Oracle’s AI Bet Is Either Genius or Insanity — There’s No Middle Ground

If you want to know whether the AI bubble is actually popping, stop watching Nvidia’s stock ticker and start watching Oracle.

Yes, Oracle. The database company. The one that’s been around since the 70s, run by the same billionaire sailor who still refuses to retire. Larry Ellison has made a move that’s either going to look like a masterstroke or a spectacular faceplant, and I’m not sure which yet.

Oracle has burned its boats. They’ve pivoted to AI, but not in the way you’d expect. They’re not building foundation models — that’s OpenAI and Anthropic’s game. They’re not a neocloud in the CoreWeave sense, though they’ve jumped into the same bare-metal rental business. They’re a SaaS company that has bet the house on a very specific vision of AI’s future, all while their traditional database business gently declines.

What makes this interesting is Oracle’s age. It’s significantly older than almost every AI competitor except Microsoft. Most legacy tech companies are tiptoeing into AI with cautious pilots and partnership announcements. Oracle is diving headfirst off a cliff.

The bet is on infrastructure. Oracle is spending billions on data centers filled with Nvidia GPUs, renting them out to AI companies that need raw compute. It’s a capital-intensive, low-margin business on the surface, but Ellison is playing a longer game. If AI workloads eventually need to sit on top of massive, reliable databases — and they will — Oracle wants to be the layer underneath.

It’s a classic Ellison move. High risk, high ego, high potential reward. The man loves being the contrarian in the room. But here’s the thing: being contrarian only works if you’re right. If AI demand softens, or if the hyperscalers eat Oracle’s lunch on pricing, those billions in GPU clusters become very expensive paperweights.

I’ve seen this pattern before. Oracle tried similar aggressive pivots in the cloud wars a decade ago and ended up playing catch-up to AWS and Azure for years. The difference this time is that Ellison seems to have identified a genuine gap: AI companies need raw compute now, and the big clouds are too rigid or too expensive for some workloads. Oracle’s bare-metal offering is simpler, faster to deploy, and priced aggressively.

Still, I’m not entirely sold. The AI infrastructure market is already crowded, and the barriers to entry aren’t as high as Oracle would like you to believe. CoreWeave, Lambda, and even Google are all competing for the same dollar. Oracle’s edge is its existing enterprise relationships — every Fortune 500 already has Oracle in their stack somewhere. If they can upsell AI infrastructure to those customers, it’s a solid play.

But that’s a big “if.” Enterprise sales cycles are slow, and AI is moving fast. Oracle’s traditional strength — selling expensive, locked-in software to CIOs — doesn’t translate perfectly to a market where startups want to spin up 10,000 GPUs overnight and cancel the contract next quarter.

I’ll give Ellison credit for guts. Most CEOs his age are coasting to retirement, not betting the entire company on a technology that might be overhyped. Whether it’s genius or insanity, it’s going to be fun to watch.