Pricing & Onboarding Reference

Strategic Manifest

Comprehensive pricing reference for cloud advisory conversations. All prices reflect April 2026 list rates in US East regions, sourced from official provider pricing pages and cross-checked against Cast AI, Cloud Price, EgressCost.com, and FinOps practitioner data. Real-world pricing varies by region, commitment, and negotiated discount (EDP, MACC, etc.) — always validate with the provider's calculator before committing. After Q2 CY2026 earnings, all four hyperscalers reported 2027 capacity largely reserved and 2028 partly reserved — GPU-tier instance availability is now the binding constraint on onboarding timelines for AI workloads at scale.

Jump to 1. Compute 2. Storage 3. Egress 4. Kubernetes 5. Free Tiers 6. Startup Credits 7. Real Workloads 8. Onboarding Costs 9. Hidden Costs 10. Advisory Take

The single most important pricing fact

List-price compute is engineered to match across all four providers

A 4-vCPU, 16GB RAM Linux instance in US East costs ~$0.19/hour on AWS, Azure, and GCP — within fractions of a cent. The providers actively monitor each other's headline rates and match them. Real cost differentiation lives in egress, block storage, Kubernetes control plane fees, commitment flexibility, and hidden infrastructure costs (NAT gateways, cross-AZ transfer, support plans). If a vendor pitch says "we're cheaper on compute," they're either talking about a non-equivalent instance, a multi-year commitment, or already-included discounts. The headline number is a wash.

1. Compute Pricing

General-purpose VM, equivalent specs (4 vCPU / 16 GB RAM), Linux, US East. On-demand vs. 1-year commit vs. 3-year commit vs. spot/preemptible.

Pricing Model	AWS · m6i.xlarge	Azure · D4s v5	GCP · n2-standard-4	OCI · VM.Standard3.Flex
On-demand4 vCPU / 16 GB, hourly	$0.192/hr ~$140/mo	$0.192/hr ~$140/mo	$0.194/hr ~$142/mo · auto SUD ~20%	$0.106/hr ~$77/mo · 2 OCPU + 16GB · per-OCPU billing
1-year commitReserved / Savings Plan / CUD	~37% off most flexible — applies to any EC2 family	~40% off Reserved tied to instance family + region	~37% off CUD by vCPU/memory in region; auto-applies	~25–30% off Universal Credits, annual commit
3-year commitMaximum discount	~57% off Savings Plan flexibility	~60% off + up to 76% w/ Hybrid Benefit	~60% off Resource-based CUD	~30% off + Oracle Support Rewards (25% rebate)
Spot / PreemptibleInterruptible workloads	up to 90% off 2-min interruption warning	up to 90% off Spot VMs · Bs not supported	60–91% off most stable spot pricing	flat 50% off simple, predictable
ARM computeProvider's own silicon	Graviton5 25–35% better price/perf vs x86	Cobalt 100 Azure ARM, parity pricing	Axion ~10% cheaper than C3	Ampere A1 cheapest ARM in market · 4 OCPU free forever
Per-second billingGranularity	Linux only EC2 Linux + ECS/Fargate	Containers only VMs are per-minute	All VMs Linux + Windows	All compute consistent
Regional price varianceSame VM, different regions	Varies ±25% London/Brazil materially higher	Varies ±20% EU/APAC premium	Varies ±15% most consistent of big 3	Globally consistent same price in every region

2. Storage Pricing

Object storage hot tier in $/GB/month. Headline rates only — request charges, retrieval fees, and minimum-storage durations add 30–70% to a real bill.

