In April 2009, Oracle Corp. made what at the time looked like a bad deal. Or at least that was the conventional wisdom. It announced that it was buying Sun Microsystems Inc. for $7.4 billion, or $5.6 billion net of Sun’s cash and debt.

TL/DR, the judgment of the community on the decision by Oracle Chief Executive Larry Ellison (pictured) was: Why would a database and enterprise software company buy a declining server company and get saddled with the resultant cost drain of a hardware business?

But here’s the rub. In 2009, Oracle didn’t mortgage its future to buy Sun. Instead it bought its future. It just took about 15 years for that to become evident.

As noted, conventional wisdom at the time was that the Sun deal would drain Oracle’s bottom line. The company, formerly known as “the dot in dot com,” saw its foundational Solaris and SPARC server businesses undermined by Linux and x86; suddenly, anyone could mount a LAMP stack web server. At the time of the acquisition, Sun was losing $100 million each month. What was to like?

In the short term, the news wasn’t terrible. In the fiscal year post-acquisition, Oracle’s total GAAP revenue rose 33%, driven by software, while revenue from the hardware (server) business declined a modest 6%. But hidden in those figures was the gem of the acquisition: Oracle’s Exadata installed base surpassed 1000 globally. Presumably, Exadata revenues were likely counted under software.

The conventional wisdom was that by buying Sun, Oracle was getting into the server business. In actuality, the Sun acquisition brought Oracle the systems expertise that would be the key, not only to its growing Exadata business, but to a future cloud business that the company had not even envisioned back in 2009.

Still, a few years after the acquisition, the reviews were still largely negative. In 2016, a post on ZDnet listed Oracle and Sun as among the worst tech mergers and acquisitions. A year earlier, a post in The Register came with the headline Win Sun, lose Sun: How Larry’s bet on old-world systems hurt Oracle.

But in that same Register article was an interesting passage, pawned off at the time as a fantasy. “Join the Reg on a journey to an alternate reality where Oracle’s cloud business is the envy of all of its competitors. It holds two trillion objects and is growing faster now than at any time in its history.” Who knew?

Fast forward to now. On Sept. 10, Oracle’s shares shot up more than 36% on the strength, not so much on current revenues (which were below Wall Street’s expectations), but on its cloud infrastructure pipeline. It reported $455 billion in “remaining performance obligations” (committed sales pipeline) for cloud services, largely driven by AI; that’s over four times that reported a year ago. Ironically, those building RPOs shouldn’t have been that much of a surprise given Oracle’s designation as cloud infrastructure provider for Open AI’s Stargate project announced earlier this year. But (not so) coincidentally, on the same day Oracle reported earnings, it was announced that $300 billion of that came from Stargate.

So, what does all this have to do with Sun? As noted above, the Sun Microsystems acquisition proved to be a systems, not server, business play. Though Exadata predated the Sun acquisition, it was previously delivered on HP hardware. Now Oracle could gain control of both the hardware and software side, and soon, its tag line for Exadata was “engineered systems”: in essence, software designed for the hardware and vice versa.

That expertise was called upon when Oracle entered the cloud business. During his Oracle OpenWorld 2008 keynote, Ellison dismissed the cloud business as “gibberish.” But a few years later, Oracle was compelled to respond. Credit Ellison: He’s willing to change course when the reality on the ground changes. Starting with the in-house expertise gained from the Sun acquisition, Oracle aggressively recruited some of the best minds from AWS and Azure to build a cloud business from scratch.

And even having chosen to dive in, Ellison was not shy about ripping up the script in midstream. Realizing that its Gen 1 architecture would not differentiate Oracle Cloud Infrastructure from the Amazon Web Services, Microsoft Azure, or Google Cloud, Ellison pulled the plug on Gen 1 and ordered a rearchitecting from the bottom up. Among the highlights, for better security it separated customer code and data from Oracle control code and provided full tenant isolation. For performance and scale, it flattened the topology, and that made possible RDMA, a protocol that allows direct memory access bypassing the operating system. With Gen 2 introduced in 2018, Gen 1 was finally sunsetted five years later.

As a result, OCI has differentiated itself with its ability to scale AI workloads. Virtually every hyperscaler runs Nvidia Corp. graphics processing units, but only OCI can gang them into “superclusters” that recently topped out at over 130,000 GPUs. The Exadata cloud business has also benefited; given customer demand, and the reality that no hyperscaler can duplicate Oracle’s RDMA optimizations, Oracle database based on Exadata infrastructure is now hosted as a native service on all three hyperscalers. From a modest start, over the past year that multicloud Oracle database business has grown 15 times.

Now of course, the laws of gravity dictate that whatever goes up must come down. There are serious questions as to whether the current AI gold rush will play out like the dot-com boom. And though Oracle’s RPOs dwarf those of its hyperscaler frenemies, the fact is that Oracle as a company is a small fraction of the size with a fraction of the revenue base of its counterparts. That means it has to commit a lot more revenue to capital investments.

The good news is that Oracle is hedging its risks by not buying buildings or real estate for OCI regions. It is only buying equipment and building infrastructure when orders actually come in. Nonetheless, the large stake on a single customer, Open AI, raises questions as to how sustainable the boom is.

Fifteen years ago, who would have thought that Sun Microsystems would turn out to be the reverse acquisition that transformed Oracle? A failing server business at the time, Sun had systems expertise that changed the script. And in the long run that expertise could change Oracle’s primary business from enterprise software to cloud infrastructure provider.

For more detail, check out the podcast I did with Matt Housley:

Tony Baer is principal at dbInsight LLC, which provides an independent view on the database and analytics technology ecosystem. Baer is an industry expert in extending data management practices, governance and advanced analytics to address the desire of enterprises to generate meaningful value from data-driven transformation. He wrote this article for SiliconANGLE.

Photo: Oracle