REVENUE MODEL
Usage-Based / Contract Revenue
PRICING MODEL
Competitive Pricing
WHY THEY WON
Consumption-based pricing for compute/storage/networking with lower published egress rates, combined with enterprise contract revenue bundling OCI with existing licensing agreements.
Publicly emphasizes lower networking/egress costs relative to AWS/Azure as a headline differentiator, with predictable published rate cards.
TARGET AUDIENCE
CUSTOMER BUYING BEHAVIOUR
tg cb
Existing Oracle Database and enterprise-applications customers, plus enterprises running database-intensive workloads.
Committee-led enterprise procurement, often initiated by an existing Oracle software renewal/migration decision.
| PRICING INTELLIGENCE
What makes this model effective & make customers pay
THE STANDARD: Competitive pricing is not undercutting the leader on the headline rate. It is finding the line item where the leader's pricing is genuinely punitive, making yours free or near-free, and forcing every comparison onto that line.
RULE 1 — ATTACK THE LOCK-IN FEE, NOT THE HEADLINE RATE.
Compute prices across major clouds are close enough that buyers negotiate them. Egress is where incumbents make leaving expensive. OCI has priced outbound transfer at roughly $0.0085/GB after a 10TB monthly free allowance against roughly $0.08-$0.09/GB at the major alternatives — a gap of around 10x that decides high-transfer workloads before compute is even discussed. Find your category's egress equivalent and zero it.
RULE 2 — A FREE ALLOWANCE THAT COVERS THE MEDIAN CUSTOMER IS A PRICE OF ZERO WITH A SAFETY VALVE.
Ten terabytes exceeds what most workloads move, so most customers experience egress as free while the meter still protects against extreme usage. This is more defensible than genuinely free, and it markets identically. Note that sources disagree on the current state: at least one 2026 report describes OCI eliminating outbound transfer charges entirely across all commercial regions from February 2026, while most contemporaneous pricing pages and comparisons still describe the 10TB-free-then-$0.0085/GB structure. Verify against Oracle's live pricing page before quoting either.
RULE 3 — GLOBAL PRICE CONSISTENCY IS A PRODUCT FEATURE FOR MULTI-REGION BUYERS.
Charging identical rates in every region, including government regions, removes an entire category of forecasting work that competitors' regional variance creates. Simplicity is worth real money to enterprises and costs the vendor almost nothing.
RULE 4 — GRANULARITY IS A PRICE CUT THAT LOOKS LIKE ENGINEERING.
Allowing scaling by a single CPU core rather than forcing a doubling of instance size means customers stop paying for capacity they do not need. The saving is real, the marketing is technical, and the competitor cannot match it without redesigning their instance families.
RULE 5 — COMMITMENT DISCOUNTS ARE HOW INFRASTRUCTURE CONVERTS PRICE ADVANTAGE INTO LOCK-IN.
Universal-credit style prepayment programmes are reported at 30-70% off pay-as-you-go depending on commitment size and term, with the deepest bands at multi-million-dollar annual commits. Treat these figures as third-party benchmark estimates rather than published list terms; Oracle does not publish a universal discount schedule.
RULE 6 — LICENSING ADJACENCY IS THE REAL WEAPON WHEN YOU OWN THE WORKLOAD.
Bring-your-own-licence economics for buyers already holding Oracle database licences produce savings no infrastructure price cut can match. When you own a piece of software the customer is already paying for, your cloud pricing can be aggressive because the margin lives upstream.
THE WILLINGNESS-TO-PAY INSIGHT: Nobody chooses infrastructure on rate cards — they choose it on the bill they are afraid of. Make the frightening line item disappear and buyers will accept a narrower service catalogue, weaker ecosystem and smaller talent pool to get it. In commoditised markets, willingness to pay is governed by the worst part of the incumbent's invoice, not the average of it.
PRICE & REVENUE
| Revenue Risk - The biggest threat to revenue stability
THE STANDARD: A backlog is a promise to spend, not a promise to earn. When contracted revenue grows faster than the capital, power and construction needed to serve it, the revenue risk moves from demand to delivery — and from the income statement to the balance sheet.
RULE 1 — RPO IS VISIBILITY, NOT SAFETY.
