By For years, Larry Ellison was Silicon Valley’s favorite dinosaur. The Oracle co-founder mocked the cloud as a marketing gimmick while Jeff Bezos, Satya Nadella, and Sundar Pichai built empires on it. Reporters cast Ellison as the old guard: a vaguely ridiculous mogul best remembered for yachts, leather jackets, and a private island, not vision.
Then the AI compute crunch arrived, and Ellison did the least fashionable thing in tech: He showed up with a hard hat. He stopped trying to win the cocktail party and set out to guarantee the one thing developers, startups, and megaclients couldn’t get enough of: compute capacity. Power first, buildings second, racks stuffed with GPUs, and signatures on contracts that lock in that capacity for years.
The market’s verdict was unambiguous.
On Sept. 9, Oracle dropped a number so large it just about made Wall Street choke: $455 billion in remaining performance obligations — contracts already signed and committed, up 359% from a year earlier. Four separate multibillion-dollar deals landed in the span of a single quarter. On Wall Street, that backlog looked less like a tech pipeline and more like a public-works budget. Investors reacted accordingly: Oracle stock recorded its biggest one-day jump since 1992.
And because Ellison still owns roughly 41% of the company, the paper math briefly made him the richest man alive. For a moment — until Tesla shares twitched and Elon Musk reclaimed the top spot — the octogenarian who once sneered at the cloud sat atop the Bloomberg Billionaires Index.
None of that made Oracle cooler. But it did make Ellison unavoidable.
The guy who once sneered at “cloud” had been rewarded for selling the hardest part of it: time and power — two things the AI boom can’t get enough of. Oracle isn’t the playground for cool developers, and Ellison doesn’t pretend it is. Ellison didn’t win back relevance by reinventing software. He did it by making himself the landlord of an AI factory economy that runs not on hype but on megawatts, racks, and signatures that stretch into the 2030s.
While everyone else sold delight, Ellison sold delivery. And in a year when AI hype kept smashing into the laws of physics, boring delivery suddenly became the hottest commodity in tech. What Oracle is selling is usable compute capacity — power, racks, and the infrastructure to run workloads when and where customers need them. That’s why the market now values Oracle less like a software vendor and more like a utility landlord with a cloud console. The product isn’t features; it’s capacity on a schedule. That’s how you get a backlog number that reads like a public-works budget and a one-day wealth spike big enough to rearrange the billionaire rankings, even if only for an afternoon.
That’s the context for Ellison’s late-career reinvention. He hasn’t gotten younger or cooler. But he’s managed to turn Oracle — a company written off as legacy a dozen times over — into the kind of entity customers may grumble about but can’t quit. That’s the trick. Not making people love you. Making them need you. And right now, they do.
From cloud laggard to power broker
The story of Ellison’s late-career turn is, at its core, about choosing the bottleneck. For decades, he sold software licenses and databases, the plumbing of corporate IT. As others rode the cloud wave in the 2010s, Oracle lagged — badly. AWS, Microsoft, and Google vacuumed up developers and mindshare. Ellison was caricatured as “lock-in Larry,” a relic of a bygone era.
What he’s doing now looks very different. Instead of chasing SaaS logos, Oracle is chasing inputs: GPUs, land, and electricity. Scarcity has flipped the script. In 2025, the rarest product isn’t clever code; it’s a data hall with power actually flowing to it. Oracle has leaned into that role. In Shackelford County, Texas, a developer is building a 1.4-gigawatt campus dubbed Frontier, and reports suggest Oracle is ready to keep it humming with gas generators if utility hookups lag — an inelegant but effective way to deliver capacity on time.
Ellison’s posture is pragmatic. He no longer talks about locking customers into his walled garden; he talks about multicloud. Oracle now cuts deals so its databases can run inside Amazon, Microsoft, or Google’s infrastructure. On the September call, CEO Safra Catz highlighted “multicloud database revenue” growing at four-digit percentages off a small base. The substance matters less than the shift: Oracle is finally selling itself as a partner, not a jailer.
Those four new multibillion-dollar agreements in a single quarter didn’t just pad remaining performance obligations (RPO); they signaled that some of the most compute-hungry customers are willing to bind themselves to Oracle for the long haul. But Ellison’s newfound indispensability could rest on one giant asterisk: Oracle’s backlog is just paper until the lights actually flick on. A $455 billion stack of obligations looks impressive on a slide; it looks a lot less certain if substations don’t connect on schedule. Oracle’s reinvention is real only if it can turn signatures into electrons, and electrons into cash.