Storage Type	AWS	Azure	GCP	OCI
Object · HotStandard, frequent access	$0.023/GB S3 Standard	$0.018/GB Blob Hot · cheapest of big 4	$0.020/GB GCS Standard	$0.0255/GB but bundled w/ free egress
Object · ColdInfrequent access	$0.0125/GB S3 Standard-IA	$0.0045/GB Blob Cold (new tier)	$0.010/GB Nearline	$0.0026/GB Infrequent Access
ArchiveLong-term, rare access	$0.00099/GB Glacier Deep Archive	$0.00099/GB Archive tier	$0.0024/GB Archive multi-region	$0.0026/GB Archive Storage
Block · SSDVM-attached	$0.10/GB EBS gp3	$0.075/GB Premium SSD	$0.17/GB Persistent Disk SSD · highest of 4	$0.0425/GB Block Volume
Vector storageRAG / embeddings	$0.06/GB S3 Vectors (GA Dec '25)	Bundled Cosmos DB / AI Search	Bundled Vertex AI Vector Search	Native Oracle DB 23ai
Request chargesPUT / GET (per 1K)	$0.005 / $0.0004 Standard tier	$0.0055 / $0.0044 Hot tier	$0.005 / $0.0004 Standard ops	$0.0034 / $0.0034 flat

3. Egress Pricing · The Silent Budget Killer

Egress (data leaving the cloud) is the most asymmetric pricing category in cloud computing. Ingress is free everywhere; egress ranges from $0.0085/GB to $0.12/GB. This is where multi-cloud strategies live or die.

First-tier internet egress, US East · per GB after free allowance

OCI

$0.008510 TB FREE/mo · ~10× cheaper

Azure

$0.087100 GB free/mo · cheapest of big 3

AWS

$0.090100 GB free/mo

GCP

$0.120100 GB free/mo · highest entry tier

Volume Scenario	AWS	Azure	GCP	OCI
10 TB/monthMid-size SaaS	$912	$868	$1,221	$0 under 10TB free tier
50 TB/monthHeavy SaaS / streaming	$4,300	$4,000	$5,200	$340 40TB billable × $0.0085
100 TB/monthEnterprise / CDN-heavy	$9,000	$8,072	$10,400	$765 90TB billable
Cross-region replication500 TB/mo (DR)	$10,000 $0.02/GB	$10,000 $0.02/GB	$5,000 $0.01/GB intra-US	Varies port-based, not per-GB
Egress when leavingPermanent migration off platform	Charged EU DMA exception only	Free waived for migrations off Azure	Charged	Mostly free 10TB/mo allowance helps

4. Kubernetes Control Plane · Small Fee, Big Multiplier

For platform teams running 10+ clusters (production / staging / per-environment / sandbox), control plane fees compound.

Item	AWS · EKS	Azure · AKS	GCP · GKE	OCI · OKE
Control planePer cluster, per month	$73/mo $0.10/hr	$0 free, always	$73/mo + 1 free zonal cluster/account	$0 free
Cost · 10 clusters/yearJust control plane	$8,760	$0	$7,884 9 paid + 1 free	$0
Maturity / DXOperational experience	Most mature deepest IAM / VPC / EBS integration	Strong best for Windows containers + Entra	Most mature Google created Kubernetes	Adequate simpler, less ecosystem

5. Free Tiers · What You Get Before Billing

All four offer free resources, but the structure is wildly different. OCI is the most generous "always free" tier — useful to know for proof-of-concept work and small client pilots.

AWS

12-month + Always-Free

No upfront credit

EC2: 750 hrs/mo t3.micro for 12 mo
S3: 5 GB for 12 mo
RDS: 750 hrs db.t2.micro for 12 mo
Lambda: 1M requests/mo always free
DynamoDB: 25 GB always free
CloudWatch: 10 metrics + 1M API req free

Year 1 only for VMs · ~$8/mo after

Azure

$200 / 30 days + 12-mo + Always-Free

$200

Credit · first 30 days

VMs: 750 hrs B1s for 12 mo
Blob: 5 GB for 12 mo
SQL DB: 250 GB for 12 mo
Functions: 1M executions/mo always free
Cosmos DB: 1,000 RU/s always free
55+ services with always-free allotments

Best $-credit for evaluation

GCP

$300 / 90 days + Always-Free

$300

Credit · first 90 days

e2-micro VM: always free in us-central1/east1/west1
Cloud Storage: 5 GB always free
BigQuery: 1 TB queries + 10 GB storage/mo free
Cloud Functions: 2M invocations/mo free
Firestore: 1 GB + 50K reads/day free
e2-micro stays free forever — only big-3 to do this

