7 Examples of IaaS Companies: A CEO’s Decision Guide [2026]

Most founders pick an IaaS provider the same way they pick an office lease — based on what’s famil­iar, what the CTO already knows, and what­ev­er free cred­its showed up in the inbox. That’s a mis­take. The spe­cif­ic exam­ples of IaaS com­pa­nies you choose today shape three things that will still mat­ter when you sell the com­pa­ny: your gross mar­gin, your switch­ing cost, and the risk dis­count a buy­er applies to your exit mul­ti­ple.

This arti­cle walks through the 7 lead­ing IaaS providers, the eco­nom­ics of each choice for a SaaS com­pa­ny between $2M and $30M ARR, and a frame­work to pick the one that match­es your stage — not the one your engi­neer­ing team finds most inter­est­ing.

Infra­struc­ture as a ser­vice (IaaS) is the lay­er of cloud com­put­ing that gives you raw com­pute, stor­age, and net­work­ing on demand. You con­trol the oper­at­ing sys­tem, the run­time, and every­thing above it; the provider han­dles the phys­i­cal hard­ware, the data cen­ter, and the plumb­ing between regions. IaaS sits below plat­form as a ser­vice (PaaS) and soft­ware as a ser­vice (SaaS) in the stan­dard four-lay­er cloud stack (FaaS → IaaS → PaaS → SaaS). If you run a B2B SaaS com­pa­ny, you’re almost cer­tain­ly a cus­tomer of at least one IaaS provider — and the cost of that rela­tion­ship shows up as the largest line item in your cost of goods sold.

Why IaaS Choice Matters More Than Most CEOs Think

Pick the wrong IaaS and two things hap­pen. Nei­ther is obvi­ous until it’s expen­sive.

First, your gross mar­gin suf­fers. Healthy B2B SaaS com­pa­nies run 70–85% gross mar­gins. Infra­struc­ture is usu­al­ly the sin­gle biggest COGS line. A poor­ly archi­tect­ed AWS set­up at $5M ARR can eas­i­ly con­sume 20–25% of rev­enue — push­ing gross mar­gin below 70%, which is the floor most strate­gic acquir­ers use to sep­a­rate “soft­ware busi­ness” from “man­aged ser­vices busi­ness.” Below 70% gross mar­gin, your rev­enue mul­ti­ple at exit drops by 30–50%.

Sec­ond, you lock your­self in. Every IaaS provider has pro­pri­etary ser­vices that make life eas­i­er and exit hard­er. AWS DynamoDB, Azure Cos­mos DB, Google Big­Query, and Ora­cle Autonomous Data­base are not portable. Once your appli­ca­tion depends on them, mov­ing off costs 6–18 months of engi­neer­ing time you don’t have. Egress fees (the cost to move your data out) add anoth­er lay­er — a 10 TB data­base trans­fer off AWS can cost $900+ in raw band­width, and that’s before the re-archi­tec­ture work.

This is why the right fram­ing isn’t “which IaaS has the most fea­tures.” It’s “which IaaS gives me accept­able gross mar­gin at my cur­rent stage with­out paint­ing me into a cor­ner I can’t afford to leave.”

IaaS vs PaaS vs SaaS vs FaaS: What You’re Actually Buying

Before the ven­dor list, a quick clar­i­fi­ca­tion. The cloud stack has four lay­ers. You will use more than one.

Lay­er	What You Man­age	What the Provider Man­ages	Typ­i­cal Exam­ple
SaaS (Soft­ware as a Ser­vice)	Just your data and users	The entire appli­ca­tion	Sales­force, Hub­Spot, Notion
PaaS (Plat­form as a Ser­vice)	Your code and data	Run­time, OS, hard­ware, mid­dle­ware	Heroku, Google App Engine, Ver­cel
IaaS (Infra­struc­ture as a Ser­vice)	OS, run­time, code, data	Phys­i­cal hard­ware, vir­tu­al­iza­tion, net­work­ing	AWS EC2, Azure Vir­tu­al Machines, GCP Com­pute Engine
FaaS (Func­tions as a Ser­vice)	Just your func­tion code	Every­thing else	AWS Lamb­da, Azure Func­tions, Google Cloud Func­tions

Most SaaS com­pa­nies run on IaaS for the core appli­ca­tion and sprin­kle FaaS and PaaS on top for spe­cif­ic work­loads (event han­dlers, mar­ket­ing sites, inter­nal tools). The rea­son IaaS is the anchor: it’s the most flex­i­ble lay­er — which means it’s the most cus­tomiz­able to your appli­ca­tion’s spe­cif­ic needs, and it’s the hard­est to out­grow.

