Late Thursday, Amazon(AMZN) Web Services said the new M4 instances—there are five in all— build on a custom Intel (INTC) Xeon Haswell processor with 2.4 GHz clock speed that can burst to 3.0 GHz when needs demand.

To back up for those who don’t speak geek, AWS sells (or actually rents) chunks of computing power, storage, and networking bandwidth on a pay-as-you-go basis. An instance is what it calls a unit of compute power.

AWS has long been seen as more of a “side business” for its parent company Amazon.com, which is now number 29 on the Fortune 500 list. But last quarter Amazon CEO Jeff Bezos put the lie to that contention when he said AWS is a $5 billion-a-year business, and growing fast.

But back to the compute instances at hand: They come in many, many variations. Most are available in a limited-use free tier for tire kickers. On-demand instances are available by the hour. For those with predictable workloads, reserved instances are discounted off on-demand price depending on a one- or three-year commitment.

True aficionados can also bid for instances on a spot market. There are instances with tons of storage attached and there are instances for graphics-intensive workloads that need 3-D visualization, for streaming graphics etc. Got it?

But there are still more choices: Most of these Amazon Elastic Compute (or EC2) instances can run on the user’s choice of Suse Linux, Red Hat(RHT) Linux, and a couple of flavors of Windows (with or without SQL database options.) Check out the Amazon EC2 page for a complete list, but to get an idea, there are nearly 40 different instance types available from Amazon’s large US-East region outside of Washington D.C.

All of this is to say that Amazon, which started renting out compute power nine years ago, offers either a rich—or bewildering—menu of options, depending on your point of view. The sheer number of choices give third parties like Cloudyn, Cloudability, and CloudCheckr a business opportunity to advise customers on how best to deploy their Amazon jobs. Amazon is doing likewise with its own Trusted Advisor dashboard.

Ben Whaley, a San Francisco-based cloud computing consultant, sees both sides of the choice/complexity question.

Amazon, he said, has to adapt to changes in the marketplace. “They have to add more options, some of their services are much better than others. Is it confusing? Yeah, sometimes it is but the alternative is to not offer enough, which is worse,” he noted via email. Besides, he added, “all the complexity keeps me employable.”

David Mytton, CEO of London-based Server Density contrasted Amazon’s strategy with Google’s cloud play.

“AWS gives you all the options and lets you choose, whereas Google tries to make as many of the choices for you, betting they have a greater understanding of the tradeoffs. I call this operational simplicity vs flexibility,” he said.

Neither Google(GOOG) Cloud Platform or Microsoft[fortune -stock symbol=”MSFT”] Azure, Amazon’s two primary cloud competitors, come close to this number of options, although both have introduced new perks of their own. Both companies, offer compute power by the minute rather than by the hour, for example. And with Google compute instances, discounts kick in automatically when usage of a given instance exceeds a certain threshold.

Subscribe to Data Sheet, Fortune’s daily newsletter on the business of technology.