Best for 90-day POC evaluation

OCI

$300 / 30 days + UNLIMITED Always-Free

$300

Credit + most generous always-free

Ampere A1: 4 OCPU + 24 GB RAM ARM (no expiry)
Block Storage: 200 GB
Object Storage: 20 GB Standard + 20 GB IA
Autonomous DB: 2 databases × 20 GB each
Egress: 10 TB/month always free
Load Balancer + VPN + Logging all included

Most generous always-free of any cloud

6. Startup Credit Programs

Each provider runs a startup program. Credits aren't free money — they're customer acquisition spend designed to build switching cost. Most experienced founders apply to multiple simultaneously and treat the credits as runway extension.

Program	Credit Ceiling	Typical Actual	Period	Requirements
Google Cloud for Startups · AI TrackMost aggressive program of 2026	Up to $350K $250K Y1 + $100K Y2	$25K–$200K	2 years	VC-backed AI startups + Google partner referral. Year 2 credits at 20% match (you spend $5, get $1).
Microsoft for Startups Founders HubBest for AI/.NET shops	Up to $150K + GitHub Enterprise + M365	$1K–$25K	12 months	$150K tier requires Microsoft accelerator partnership or strong VC. Bootstrapped path: $1K → $5K. Credits expire after 12 months.
AWS ActivateLargest reach	Up to $100K "Portfolio" tier	$1K–$25K	2 years	Activate Provider (VC/accelerator) referral required for full $100K. Self-serve "Founders" path is much smaller. Lifetime caps enforced.
Oracle for StartupsLess publicized	$5K–$500K tiered by stage	$5K–$50K	2 years	Less aggressive marketing. Strongest for AI startups needing GPU capacity given OCI's GPU economics. Often combined w/ NVIDIA Inception.

7. Real-World Workload Cost Scenarios

Three configurations clients commonly ask about. List pricing only — no commitment discounts applied. Teal-highlighted cell is the low-cost winner for each scenario.

Scenario 1 · Early-stage SaaS

2× app servers (2 vCPU/4GB each) · 1× managed Postgres HA (2 vCPU/4GB, 100GB) · 200GB object storage · 500GB/mo egress · CDN

AWS

~$420/mo

EC2 + RDS + S3 + CloudFront

Azure

~$415/mo

VMs + Azure SQL + Blob

GCP

~$380/mo

6–10% cheaper · auto SUD

OCI

~$300/mo

Cheapest list · fewer managed services

Scenario 2 · Enterprise data analytics pipeline

8 vCPU/32GB analytics VM · 10TB object storage · 50TB/mo egress · Data warehouse · 5TB queries/month

AWS

~$8,500/mo

Redshift instance-hour billing

Azure

~$8,200/mo

Synapse + Fabric

GCP

~$5,500/mo

BigQuery $5/TB scan beats per-hour 100×

OCI

~$6,200/mo

Autonomous DW + free egress saves ~$4K/mo

Scenario 3 · LLM production inference (medium)

Llama 70B-class · 4× A100 GPUs · 15TB/mo egress (high traffic API)

AWS

~$13,570/mo

p4d.24xlarge fractional

Azure

~$12,367/mo

ND A100 v4

GCP

~$11,200/mo

A2 instances · auto SUD

OCI

~$8,838/mo

35% cheaper than AWS · free egress

8. Onboarding Cost Ladder · The First 90 Days

The line items most commonly underbudgeted in client engagements. These costs are one-time or front-loaded — not part of the ongoing run-rate, but they determine whether year-1 ROI numbers actually land.