If you’re ear­ly ($2M ARR) and you want to ship fast with a small team, a pure PaaS like Heroku or Ver­cel often beats IaaS because you don’t spend engi­neer­ing time on infra­struc­ture. The moment your bill hits ~$5K/month on a PaaS, do the math on mov­ing to IaaS. That inflec­tion usu­al­ly shows up between $2M and $5M ARR.

The 4 Questions to Ask Before Picking an IaaS

These are the ques­tions I wish more CEOs forced on their CTOs before the ven­dor deci­sion gets locked in.

1. What does this choice do to my gross mar­gin in 18 months? Mod­el the infra­struc­ture bill against your ARR plan. If your gross mar­gin drops below 75% at next year’s ARR, you picked wrong. This is the sin­gle most impor­tant ques­tion, and it’s the one almost nobody asks — because engi­neers default to “AWS has every­thing” and founders assume the engi­neers know the cost impli­ca­tions. They usu­al­ly don’t, because the per­son choos­ing the provider isn’t the per­son read­ing the month­ly bill against the P&L.

2. How much would it cost to leave? If you had to migrate off this provider in 12 months, what would it cost in engi­neer­ing hours + egress fees + re-archi­tec­ture + down­time risk? If the answer is “we can’t, we’re stuck,” you just dis­cov­ered your lock-in expo­sure. Buy­ers will find this at dili­gence and apply a risk dis­count. Accept­able switch­ing cost for most $10M ARR SaaS com­pa­nies is 3–6 months of eng time. If you’re over a year, pare back pro­pri­etary ser­vice usage before it gets worse.

3. Does this provider’s geog­ra­phy match my cus­tomers’ com­pli­ance needs? If you sell into Euro­pean B2B, you need data res­i­den­cy in the EU. If you sell into health­care or gov­ern­ment, you need HIPAA/Fe­dRAMP-com­pli­ant regions. Not every provider has every region; not every ser­vice is avail­able in every region. Map your top 5 tar­get cus­tomer seg­ments to their com­pli­ance require­ments, then fil­ter the IaaS list before you decide.

4. What’s my engi­neer­ing team actu­al­ly good at? A 6‑person eng team choos­ing Google Cloud because Kuber­netes is “the mod­ern way” usu­al­ly ends up with a worse out­come than the same team choos­ing AWS because half of them already know it. The right IaaS is the one your team will oper­ate well — not the one with the best tech on paper. This is where most PE-adja­cent SaaS com­pa­nies get the deci­sion right: they default to what­ev­er their team ran at pre­vi­ous com­pa­nies and ship. Ear­ly-stage star­tups get it wrong when they chase what­ev­er the lat­est engi­neer­ing blog post cel­e­brates.

The 7 Leading Examples of IaaS Companies

Here’s the ven­dor-by-ven­dor break­down. For each, I cov­er what it’s best for, pric­ing at the CEO Don­ny scale ($5M ARR, ~$40M ARR tra­jec­to­ry), the CEO-lev­el trade­off, and when to recon­sid­er.

1. Amazon Web Services (AWS)

What it’s best for: Default choice for SaaS com­pa­nies at any stage. Deep­est ser­vice cat­a­log, largest tal­ent pool, broad­est geo­graph­ic cov­er­age (36+ Regions, 114+ Avail­abil­i­ty Zones as of 2026). If you hire an expe­ri­enced engi­neer in a sec­ondary U.S. city, they almost cer­tain­ly know AWS.

Pric­ing real­i­ty at $5M ARR: A typ­i­cal mul­ti-ten­ant B2B SaaS work­load on AWS runs $8K–$15K per month. That’s rough­ly 2–3.6% of ARR on infra­struc­ture — inside the healthy range. The com­mon fail­ure mode is not AWS itself, it’s the NAT gate­way, inter-AZ data trans­fer, and over-pro­vi­sioned reserved instances that qui­et­ly push the bill 40% high­er than it needs to be. Cost-engi­neer your set­up once per year and you’ll stay inside the band.

