Most companies making a move into cloud computing start out wanting to use multiple cloud providers, but end up deciding to go with one, according to Andy Jassy, the Amazon long-timer who was promoted to CEO of Amazon Web Services in April.
Clearly he has a dog in this fight. AWS is the biggest and most mature public cloud provider—but now competes with Microsoft Azure and Google Cloud Platform, neither of which existed 10 years ago when Amazon (AMZN) launched this new category of computing.
Amazon, like Microsoft (MSFT) and Google (GOOG) has a worldwide network of data centers stuffed with servers, storage, and networking that customers can rent as needed.
Amazon Customers Worry About Cloud Lockin
Most of the big companies I talk to about cloud deployments are leery of putting all their eggs in one cloud basket. They feel that by using more than one cloud provider, they can keep the other provider honest. Most of these companies have faced vendor lock-in before and feel abused by IT suppliers that made it difficult to move to a rival product or even integrate their technology with that of another company.
Oracle (ORCL), VMware (VMW), Microsoft are all often criticized for these practices.
Get Data Sheet, Fortune’s daily newsletter on the business of technology.
But Jassy swears AWS is different from its predecessors. The lock-in question is very understandable, he told attendees of the AWS Public Sector Summit in Washington D.C. on Tuesday. “People are looking at being locked into Oracle for the last few years so you can understand the trepidation.”
Here he was interrupted by audience laughter. “I wasn’t really trying to be funny. But vendors have used proprietary software, licensing and punitive audits to keep customers,” he noted.
Jassy said when most companies do the analysis they realize that using multiple providers means they have to standardize on the most basic cloud components. In AWS’s case that would be EC2 servers and S3 storage, but not all the workflow, messaging, and database services can layer atop those resources.
That means the company can’t reap the full benefits of AWS, Jassy said. In fact, many AWS customers do deploy just in this way. They love the cheap and flexible Amazon compute and storage, but eschew thise aforementioned higher level services because they know that will make exiting difficult.
Secondly, managing multiple sets of technology sucks up time and resources that could be better spent elsewhere, Jassy said. Like a travel company building better reservation services, for example.
And finally splitting work up between clouds means less of a discount on any one cloud. “You give up buying leverage. All the providers have volume discounts so if you split across clouds you get less leverage.”
He also maintained that Amazon’s various migration tools that move corporate data into AWS work both ways, although it’s worth noting that Amazon’s SnowBall data migration tool started out as a one-way offering that added return trip capability (data egress) months later.
For better or worse, the perception remains that AWS still makes it easy to schlep your stuff into its cloud and less easy to get it out.
With all respect to AWS, the IT powers of the past—the Oracles, Microsofts and VMwares of the world—do not worry potential cloud customers so much right now. In this category AWS is the too-big-to-fail company. And as customer-focused as AWS may be, that power is worrisome to those customers. Ironically, those same companies who cursed heavy-handed tactics by Microsoft and Oracle may now view those companies’ cloud efforts as a hedge against Amazon.
That’s one reason Pivotal, the EMC (EMC)-VMware spin-off and other companies are pitching technology that runs across multiple clouds and inside corporate data centers. That ability, if it works as advertised, may assure would-be cloud customers that they really aren’t locked into any one provider.