Cost Category	Typical Range	What Drives It
Discovery & assessmentFirst 2–4 weeks	$5K–$50K	Inventory current state, dependency mapping, TCO modeling, risk assessment. AWS Migration Evaluator and Azure Migrate automate ~70% of inventory but humans still validate. Skip this and you'll re-do it later at 3× the cost.
Account / landing zone setupFoundation	$10K–$80K	Multi-account / management group hierarchy, IAM/Entra/Cloud IAM, networking (VPCs/VNets/VCN), guardrails, FinOps tagging. Done badly here = years of cleanup. AWS Control Tower / Azure Landing Zones / GCP Foundation accelerate this.
Data migrationBandwidth + tools	$5K–$200K+	Ingress is free, but transferring 100 TB+ takes weeks over the internet. AWS Snowball ($300–$1K/device), Azure Data Box (similar), Google Transfer Appliance, OCI Data Transfer all charge per appliance + shipping. Database Migration Service tools are usually free but compute/replication isn't.
Training & certificationPer engineer	$500–$3K each	Cloud Practitioner / AZ-900 / Google Cloud Digital Leader: $100 exam + ~40 hrs prep. Solutions Architect / AZ-104 / Professional Cloud Architect: $165–$200 exam + 80–120 hrs. Plan 4–8 weeks of part-time study per engineer. For a 10-person platform team, budget $20K–$50K total in fees + opportunity cost.
Application refactoringThe biggest variable	$20K–$2M+	Lift-and-shift: cheapest, $5K–$50K per app. Refactor for cloud-native: $50K–$500K per app. Rewrite (e.g., monolith → microservices): $200K–$2M+. Most clients underestimate by 2–3× because the long tail of integration testing, monitoring, and runbook conversion gets skipped in initial scoping.
Professional services / SIExternal help	$25K–$1M+	Hourly rates: $150–$300 senior cloud architect; $200–$400 partner-firm rates. AWS, Azure, GCP all have funded migration programs (MAP, AMMP, RAMP) that offset 25–50% of partner-led migration costs for committed-spend customers. Clients leave this money on the table 80% of the time.
Support tier subscriptionRequired for production	$100–$15K/mo	Basic tier is included. Production needs Business / Standard ($100/mo or 3% of bill, whichever higher). Enterprise (24/7 TAM) is $15K/mo minimum or 10% of bill. Azure Premier Support and GCP Premium are similar. Most clients pick "Developer" to save money, then need Enterprise for production go-live and pay for both. Budget Business minimum from day 1.
Compliance & auditRegulated industries	$15K–$500K	HIPAA, SOC 2, PCI-DSS, FedRAMP, GDPR. Each cloud has compliance certs covering the platform, but YOUR workloads need separate audits. Budget for AWS Audit Manager / Azure Compliance Manager / Compliance Reports + external auditor + remediation. FedRAMP High alone is a 12–18 month, $500K–$2M effort.
FinOps tooling & governanceOften skipped, then added year 2	$0–$50K/yr	Native: AWS Cost Explorer / Azure Cost Management / GCP Cost Management — all free. 3rd-party (Vantage, Apptio, Finout, CloudZero, Spendark) typically 1–3% of cloud spend or flat $1K–$10K/mo. Pay-back in 60–90 days for any client over $50K/mo cloud spend.
Cross-cloud egress (if migrating)The "exit tax"	$0–$50K+	100 TB out of AWS = ~$9,000 in egress alone, plus rebuilding destination infra. Azure waives egress for permanent migrations off the platform (1 of the rare anti-lock-in moves). GCP and AWS don't. Multi-cloud architectures should sit data on the cloud with the most expensive egress (or use OCI's 10TB/mo free).

9. Hidden Costs the Calculators Miss

These don't appear on official pricing pages but show up in real bills. Account for them when modeling client spend.

NAT Gateway fees

$32–$45/mo per NAT

+ per-GB processing on AWS/Azure. A typical 3-AZ HA setup is $130/mo before any traffic. GCP Cloud NAT is similar. OCI NAT is included free.

Cross-AZ data transfer

$0.01/GB

Charged when traffic crosses availability zones inside the same region. Buried in "EC2-Other" on AWS bills. For chatty microservices this adds 5–15% to compute costs.

VPC endpoints

$7/mo + $0.01/GB

Per endpoint, per AZ on AWS. Restricting outbound internet means paying for endpoints to every service. Stack a 3-AZ + 10 services = $210/mo before any traffic.