CEO-lev­el trade­off: AWS gives you the deep­est lock-in of any IaaS. DynamoDB, Lamb­da, SQS, S3 event trig­gers, and IAM poli­cies are all pro­pri­etary. The trade­off is that you get veloc­i­ty today in exchange for switch­ing cost lat­er. Most com­pa­nies accept that trade­off through $30M ARR; beyond that, strate­gic acquir­ers start ask­ing about mul­ti-cloud readi­ness as a risk-reduc­tion sig­nal.

When to recon­sid­er: When your AWS bill cross­es $50K/month and you haven’t done a cost-engi­neer­ing audit in the last 12 months. Also when you’re rais­ing an insti­tu­tion­al round and the VC dili­gence team push­es on sin­gle-provider con­cen­tra­tion risk. See cloud ser­vice providers for a deep­er treat­ment of provider selec­tion beyond the IaaS lay­er.

Key fea­tures to know: EC2 com­pute, S3 stor­age, RDS man­aged data­bas­es, VPC net­work­ing, IAM access con­trol, Cloud­Watch mon­i­tor­ing, Reserved Instances and Sav­ings Plans for 30–60% dis­counts on com­mit­ted-use pric­ing.

2. Microsoft Azure

What it’s best for: SaaS com­pa­nies sell­ing into the enter­prise, espe­cial­ly if your buy­ers are already deep in the Microsoft ecosys­tem (Office 365, Dynam­ics, Active Direc­to­ry). The Azure AD inte­gra­tion sto­ry is the sin­gle biggest rea­son enter­prise SaaS com­pa­nies pick Azure — it removes a pro­cure­ment objec­tion.

Pric­ing real­i­ty at $5M ARR: $7K–$13K per month for a com­pa­ra­ble work­load. Azure’s com­mit­ted-use dis­counts (reserved instances + sav­ings plans) tend to be 3–5 points bet­ter than AWS’s at the high end. If you’re nego­ti­at­ing an Enter­prise Agree­ment at $15M+ ARR, Azure dis­counts are often the most aggres­sive of the big three.

CEO-lev­el trade­off: Azure’s devel­op­er expe­ri­ence is less pol­ished than AWS’s, which means your team will move slow­er for the first 3–6 months. In exchange, you get eas­i­er pro­cure­ment con­ver­sa­tions with enter­prise buy­ers, bet­ter SSO/identity inte­gra­tion, and typ­i­cal­ly bet­ter com­mer­cial terms at scale. If your GTM depends on land­ing For­tune 1000 cus­tomers, Azure often pays for itself in short­ened sales cycles.

When to recon­sid­er: If your cus­tomer base is SMB-dom­i­nat­ed and none of them care about Microsoft inte­gra­tion, you’re pay­ing the Azure com­plex­i­ty tax with­out get­ting the enter­prise deal ben­e­fit. Switch to AWS or GCP.

Key fea­tures to know: Vir­tu­al Machines, Blob Stor­age, Azure SQL, Azure AD (now Entra ID), Azure Kuber­netes Ser­vice, Log­ic Apps, Azure Mon­i­tor, Hybrid Ben­e­fit for Win­dows Serv­er licens­es.

3. Google Cloud Platform (GCP)

What it’s best for: Data-heavy SaaS work­loads and AI-first prod­ucts. Big­Query is the sin­gle best cloud data ware­house, full stop. If your pro­duc­t’s dif­fer­en­ti­a­tion involves ana­lyt­ics, ML, or real-time data pro­cess­ing, GCP earns its place on the short list.

Pric­ing real­i­ty at $5M ARR: $6K–$12K per month for gen­er­al work­loads; often 15–25% cheap­er than AWS for equiv­a­lent com­pute. GCP’s sus­tained-use dis­counts apply auto­mat­i­cal­ly — you don’t need to nego­ti­ate or pre-com­mit like you do with AWS Reserved Instances. The $300 free cred­it for new cus­tomers is real but irrel­e­vant at your scale.

