Those companies assemble and sell massive arrays of servers, storage, and networking to businesses—most of which don’t want to build more of their own data centers. Towards that end, those three cloud superpowers alone spent roughly $31 billion last year to extend their data center capacity around the world, according to the Wall Street Journal, which tabulated that total from corporate filings.
But Mark Hurd, Oracle’s co-chief executive, would beg to differ. In his view, there are data centers and then there are data centers. And Oracle’s data centers, he said, can be more efficient because they run Oracle hardware and supercharged databases.
“We try not to get into this capital expenditure discussion. It’s an interesting thesis that whoever has the most capex wins,” Hurd said in response to a question from Fortune at a Boston event on Tuesday. “If I have two-times faster computers, I don’t need as many data centers. If I can speed up the database, maybe I need one fourth as may data centers. I can go on and on about how tech drives this.”
Get Data Sheet, Fortune’s technology newsletter.
“Our core advantage is what we’ve said all along, which is that it’s about the intellectual property and the software, not about who’s got the most real estate,” Hurd added. “We have spent billions over the past year, but in isolation, that’s a discrete argument that I find interesting, but not fascinating.”
Following up via email, Hurd said: “This isn’t a battle of capex. This is about R&D, about technology, software, innovation and IP; and then the capex to make it work.”
Oracle has said it runs its data centers on Oracle Exadata servers, which are turbocharged machines that differ fundamentally from the bare-bones servers that other public cloud providers deploy by the hundreds of thousands in what is called a scale-out model. The idea is that when a server or two among the thousands fail—as they will—the jobs get routed to still-working machines. It’s about designing applications that are easily redeployed.
Oracle is banking more on what techies call a “scale-up” model in which fewer, but very powerful computers—in Exadata’s case each with its own integrated networking and storage—take on big workloads.
Oracle execs, including executive chairman Larry Ellison, have argued that Oracle’s big machines can actually work cheaper and more efficiently than the other public cloud configurations. Many industry analysts have their doubts on that, maintaining Oracle must spend much more to catchup with Amazon. Toward that end, in January, Oracle announced plans to add three new data center farms within six months and more to come.
There are those who think that Fortune 500 companies relying on Oracle databases and financial applications give Oracle an advantage because they are loath to move those workloads to another cloud provider—despite AWS wooing them with promises of easy migrations other perks.
In late March, AWS chief executive Andy Jassy claimed the company had converted 22,000 databases from other vendors to its own database services. AWS (amzn) does not break out which databases those customers had been using.
Hurd took up that point as well: “How much database market will Oracle lose to [Amazon] Aurora? My guess is close to zero.” (Aurora is one of several database options that AWS offers.)
“The third largest database in the world is IBM DB2, and it’s been going out of business for 20 years,” Hurd said in a characterization that IBM (ibm) would dispute. “If it was so easy to replace databases, DB2 market share would be zero.”
That is because most databases—which companies rely on as the basis for core accounting and financial operations—run custom programming, which is hard to move.
Still, in Amazon, Oracle—as well as IBM, Microsoft, and virtually every legacy information technology provider—faces a huge challenge. AWS is on track to log more than $14 billion in revenue this year.
Note: (April 12, 2017 12:55 p.m.) This story was updated to add an additional quote from Oracle’s Mark Hurd.