Logging & monitoring

$2.76/GB ingested

Azure Monitor and CloudWatch Logs both bill for ingest. A noisy app at 100 GB/mo of logs costs $276/mo on Azure. Most teams discover this after a debug session.

Idle resources

35–40% waste

AWS reports 35–40% of EC2 instances are at least one size larger than the workload requires. Unattached EBS volumes, stopped instances with storage, idle Elastic IPs add 5–15% to most bills.

Aurora / Cosmos DB premium

2–3× managed RDS

"Cloud-native" databases (Aurora, Cosmos, AlloyDB) are 2–3× the cost of standard managed RDS/SQL. Worth it for the SLA, but easy to under-budget at scoping time.

Snapshot storage growth

Compounds monthly

EBS / managed disk snapshots accrue silently. Daily snapshots × 30-day retention × multiple environments = thousands per month with no policy. Always set lifecycle rules.

Data egress on inference

$0.09/GB · response tokens

LLM API calls return tokens that count as egress when delivered to end users on internet. High-traffic GPT/Claude apps can hit $1,000+/mo on egress alone before model costs.

10. What This Means for Advisory

Six positioning takeaways for JCB client conversations. These are the practical implications of everything above.

Don't lead with compute pricing

It's a wash and the client will eventually figure that out. Lead with egress economics, support tier, and the onboarding ladder. Those are the differences that compound into 6-figure annual deltas.

OCI is the egress play

For any client moving more than 10 TB/month off platform, OCI's 10 TB free + $0.0085/GB beyond is structurally 10× cheaper. For data-heavy SaaS, video, IoT, or analytics — model it. The savings often pay for migration within 12 months.

Onboarding is 70% of year-1 cost

For most enterprises, the first 12 months are ~30% run rate, ~70% migration / refactoring / training / professional services. Set client expectations correctly. The "cloud is cheaper" narrative gets blown up if year-1 sticker shock hits before steady state.

Always pursue funded migration programs

AWS MAP, Azure Migrate & Modernize (AMMP), GCP Rapid Migration Program (RAMP) all offset 25–50% of partner-led migration spend for committed-spend customers. Most clients don't know these exist. JCB should walk every Azure prospect through AMMP eligibility on the first call.

EDP / MACC negotiation unlocks at $1M/yr

AWS Enterprise Discount Program: 10–25% off list at $1M+/yr commit. Azure MACC: similar. Critical leverage moment is at $50K/mo run rate — that's when you should start pulling competing quotes from rival clouds before signing. This is the highest-ROI conversation to have with clients.

Support tier is non-optional for production

Most clients pick the cheap support tier to save money, then need it upgraded for production go-live and pay double. Build "Business / Standard" support into the day-1 budget ($100/mo minimum or 3% of bill). For regulated workloads, Enterprise tier ($15K/mo) is a non-negotiable line item.

Sources: AWS, Azure, Google Cloud, Oracle official pricing pages (April 2026) · Cast AI Cloud Pricing Comparison · CloudPrice.net · EgressCost.com · BitsFromBytes Cloud Hosting Calculator · Tech Insider AWS vs Azure 2026 · CloudZero Oracle Cloud Pricing Guide · SquareOps GCP-to-AWS migration guide · a16z infrastructure cost analysis · AWS Application Migration Service pricing · Spendark egress cost guide.
Important: All prices reflect April 2026 list rates and change without notice. Cloud providers update pricing pages without versioning. Always validate against the provider's official calculator before client commitments. Negotiated discounts (EDP, MACC, BYOL) can move the numbers 10–30%.
H1 2026 capacity note: After Q2 CY2026 earnings, all four hyperscalers reported 2027 capacity largely reserved and 2028 partly reserved. GPU-tier instance availability (H100, GB200, Trainium 2/3, Blackwell) is now the binding constraint on onboarding timelines for AI workloads at scale. Plan reservation conversations 6-12 months earlier than 2024-25 norms.
Compiled for JCB Technology · Cloud Advisory · Last reviewed August 03, 2026 · Refresh cadence: quarterly post-earnings for run-rate; semi-annually for onboarding ladder.