CEO-lev­el trade­off: Small­er tal­ent pool than AWS. Hir­ing a senior GCP engi­neer in a sec­ondary city is hard­er and more expen­sive than hir­ing the equiv­a­lent AWS per­son. The com­pute is cheap­er; the hir­ing is not. Net-net, GCP tends to be the right choice for data-dif­fer­en­ti­at­ed prod­ucts and the wrong choice for run-of-the-mill CRUD appli­ca­tions.

When to recon­sid­er: If your prod­uct is a stan­dard B2B SaaS CRUD app (forms, dash­boards, work­flows) with no data dif­fer­en­ti­a­tion, you’re pay­ing in hir­ing dif­fi­cul­ty for com­pute sav­ings you don’t need. AWS is usu­al­ly the bet­ter answer.

Key fea­tures to know: Com­pute Engine, Cloud Stor­age, Big­Query, Kuber­netes Engine (GKE — Google invent­ed Kuber­netes, and it shows), Cloud Run, Ver­tex AI, sus­tained-use auto­mat­ic dis­counts.

4. Alibaba Cloud

What it’s best for: SaaS com­pa­nies with mean­ing­ful rev­enue from Chi­nese, South­east Asian, or Mid­dle East­ern enter­prise cus­tomers. Aliba­ba has the deep­est cov­er­age in APAC by a wide mar­gin.

Pric­ing real­i­ty at $5M ARR: $4K–$9K per month equiv­a­lent, often 30–40% cheap­er than the U.S. big three at com­pa­ra­ble capac­i­ty. Cost advan­tage is real but comes with oper­a­tional com­plex­i­ty — Eng­lish doc­u­men­ta­tion qual­i­ty has improved but still lags AWS/Azure/GCP, and sup­port is time-zone con­strained for U.S. cus­tomers.

CEO-lev­el trade­off: If a dou­ble-dig­it per­cent­age of your rev­enue comes from APAC cus­tomers, Aliba­ba can cut infra­struc­ture costs 30%+ for that por­tion of your traf­fic. If your cus­tomer base is 95% U.S. and Europe, the com­plex­i­ty isn’t worth the sav­ings. Most $5M–$15M ARR SaaS com­pa­nies don’t have enough APAC rev­enue to jus­ti­fy Aliba­ba as a pri­ma­ry or sec­ondary provider.

When to recon­sid­er: If you don’t serve APAC, don’t use Aliba­ba. If you do serve APAC and it’s more than 20% of rev­enue, run the cost com­par­i­son — the answer is often “yes” but the oper­a­tional over­head is real.

Key fea­tures to know: Elas­tic Com­pute Ser­vice (ECS), Object Stor­age Ser­vice (OSS), ApsaraDB, Cloud­Mon­i­tor, Auto Scal­ing, Hybrid Back­up Recov­ery, exten­sive APAC region cov­er­age (200+ countries/territories served).

5. IBM Cloud

What it’s best for: Reg­u­lat­ed indus­tries — bank­ing, insur­ance, health­care — where IBM’s enter­prise sales motion and com­pli­ance cer­ti­fi­ca­tions mat­ter more than devel­op­er expe­ri­ence. Wat­son AI is a cred­i­ble dif­fer­en­tia­tor for spe­cif­ic AI work­loads, par­tic­u­lar­ly nat­ur­al lan­guage pro­cess­ing in reg­u­lat­ed con­texts.

Pric­ing real­i­ty at $5M ARR: $8K–$16K per month. Not the cheap­est; rarely the most expen­sive. Pric­ing is heav­i­ly nego­tiable if you have an IBM enter­prise rela­tion­ship already — their sales team will cut deals to keep you on-plat­form. “Always Free” ser­vices plus $200 cred­it for 30 days is avail­able for new cus­tomers, though not rel­e­vant at scale.

CEO-lev­el trade­off: IBM Cloud’s small­er mar­ket share means a small­er tal­ent pool and slow­er ser­vice-cat­a­log growth. But if your cus­tomer base is For­tune 500 reg­u­lat­ed indus­tries, IBM gives you a sales-side cred­i­bil­i­ty boost and some­times a pro­cure­ment short­cut. If nei­ther applies, IBM is rarely the right pri­ma­ry IaaS.

