Ellison has been unusually loud about Oracle's AI ambitions, pushing what he calls the "AI factory" model: massive data centers stuffed with GPUs, rented out for eye-watering sums. He's floated promises about AI that monitors cattle, diagnoses cancer, and tracks every human heartbeat. The marketing is classic Oracle—bombastic, absolute, impossible to verify. But the spending is real. Oracle's capex now looks more like a hyperscaler's than a legacy database company's.
The actual question isn't whether AI is a bubble. It's whether Ellison's specific bet—building GPU capacity at a scale that goes toe-to-toe with Microsoft and Amazon—becomes the trigger. The optimistic read: the cloud infrastructure shortage is genuinely painful, and whoever locks up silicon now owns the next decade of enterprise AI. The pessimistic read: we're seeing a fiber-optic rerun, where everyone overbuilt because they assumed the demand curve was a straight line up.
What makes Ellison such a perfect face for this cycle is his rhetoric. He doesn't talk in incremental improvements. He talks about "infinite demand" and machine intelligence that makes humans look like house pets—all arriving within a few years. That isn't a sober forecast; it's a hype sermon. And when a legend of enterprise software starts preaching, the market tends to listen—until it stops.
There's also the personal layer. Ellison has never forgiven himself for letting AWS and Azure run away with the cloud. AI looks like his second chance. Oracle's infrastructure business isn't just a profit center; it's an ego project. That makes the stock more reactive than the balance sheet alone would suggest. Buying Oracle as a pure AI play means buying Ellison's ability to outspend, outtalk, and outlast his rivals.
I don't know if this ends in a crash. The real demand for AI compute isn't fabricated—companies are burning tokens and inference budgets like there's no tomorrow. But the market cap implies Oracle will keep winning hyperscale contracts forever, which is far from certain. One missed renewal, one cooling macro window, and the
AI's First Tremors Are Already Showing Up in Digital Revenue 3h ago
[Training Data, Not Compute: Ex-OpenAI Researcher Bets $100B 5h ago](/en/news/4491/)
[OpenAI Slashes GPT-5.6 Luna Price 80%: A Pricing Deep Dive 5h ago](/en/news/4489/)
[Leopold Aschenbrenner Got AGI Wrong: A Cautionary Tale 6h ago](/en/news/4480/)
[GPT-5.6 Revenue Surge: How July Beat Q2 15h ago](/en/news/4448/)
OpenAI and the US Government Just Rediscovered the "Blank Map" 18h ago
Next Three organizations got breached in a controlled exercise — and →