TL;DR
SaaS and PaaS are just two entries in a genuinely large family. The umbrella term for all of it is XaaS — “Anything as a Service” and by one count there are over 50 distinct “-aaS” categories in active use. Most of them boil down to the same underlying idea: someone else runs the infrastructure, and you pay for the outcome instead of owning the machinery. Here’s the landscape that actually matters, from the three foundational models to the ones reshaping how software gets built in 2026.
The Original Three, and the One Sentence That Explains Them
Before the acronym explosion, there were three formally defined models, and the cleanest way to tell them apart is this: IaaS gives you a machine, PaaS gives you a runtime, SaaS gives you an outcome.
IaaS (Infrastructure as a Service) — you get raw computing resources: servers, storage, networking, delivered over the internet instead of racked in a physical data center. You still manage the operating system and everything above it. AWS EC2, Microsoft Azure, and Google Compute Engine are the standard examples. Best fit: teams that need real control and flexibility, or specialized/regulated workloads that don’t fit a pre-built platform.
PaaS (Platform as a Service) — infrastructure plus a pre-built environment for developing, testing, and deploying applications, without managing the underlying servers yourself. Heroku, Google App Engine, and AWS Elastic Beanstalk are classic examples. Best fit: teams that want to build and ship fast without owning infrastructure decisions.
SaaS (Software as a Service) — a complete, ready-to-use application delivered over the internet. No setup, no infrastructure decisions, just log in and use it. Salesforce, Dropbox, HubSpot. SaaS remains the dominant model by far the average organization now manages around 305 separate SaaS apps, and large enterprises run closer to 700.
The Two Tiers That Didn’t Exist on the Old Textbook Slide
Most people learned “IaaS vs. PaaS vs. SaaS” as a fixed, three-tier model. That model was accurate for explaining cloud computing a decade ago. It’s incomplete now — two more tiers have become genuinely mainstream since.
FaaS (Functions as a Service) — the most abstract tier below PaaS. Instead of deploying a full application, you deploy a single function, and the platform handles scaling, request routing, and execution automatically you pay per invocation, down to the millisecond. AWS Lambda (the original, from 2014), Cloudflare Workers, and Vercel Functions are the standard examples. It’s especially strong for event-driven work webhooks, scheduled jobs, image processing though it’s a worse fit for long-running processes that need a persistent connection.
Agentic-PaaS — a genuinely new category that didn’t exist even a few years ago: platforms built specifically to expose themselves to AI agents as first-class users, not just human developers. As AI agents increasingly write and deploy code autonomously, platforms are being redesigned around that reality rather than retrofitted for it.
The Specialized “-aaS” Categories Worth Actually Knowing
Beyond the core tiers, dozens of narrower categories have emerged, each solving one specific problem as a subscription rather than something you build in-house. A few of the ones that show up most in real infrastructure decisions:
- DaaS (Desktop as a Service) — full virtual desktops delivered over the internet, useful for remote teams or temporary/contractor access without provisioning physical hardware.
- DBaaS (Database as a Service) — a managed database where the provider handles backups, scaling, and maintenance, so a team doesn’t need dedicated database administration.
- SECaaS (Security as a Service) — outsourced security functions antivirus, intrusion detection, identity management delivered as a subscription rather than run in-house. This market alone was projected to grow from $12.4 billion in 2021 to $23.8 billion by 2026.
- UCaaS (Unified Communications as a Service) — voice, video, messaging, and conferencing bundled into one cloud-delivered platform rather than separate on-premises systems.
- DRaaS (Disaster Recovery as a Service) — a provider handles backup and recovery infrastructure, so a company doesn’t have to build and maintain a duplicate emergency system itself.
- AIaaS (AI as a Service) — access to AI models and capabilities via API rather than training or hosting your own this is functionally what most companies building AI products are actually doing, whether or not they use the label.
Where “Results as a Service” Actually Fits
This is a newer, less formally standardized framing, and it’s worth being precise about what it actually means: rather than paying for access to a tool (the SaaS model) or infrastructure (IaaS/PaaS), you pay for a completed outcome. The provider handles everything required to produce that result, and pricing is tied to the result itself rather than to usage or seats.
It’s less a single formal category on the NIST-style list and more a direction several categories are drifting toward a natural extension of AIaaS in particular, where an AI agent doesn’t just provide a tool to use, it actually completes the task and you pay for the completed work. Think of it as the logical endpoint of the “IaaS gives you a machine, PaaS gives you a runtime, SaaS gives you an outcome” progression pushed one step further, where even the use of the software is abstracted away and only the result remains.
How to Actually Pick, Instead of Chasing the Newest Acronym
A simple way to sort through this instead of trying to memorize 50 categories:
- Need raw control and flexibility, or have a regulated/specialized workload? IaaS.
- Want to build and ship an application without managing servers? PaaS.
- Need a complete, ready-to-use tool for your team right now? SaaS.
- Have small, event-driven tasks that don’t need a persistent server running? FaaS.
- Building something AI agents will interact with directly, not just humans? Look specifically for agentic-PaaS support.
- Want a specific narrow problem (security, backups, communications) handled entirely by someone else? Check whether a dedicated “-aaS” category already exists for it before building it in-house.
Know Your Layer Before You Pay for the Next One
The acronym pile can look overwhelming, but nearly all of it reduces to the same underlying trade: less control in exchange for less operational burden, at increasing levels of abstraction machine, runtime, application, function, and now, increasingly, outcome. Most real infrastructure decisions in 2026 aren’t “pick one model” they’re a genuine mix: SaaS for the tools your team uses daily, PaaS or FaaS for what you’re actually building, and IaaS reserved for the specific workloads that need full control. Know which layer you actually need before you get talked into paying for the layer above it.
Related Buzz: We also covered [The Metric Investors Actually Look at First]