When to recon­sid­er: Out­side of reg­u­lat­ed enter­prise con­texts, IBM Cloud is usu­al­ly not the right pri­ma­ry provider for a grow­ing B2B SaaS com­pa­ny. If you land­ed here because of a lega­cy IBM rela­tion­ship, bud­get 6–12 months to migrate to AWS or Azure.

Key fea­tures to know: Vir­tu­al Servers (bare met­al and VM), Cloud Object Stor­age, Cloud Data­bas­es, Wat­son AI ser­vices, Kuber­netes Ser­vice, strong hybrid and mul­ti-cloud sup­port (inte­grates with AWS, Azure, GCP), built-in encryp­tion and key man­age­ment.

6. Oracle Cloud Infrastructure (OCI)

What it’s best for: SaaS com­pa­nies run­ning Ora­cle data­bas­es or Java-heavy enter­prise work­loads. Ora­cle has aggres­sive­ly priced OCI to com­pete with AWS, and the “Always Free” tier is gen­uine­ly use­ful for dev envi­ron­ments. Ora­cle’s Autonomous Data­base is the most com­pelling pro­pri­etary ser­vice OCI offers.

Pric­ing real­i­ty at $5M ARR: $6K–$11K per month; often 20–30% cheap­er than AWS for equiv­a­lent com­pute, espe­cial­ly if you have an Ora­cle license port­fo­lio already. Ora­cle’s Uni­ver­sal Cred­its mod­el (com­mit once, spend across ser­vices) is gen­uine­ly more flex­i­ble than AWS’s Sav­ings Plans.

CEO-lev­el trade­off: OCI’s rapid­ly matur­ing ser­vice cat­a­log still lags AWS by sev­er­al years in some cat­e­gories. The cost sav­ings are real; the ser­vice breadth is not. If your appli­ca­tion is Ora­cle-data­base-heavy, OCI is often the cheap­est and oper­a­tional­ly eas­i­est option. If it isn’t, you’re giv­ing up capa­bil­i­ty for cost.

When to recon­sid­er: If your appli­ca­tion has no Ora­cle foot­print and your team has no Ora­cle expe­ri­ence, OCI is rarely the right pri­ma­ry provider. The cost sav­ings won’t off­set the ser­vice-cat­a­log gap for a typ­i­cal B2B SaaS work­load.

Key fea­tures to know: Bare met­al com­pute, Autonomous Data­base, Object Stor­age, Vir­tu­al Cloud Net­work (VCN), Kuber­netes Engine, “Always Free” tier, Uni­ver­sal Cred­its for com­mit­ted-use pric­ing.

7. DigitalOcean

What it’s best for: Ear­ly-stage SaaS com­pa­nies under $5M ARR that want sim­ple, cheap, pre­dictable infra­struc­ture. Dig­i­talO­cean’s pric­ing is rad­i­cal­ly sim­pler than AWS’s (flat per-hour rates, no egre­gious egress fees at small scale, no PhD in AWS billing required).

Pric­ing real­i­ty at $5M ARR: $3K–$7K per month for a com­pa­ra­ble work­load — rough­ly half what you’d pay on AWS. The catch: the ser­vice cat­a­log is much small­er. You’ll have to build, buy, or work around ser­vices AWS pro­vides out of the box (man­aged Redis, advanced net­work­ing, sophis­ti­cat­ed IAM, deep mon­i­tor­ing).

CEO-lev­el trade­off: Dig­i­talO­cean trades breadth for sim­plic­i­ty and cost. Per­fect for a 3–10 per­son engi­neer­ing team that wants to ship the prod­uct with­out becom­ing infra­struc­ture spe­cial­ists. Less per­fect for a 30-per­son team that needs enter­prise-grade net­work­ing, com­pli­ance, and man­aged ser­vices.

When to recon­sid­er: When you land your first enter­prise cus­tomer that demands SOC 2, HIPAA, or FedRAMP. Dig­i­talO­cean has SOC 2 but it’s oper­a­tional­ly thin­ner than AWS’s or Azure’s com­pli­ance tool­ing. Also when you cross $10M ARR and your engi­neer­ing team spends mean­ing­ful time work­ing around miss­ing ser­vices.