Remaining performance obligations went from $138B (close of FY2025) to $455B, $523B, $553B and $638B across FY2026 — up 363% year-on-year at Q4. That is more than eight years of current revenue committed, and every dollar of it requires data centres that do not exist yet.
RULE 2 — DECELERATING SEQUENTIAL ADDITIONS ARE THE REAL SIGNAL INSIDE A RISING BACKLOG.
The quarterly increments were roughly +$300B, then +$68B, then +$29B, then +$85B. A backlog can keep setting records while the booking rate collapses; read the deltas.
RULE 3 — CUSTOMER CONCENTRATION IS THE DEFINING EXPOSURE, AND IT IS ACKNOWLEDGED.
A single reported ~$300B five-year OpenAI contract dominates the backlog. Morningstar named concentration explicitly. When one counterparty's financing plans move, your revenue schedule moves with them.
Evidence: the Abilene, Texas campus was reported at roughly 1.2GW of eventual capacity, some 800MW below the originally planned 2GW, attributed to financing constraints at the customer.
RULE 4 — RENTING GPUs EARNS A WORSE MARGIN THAN SELLING LICENCES, AND THE MIX SHIFT IS PERMANENT.
OCI grew 68% to $4.1B (Q2 FY26) and 84% to $4.9B (Q3 FY26) while software licences and support declined ~3%. Non-GAAP operating margin slipped to 43% from 44%. High-growth revenue that dilutes margin is still dilution.
RULE 5 — THE CASH COST OF GROWTH IS THE ACUTE RISK, NOT THE REVENUE.
Trailing twelve-month free cash flow was reported at negative $13.18B, debt at roughly $153.1B, with plans to raise $45-50B in a single fiscal year for capacity. Shares fell about 10% after the Q4 print on the funding plan, and the stock was down roughly 23% year-to-date in early March 2026.
RULE 6 — PREPAYMENT AND CUSTOMER-SUPPLIED HARDWARE MITIGATE THE RISK AND RELOCATE IT.
Oracle states most of the Q3/Q4 RPO increase involved customers prepaying for GPUs or supplying them directly. That reduces Oracle's equipment funding and increases dependence on those specific customers' continued solvency and commitment.
RULE 7 — DELIVERY PACE IS NOW A REPORTED OPERATING METRIC.
1.2GW delivered across all of FY26, with close to 1GW targeted in Q1 FY27 alone. When the constraint is power, land and construction, execution risk is the revenue risk.
WHERE SOURCES DIFFER: analyst estimates of eventual Stargate-related annual revenue range from roughly $30B to $60B, with material recognition not expected before FY2028. Price targets in mid-2026 ranged from $190 to $330 on the same facts.
Where the model can break
4
MOTION
GROWTH EXPANSION MODEL
COMPETITIVE STRATEGY
motion ge cs
Ecosystem Expansion
HOW THEY EXPAND
Expanded from its captive Oracle base into multicloud interconnect partnerships with Azure and Google Cloud, winning workloads even inside accounts committed to a competing hyperscaler.
Focus Strategy
HOW THEY COMPETE
Rather than compete across the full breadth of general-purpose cloud services, focuses specifically on database and enterprise-application workloads.
GROWTH ENGINE
GTM
ge n gtm
Partnership Growth
Loop: multicloud interconnect deals let OCI win Oracle-database workloads inside accounts otherwise committed elsewhere → each deployment demonstrates measurable benefits → results become case-study evidence for the next account. Depends on Oracle's continued incentive to interconnect with hyperscalers.
Direct enterprise sales, public price-comparison marketing, and high-visibility multicloud interconnect partnership announcements.
SUSTAINING MOATS
Switching Costs, High Customer Lock-In, Brand Power, Technology Advantage (complex enterprise scenarios)
moat
Because OCI's architecture is purpose-built around Oracle Database's performance characteristics, that advantage is structurally difficult for AWS/Azure to fully replicate.
| MOAT INTELLIGENCE
THE STANDARD: A late entrant to a scale market cannot win on breadth. It wins by being the ONLY ACCEPTABLE HOME for workloads it already owns elsewhere — and by pricing the one dimension the leaders profit from most.
RULE 1 — YOUR INSTALLED BASE IN AN ADJACENT CATEGORY IS THE ONLY DEFENSIBLE BEACHHEAD A LATE ENTRANT HAS.