The biggest question mark is concentration. The Wall Street Journal reported that a huge slice of that backlog is a single customer — OpenAI — agreeing to spend $300 billion of compute over five years starting in 2027. That’s whale hunting on steroids. It’s great if the whale keeps swimming; it’s disastrous if it migrates for the season. For now, Ellison gets credit for landing the catch. But the perception of being “the OpenAI cloud” isn’t diversification. It’s dependency. Whale accounts make for glorious top lines and nervous CFOs. If a single AI tenant represents an outsized share of the stack, Ellison’s “overflow valve for the industry” pitch risks turning into a “whale-in-a-bathtub” reality.
It’s a clever workaround but also a reminder: Ellison’s reinvention is built on concrete and transformers, not just code. If deadlines slip, regulators balk, neighbors complain — suddenly, those glowing backlog numbers start to look like IOUs. RPO converts into revenue over years, and in Oracle’s case, much of it won’t show up until later this decade. That’s why analysts immediately flagged the risk ledger. Concentration risk: If OpenAI is the whale behind much of that spike, what happens if it stumbles or diversifies its compute elsewhere? Conversion risk: $455 billion on paper is only as good as the substations, transformers, and racks that turn it into cash. And power risk: Temporary gas plants may solve deadlines, but they invite regulatory scrutiny and environmental blowback.
The cloud is no longer defined by APIs and dashboards. It’s defined by who can spin up 20 megawatts of power and a few thousand GPUs without waiting a decade for utility approvals. Suddenly Ellison’s taste for the industrial — steel, substations, long contracts — looks prescient. Oracle’s Shackelford County, Texas, move would have gotten him laughed off the stage 10 years ago. Today it reads like operational competence.
This isn’t your parents’ Oracle, minting license fees with enviable margins. Capital expenditures have ballooned; free cash flow is projected to dip negative as billions are poured into new capacity. That’s not unusual for utilities. But it’s unusual for a software company whose appeal once rested on 30%-plus operating margins and steady license renewals. Investors are, in effect, funding Ellison’s reinvention on the belief that those multiyear deals will convert. Ellison is asking investors to think less like software analysts and more like utility regulators. So far they’ve obliged. But investor patience is a short fuse. Miss too many energization dates, and the market turns on you faster than an overworked generator.
When spreadsheets meet substations
Markets love a clean story, and Ellison gave them one. A massive backlog. A marquee customer. A stock surge. A billionaire shuffle.
Ellison has always had a taste for control. Owning the database market wasn’t enough; he wanted to own the hardware that ran it, the salespeople who hawked it, the campus they worked in, even the Hawaiian island he retreated to when he was done with them. Control was the through-line.
In the 2010s, Ellison’s instincts looked maladjusted. He mocked the cloud — “Maybe I’m an idiot, but I have no idea what anyone is talking about. What is it? It’s complete gibberish. It’s insane. When is this idiocy going to stop?” — while Salesforce’s Marc Benioff railed against Oracle’s “false clouds” while riding the real thing to glory (and turning Ellison into a pantomime villain). Analysts routinely described Oracle as a laggard. The company wasn’t trendy, it wasn’t growing, and it certainly wasn’t where young engineers wanted to be.
For much of the past decade, the Ellison name was in the headlines less because of Larry and more because of his children. David Ellison parlayed family billions into Skydance, a studio that financed “Mission: Impossible” and “Top Gun: Maverick” before muscling its way into a Paramount takeover fight. Megan Ellison chased a different corner of Hollywood, underwriting prestige fare from “Zero Dark Thirty” to “Phantom Thread” through Annapurna Pictures. For a while, it felt like the Ellisons who mattered were the ones writing checks in Los Angeles, not the one in Redwood Shores.
Ellison’s greatest public trick has always been the mix of showman and street fighter. The version of him that wins this decade is less pirate, more project manager — the executive who can charm a city council on Tuesday and demand a substation delivery on Thursday, all while keeping enough peace with rivals to keep the multicloud escape hatches open. That’s not the myth he built in his youth, but it’s the role the market happens to need now.
By leaning into what once made him look outdated — the industrial, the unglamorous — Ellison has stumbled into relevance again. Oracle isn’t the trendiest place to work. Its developer halo is dim. But in an economy where AI demand is bottlenecked by transformers and substations, being reliable and boring is a competitive advantage.
So what do we call this phase of Ellison’s career? Act One was the database kingpin years, where Oracle minted money selling corporate plumbing. Act Two was the cloud skeptic turned late adopter, the era of “false clouds” and missed growth curves. Act Three is shaping up as something stranger: Ellison as a utility baron, selling electrons and floor space dressed in the language of the cloud.
The risks are real, the timelines are long, and the crown of “world’s richest” was fleeting. But Ellison has already rewritten the narrative. The man who once sneered at the idea of the cloud is now selling the part everyone else forgot to brand: time to power. For the moment, the lights are pointed his way. Not because he dazzled anyone. Because he promised something simple and scarce — and convinced the world he could deliver it. And in a business addicted to the next shiny thing, that’s the oldest trick in the book.
This article first appeared here.
shannonigans 