Key fea­tures to know: Droplets (sim­ple VMs), Man­aged Data­bas­es (Post­gres, MySQL, Redis), Spaces (S3-com­pat­i­ble object stor­age), Kuber­netes, App Plat­form (PaaS lay­er on top of IaaS), clear flat-rate pric­ing.

Hon­or­able men­tions worth know­ing about:

• Lin­ode (acquired by Aka­mai in 2022) — very sim­i­lar to Dig­i­talO­cean in pric­ing and phi­los­o­phy, now with Aka­mai’s glob­al edge net­work bolt­ed on. Strong option for con­tent-deliv­ery-heavy work­loads.
• Het­zn­er Cloud — Euro­pean provider with the most aggres­sive pric­ing in the mar­ket. Best-in-class price-per-vCPU but more man­u­al oper­a­tions. Good for cost-sen­si­tive Euro­pean SaaS com­pa­nies under $3M ARR.

Picking the Right IaaS for Your Stage

Here’s how the sev­en exam­ples of IaaS com­pa­nies map to typ­i­cal SaaS stages. This isn’t a hard rule — excep­tions are every­where — but it’s a good start­ing point.

ARR Stage	Pri­ma­ry Rec­om­men­da­tion	Why	When to Move
Under $1M ARR	Dig­i­talO­cean, Het­zn­er, or a PaaS (Heroku, Ver­cel)	Opti­mize for engi­neer­ing time, not cloud cost. You’ll lose more mon­ey to slow ship­ping than to over­pay­ing for infra­struc­ture.	Move when your bill hits ~$3K–$5K/month or you need com­pli­ance cer­ti­fi­ca­tions the PaaS can’t offer.
$1M–$5M ARR	AWS or Dig­i­talO­cean, depend­ing on team	AWS if the team knows it and you need the ser­vice cat­a­log. Dig­i­talO­cean if the team is small and ship­ping speed mat­ters more than fea­tures.	Move off Dig­i­talO­cean when a seri­ous cus­tomer asks for SOC 2 con­trols DO can’t clean­ly sup­port, or when the cost delta vs. AWS inverts (above ~$10M ARR it usu­al­ly does).
$5M–$15M ARR	AWS (default), Azure (if enter­prise GTM), GCP (if data/AI prod­uct)	This is the stage where IaaS choice com­pounds. Get the pri­ma­ry provider right and stop sec­ond-guess­ing.	Move only if you hit the gross mar­gin ceil­ing or you can’t hire tal­ent for your cur­rent stack. Both are rare before $15M ARR.
$15M–$30M ARR	Stay with pri­ma­ry + start mul­ti-cloud readi­ness	Cloud con­cen­tra­tion becomes a dili­gence item at this range. You don’t need to run mul­ti-cloud, but you should have a believ­able migra­tion plan.	When strate­gic buy­ers start show­ing up. The con­ver­sa­tion turns into “how risky is this depen­den­cy?”
$30M+ ARR	Mul­ti-cloud or nego­ti­at­ed Enter­prise Agree­ment	At this scale, your IaaS bill is a real line item on the exit mod­el. Nego­ti­ate aggres­sive­ly. AWS and Azure EAs at $1M+/year of com­mit­ted spend typ­i­cal­ly come with 20–35% dis­counts.	Ongo­ing — this is a pro­cure­ment func­tion, not a one-time deci­sion.

For a deep­er treat­ment of how cloud deci­sions scale with rev­enue, see how to scale a SaaS busi­ness.

How IaaS Choice Shows Up in Your Exit

This is the sec­tion every founder wish­es some­one had shown them ear­li­er.

When you sell a SaaS com­pa­ny, the buy­er’s mod­el starts with rev­enue and applies a mul­ti­ple. Rev­enue mul­ti­ples range from 2x for weak busi­ness­es to 15x+ for best-in-class. The mul­ti­ple is dri­ven by six levers: rev­enue nature, growth rate, mar­gins, risk, com­pet­i­tive advan­tage dura­bil­i­ty, and mar­ket size cap. IaaS choice direct­ly affects two of them — mar­gins and risk.