Oracle did not win general-purpose cloud share. It won the right to host Oracle databases, where licensing terms, performance guarantees and support accountability make the vendor's own cloud the path of least resistance for its own customers.
RULE 2 — ATTACK THE COMPETITOR'S MARGIN LINE, NOT THEIR HEADLINE PRICE.
Egress fees are the classic hostage mechanism in cloud. Pricing that dimension aggressively is a strategy aimed at the incumbent's business model rather than their price list — the one attack a leader finds genuinely painful to match.
RULE 3 — CONTRACTED BACKLOG IS A CLAIM, NOT A MOAT, UNTIL IT IS DELIVERED.
Enormous multi-year commitments require capex, power and silicon delivered on schedule. Judge these by revenue recognised and capacity energised, not by the headline booking.
RULE 4 — CONCENTRATED CUSTOMERS ARE CONCENTRATED RISK.
When a small number of AI labs represent a large share of committed backlog, you have swapped diversified enterprise revenue for counterparty exposure to companies that are themselves pre-profit. State that plainly in any assessment.
RULE 5 — SOVEREIGNTY AND REGION COVERAGE ARE A REGULATORY MOAT, and the one place a challenger can beat scale leaders on a specific deal.
EVIDENCE (with limits stated):
- OCI is Oracle's second-generation cloud platform, positioned on price/performance, dramatically cheaper data egress than the largest hyperscalers, and dedicated/sovereign region deployment options — including running OCI inside a customer's own data centre.
- Its most defensible position is Oracle-workload gravity: Oracle Database, Exadata and Fusion applications customers face licensing, performance and support incentives that favour Oracle's own infrastructure.
- I DID NOT VERIFY ORACLE'S CURRENT RPO, OCI REVENUE RUN-RATE, CAPEX GUIDANCE OR NAMED AI CUSTOMER COMMITMENTS IN THIS PASS. These figures have moved materially and repeatedly through 2025-2026 and are reported inconsistently in secondary coverage. Take them from Oracle's latest 10-Q or earnings release, not from summaries.
- Structural position: OCI remains a distant challenger to AWS and Azure in general-purpose market share by every published estimate, while growing considerably faster from a smaller base.
THE SIGNAL TO COPY: Oracle's cloud moat is not technology — it is that a large population of enterprises already run their most critical, hardest-to-move system on Oracle software. The transferable rule is that a late entrant needs a hostage, not a differentiator. If you have no adjacent installed base to convert, do not enter a scale market late.
Why this company remains defensible
ARR & TAKEAWAY
ARR Journey - what to do at each stage
PRE-$1M ARR — DO NOT ENTER THIS CATEGORY FROM ZERO
State the honest instruction first: hyperscale infrastructure is not a startup category. Entry requires capital measured in billions, and no founder should read this row as a template for building a cloud.
Read it instead as a lesson in second-mover strategy. If you are entering a category where the leaders are years ahead, you must find the workloads their architecture serves worst, not build a cheaper version of what they already do well.
Identify the customers who structurally cannot use the leader — competitors of the leader's parent, regulated industries, sovereignty-bound workloads.
REFUSE: a general-purpose positioning against an entrenched general-purpose leader.
$1–5M ARR — THE TRANSFERABLE MOVE: START FROM YOUR OWN INSTALLED BASE
If you have an existing customer base in an adjacent category, that base is your entry, not a new market. Oracle's route into cloud ran through the databases and enterprise applications its customers already ran and could not easily move.
Make migration from your own legacy product to your new product the cheapest path available, and fund it.
Price the transition, not the product: guaranteed cost reduction against the customer's current bill is a far stronger argument than feature comparison.
WATCH: the proportion of existing customers who have moved any workload at all.
$5–10M ARR — DIFFERENTIATE ON ARCHITECTURE THE LEADERS CANNOT COPY CHEAPLY
Build on a genuine architectural difference and price against it. Predictable pricing, low egress fees and bare-metal performance are choices with real economics behind them, not marketing positions.
Publish the price comparison explicitly. In a category where the leaders' billing is famously opaque, transparency is differentiation.
Target the workloads where the leaders' pricing hurts most: high-egress, high-bandwidth, GPU-intensive, data-heavy.
