Google’s Custom Machine Types, which have been in test mode since November, let the buyer configure computing power with different amounts of processing power and memory to best suit the application at hand.

To back up a tad, public cloud providers, as exemplified by AWS (AMZN), Microsoft (MSFT) Azure and Google (GOOG) Cloud Platform, pool lots and lots of servers and then rent out that shared computing capacity by the hour or, in Google’s and Microsoft’s cases, by the minute.

And yet that’s apparently not granular enough for some customers who don’t want to pay for any unused capacity at all, ever, even if it’s by the minute. So these new machine types let users build the type of computing they need, adding or subtracting memory to the compute node as needed.

Basically what Google has done is slice and dice basic computing capacity and the associated memory needed to perform tasks in ways that best suit their workloads. So a developer can choose to run up to 32 virtual central processing units (vCPU is Google’s term for a basic computing unit), and allocate memory from 6.5 gigabytes per vCPU on up.

That means, as Google said last fall, that customers can move a particular job to another, more appropriate vCPU configuration if the job changes.

A theoretical analogy would be if your laptop could fire up a newer, more powerful Intel (INTC) processor just for compute-intensive tasks and more RAM when memory is required, but then take those resources down (and stop paying for them) when they’re not needed.

GCP users can also opt to run their vCPUs for a month at a time. If their job runs at full, 100% capacity, Google can cut their bill significantly.

MORE: Google is serious about the enterprise, says cloud chief

In a competitive context, these new computing units are an attempt by Google to woo customers from market leader AWS, which offers a huge array of computing instances—so huge that things can get confusing fast. All of these cloud providers tread a fine line between offering a ton of choices and trying to keep things simple for harried users. It’s not an easy feat.

Google’s issue is that, while developers love its whizzy cloud-based tools like its BigTable NoSQL database and BigQuery analytics, its cloud platform per se isn’t really seen as an enterprise cloud choice.

AWS used to be viewed in much the same way but Amazon is well into year five or so of a huge enterprise sales push. Microsoft, on the other hand, already had tons of big business customer relationships in place by by virtue of its decades-old Windows, Office, SQL Server, and Exchange Server businesses.

For more on Google, watch:

Many big companies still view Google slightly askance wondering if the search and ad giant is serious about selling them cloud. Google, after all, is known for launching and then nixing products, the most recent of which is the Google Search Appliance, the company’s first product focusing on bigger business customers.

Lydia Leong, a Gartner VP and distinguished cloud analyst, recently told Fortune it’s not that Google doesn’t understand big customers. It has, after all, hired considerable tech talent from Oracle (ORCL) and other big-business focused software providers.

“It’s more that Google to date hasn’t had the willpower to do what they need to do to appeal to these big companies,” Leong said.

Some Google partisans even acknowledge this was the case, but counter that Google’s cloud thinking started to shift two years ago when it hired former Red Hat (RHT) chief technology officer Brian Stevens as vice president of cloud platforms.

Get Data Sheet, Fortune’s technology newsletter.

That message was reinforced a year later when it named VMware (VMW) co-founder Diane Greene to head up the Google Enterprise business, which includes GCP.

Both Red Hat and VMware focus exclusively on big-business customers.

Expect Google to talk up its cloud as a good enterprise option next month at its Google Next conference in San Francisco.