Mar­gins. A com­pa­ny run­ning 82% gross mar­gin gets a mate­ri­al­ly high­er mul­ti­ple than the same com­pa­ny at 68% gross mar­gin. The line between “soft­ware busi­ness” (high mul­ti­ple) and “ser­vices busi­ness” (low mul­ti­ple) is usu­al­ly drawn at 70% gross mar­gin. Infra­struc­ture is the biggest lever you con­trol; a 500 bps (5 per­cent­age point) improve­ment in gross mar­gin at $10M ARR can add $10M–$30M to enter­prise val­ue.

Con­crete math. Imag­ine two $10M ARR SaaS com­pa­nies grow­ing at 40%. Com­pa­ny A runs 80% gross mar­gin. Com­pa­ny B runs 68% gross mar­gin because its AWS set­up is poor­ly engi­neered and man­age­ment nev­er pri­or­i­tized cost opti­miza­tion. At exit, Com­pa­ny A trades at 8x rev­enue ($80M), Com­pa­ny B at 5x rev­enue ($50M). Same rev­enue, same growth rate, same prod­uct — $30M of enter­prise val­ue dif­fer­ence, dri­ven almost entire­ly by infra­struc­ture dis­ci­pline.

Risk. Sin­gle-provider con­cen­tra­tion is a risk fac­tor. Acquir­ers ask about it. If 100% of your pro­duc­tion work­load runs on one AWS account in one region with no believ­able migra­tion plan, the dili­gence team applies a risk dis­count — typ­i­cal­ly 5–15% off the mul­ti­ple, depend­ing on how crit­i­cal the depen­den­cy is to the prod­uct. This is why the $15M–$30M ARR stage is the right time to start mul­ti-region (at min­i­mum) and prefer­ably mul­ti-cloud-readi­ness work. For the broad­er fram­ing, see SaaS exit strat­e­gy.

The P&L tim­ing angle. Buy­ers typ­i­cal­ly val­ue the com­pa­ny on the trail­ing 12 months of finan­cials, with that win­dow start­ing ~6 months before close. If you’re going to exit in 18 months, the infra­struc­ture cost opti­miza­tion you start today hits the trail­ing 12 months they’ll actu­al­ly val­ue. Inverse: sav­ings you real­ize the month after close don’t mat­ter for your out­come. Time the work.

Common Mistakes When Choosing IaaS Examples

These are the five mis­takes I see most often — every one of them pre­ventable.

1. Let­ting the CTO pick alone. The CTO will pick the stack that’s eas­i­est to build on. The CEO’s job is to make sure the choice is also the cheap­est to oper­ate and the eas­i­est to exit. If you’re not in the room for this deci­sion, you’re trust­ing a stack choice that will show up in your P&L for the next five years.

2. Ignor­ing egress costs. AWS charges ~$0.09/GB to move data out of a region. At 50 TB/month of egress (not unusu­al for a mid-sized API busi­ness), that’s $4,500/month — $54K/year — that nobody bud­gets for. Mul­ti­ply by 3x if your archi­tec­ture has inter-AZ chat­ter. Read the bill line by line at least once per quar­ter.

3. Over-com­mit­ting on reserved instances. Reserved Instances and Sav­ings Plans give you 30–60% off if you com­mit to a year or three years of spend. Teams rou­tine­ly over-com­mit at the peak of a growth spurt, then car­ry unused capac­i­ty when growth slows. Com­mit for no more than 70% of your run-rate usage; leave the buffer for spot/on-demand.

4. Treat­ing “free tier” as free for­ev­er. Every provider’s free tier has a cliff. GCP’s $300 cred­it expires. AWS Free Tier drops after 12 months. Bud­get like the free tier does­n’t exist; treat it as a one-time bonus if it hap­pens to apply.

5. Not seg­ment­ing cloud cost. Com­pa­ny-wide infra­struc­ture COGS as a per­cent of rev­enue is a use­less num­ber. Seg­ment it by cus­tomer tier, prod­uct line, and region. Your enter­prise cus­tomers usu­al­ly cost 3–5x more to serve than your SMB cus­tomers; if your pric­ing does­n’t reflect that, your enter­prise seg­men­t’s unit eco­nom­ics are worse than the blend­ed num­ber sug­gests. Seg­ment every­thing — cloud cost includ­ed.