$10–50M ARR — WIN THE PARTNERSHIP THE LEADERS CANNOT OFFER
Make yourself the neutral option for companies that compete with the leaders' parents, and make multi-cloud interoperability a product rather than a concession. (Oracle's interconnect and database services running inside competing clouds is the clearest expression of this: sell your differentiated layer on someone else's infrastructure.)
Sell through the applications your customers already run rather than as raw infrastructure.
WATCH: workloads that started with you versus workloads that migrated. The second cohort proves the thesis; the first only proves availability.
$50–100M ARR — CAPACITY COMMITMENTS ARE THE BUSINESS MODEL NOW
In AI-era infrastructure, revenue is contracted forward through multi-year capacity commitments, and the constraint is your ability to finance and build data centres, not to sell.
Understand the risk that comes with it: large forward commitments concentrate revenue in a handful of counterparties and require enormous capital expenditure ahead of the cash.
Report contracted-but-unrecognised revenue separately and expect the market to scrutinise the gap between backlog and revenue.
$100M+ ARR — READ THE CONCENTRATION RISK, NOT THE BACKLOG
OCI has grown rapidly from a late start and Oracle has reported very large increases in remaining performance obligations driven by AI infrastructure commitments; the specific figures move every quarter and any number quoted here would be stale, so verify against the most recent 10-Q before using one.
The transferable instruction is about customer concentration: a backlog dominated by a small number of AI buyers is a different risk profile from diversified enterprise cloud revenue, and it should be underwritten as counterparty risk, not as demand.
For founders, the durable lesson is the second-mover one: you do not beat an entrenched leader by being cheaper at their game. You win the workloads their architecture, their pricing or their corporate parent makes impossible for them to serve.
COPY PLAYBOOK : What Worked → What Failed → What to Replicate → What to Avoid
THE STANDARD: You cannot win a hyperscale infrastructure race on breadth. You win the WORKLOADS where your existing software is the system of record, and you price on the specific line items the leaders are known to overcharge for.
HOW TO COPY — THE SEQUENCE:
1. Do not enter a capital-intensive infrastructure category unless you already own a workload the incumbents cannot run natively — for Oracle, its own database estate.
2. Lead with the customers facing a forced migration decision, where the alternative to you is a re-platforming project rather than a status quo.
3. ATTACK ON THE PRICING SORE POINTS, NOT THE PRICE LIST. Egress fees and database licensing are where the leaders' customers feel gouged; a differentiated price on a hated line item beats a lower headline rate.
4. Accept multicloud rather than fight it. Interconnects and in-cloud database services placed inside a competitor's data centre win workloads inside accounts that will never leave the competitor.
5. Sell the licensing economics: running your own database on your own cloud can be made structurally cheaper than running it on a rival's, and that gap is a policy decision you control.
6. Compete for the AI capacity buildout only if you can fund data centres; capacity commitments, not features, are the currency.
WHAT WORKED:
- Entering through the installed base rather than through developers, which is the only viable entry for a late arrival to cloud.
- Using multicloud interconnect as a wedge, capturing database workloads inside accounts primarily committed to AWS, Azure or Google.
- Aggressive, published price differentiation on egress and compute, which gave enterprise procurement a quantified argument to run a second cloud.
WHAT DID NOT WORK / THE CAUTIONS:
1. LATE ENTRY INTO A CONSOLIDATED INFRASTRUCTURE MARKET IS ONLY SURVIVABLE WITH A BALANCE SHEET AND A CAPTIVE WORKLOAD. Almost every telco and enterprise-vendor cloud that lacked one exited the category; a founder cannot replicate this playbook at all, and should read it as a lesson about structural advantage rather than a template.
2. DEVELOPER MINDSHARE WAS NEVER WON. OCI's strength is procurement-led and licensing-led; the bottom-up developer adoption that built AWS remains with the incumbents, which caps greenfield workloads.
3. PRICE-LED POSITIONING INVITES MATCHING. Hyperscalers can adjust egress and commit discounts at will; a price advantage held against better-capitalised rivals is a policy, not a moat.
4. CURRENT OCI-SPECIFIC METRICS SHOULD BE READ FROM ORACLE'S FILINGS DIRECTLY. Cloud infrastructure figures, capacity commitments and backlog in this segment have moved fast and are frequently misreported by secondary sources; treat any single third-party number as an estimate.
bottom of page