Frequently Asked Questions About Examples of IaaS

What’s the cheap­est IaaS for a SaaS start­up? For star­tups under $1M ARR, Het­zn­er Cloud is the cheap­est legit­i­mate option, fol­lowed by Dig­i­talO­cean and Lin­ode. Below ~$3K/month of spend, the engi­neer­ing time cost of opti­miz­ing a cheap­er provider out­weighs the sav­ings — just pick whichev­er your team already knows.

What’s the best IaaS for an AI-first SaaS prod­uct? GCP for most AI work­loads. Ver­tex AI and Big­Query are gen­uine­ly cat­e­go­ry-lead­ing. AWS is a cred­i­ble sec­ond choice (Bedrock, Sage­Mak­er). Azure Ope­nAI is the right call if your pri­ma­ry AI depen­den­cy is on Ope­nAI mod­els and you need enter­prise-grade SLAs around them.

When should I move off AWS? Only when you have a con­crete pain point that AWS can’t solve and an alter­na­tive provider can. “We should con­sid­er leav­ing AWS” is not a migra­tion trig­ger. “Our Euro­pean cus­tomers need EU data res­i­den­cy and our AWS archi­tec­ture makes that expen­sive” is. See cloud ser­vice providers for the full deci­sion frame­work.

Is IaaS the same as cloud com­put­ing? No. Cloud com­put­ing is the umbrel­la term for all on-demand com­pute ser­vices, includ­ing IaaS, PaaS, SaaS, and FaaS. IaaS is one lay­er of the cloud stack — the infra­struc­ture lay­er.

Do I need mul­ti-cloud? Prob­a­bly not until $30M+ ARR. Run­ning mul­ti-cloud ear­ly is a dis­trac­tion that slows prod­uct veloc­i­ty. But you should be mul­ti-cloud ready by $15M ARR — mean­ing your stack uses enough portable ser­vices (Post­gres over DynamoDB, Kuber­netes over ECS, S3-com­pat­i­ble stor­age) that a 6‑month migra­tion is fea­si­ble if you need­ed it.

How do I know if my IaaS bill is too high? Bench­mark infra­struc­ture COGS as a per­cent of ARR. For most B2B SaaS at $5M–$15M ARR, healthy is 3–7% of ARR on infra­struc­ture. Above 10% is a sign of poor cost engi­neer­ing, over-pro­vi­sion­ing, or a bad­ly archi­tect­ed appli­ca­tion. Below 2% is unusu­al and some­times a sign you’re under-invest­ing in reli­a­bil­i­ty.

The Infrastructure Question CEOs Actually Care About

When you strip away the ven­dor mar­ket­ing and the engi­neer­ing team’s pref­er­ences, the IaaS deci­sion comes down to three num­bers:

• What per­cent of ARR does this cost me? (Tar­get: under 7% at $5M–$15M ARR)
• How long would it take to leave? (Tar­get: under 6 months of eng time at $10M ARR, under 12 months at $25M ARR)
• How does this affect my mul­ti­ple when I sell? (Tar­get: in the top quar­tile on gross mar­gin, below the thresh­old on con­cen­tra­tion risk)

Opti­mize for those three answers. The ven­dor you pick mat­ters less than the dis­ci­pline you apply once you’ve picked one. AWS, Azure, and GCP all pro­duce excel­lent out­comes for SaaS com­pa­nies that man­age them well. All three pro­duce poor out­comes for com­pa­nies that don’t. The dif­fer­ence between a 68% gross mar­gin SaaS com­pa­ny and an 82% gross mar­gin SaaS com­pa­ny is almost nev­er the provider — it’s the man­age­ment.

If you’re seri­ous about this, put infra­struc­ture cost on your month­ly P&L review, hire or con­tract a cloud cost spe­cial­ist once your bill hits $30K/month, and run a for­mal cost-engi­neer­ing audit every 12 months. That dis­ci­pline is worth more at exit than any ven­dor you could pick.

The exam­ples of IaaS com­pa­nies in this arti­cle — AWS, Azure, GCP, Aliba­ba, IBM, Ora­cle, and Dig­i­talO­cean — are the tools. The eco­nom­ics are the out­come. Spend your time on the eco­nom­ics.