The Elephant Dancing to AI's Tune

There's a company so old and so large that everyone forgot it's actually an AI giant too — database and data-center powerhouse Oracle. Its founder, Larry Ellison, recently saw his net worth spike to $340 billion, briefly making him the richest man in the world. He only held the title for a short while before Elon Musk overtook him again, dropping him to the world's second-richest person — but it was enough to make him the talk of the town.

Customers Rush to Lock In Capacity — $450 Billion in Future Orders Materialize

For a company at roughly $800 billion in market value, a huge single-day swing is generally unlikely. So it was genuinely shocking that Oracle managed to surge 40% in one go. The main driver: Oracle's future revenue backlog — its RPO (remaining performance obligations, i.e., orders not yet delivered) — hit a staggering $450 billion, a full 230% above what was projected just one quarter earlier. That's an increase of over a hundred billion dollars in future bookings in a single quarter — you could practically invest in this with your eyes closed.

These massive orders mainly represent customers pre-booking Oracle's data-center capacity and compute resources, locking in contracts ahead of time. Over the next five years, Oracle will effectively be building data centers and selling that cloud capacity to customers simultaneously.

This is also exactly what critics of these deals point to: Oracle's biggest customer, OpenAI, reportedly has only around $20 billion in annualized revenue right now — so how do you sign a five-year, $300 billion contract with them? This kind of criticism echoes the skepticism aimed at the various American tech giants' Stargate project earlier this year — a plan to spend $500 billion building AI infrastructure without, critics note, actually having that money in hand. I think this line of thinking underestimates both the fundraising capacity of American tech giants and the sheer national-level commitment behind pushing US AI infrastructure to a commanding lead.

When the dust settles, I suspect both the critics and the believers will turn out to be right — probably at the same time. Some tech giants will indeed fail down the road, but the AI infrastructure built with all that money will genuinely prove useful, creating even more wealth in the process.

What makes Oracle interesting is that its founder, Ellison, is still running the company he started — a rarity in the hyper-competitive tech industry. He's practically a mythical creature who did business with Steve Jobs himself. Even his protégé, Marc Benioff, went on to found the wildly successful cloud company Salesforce.

Benioff was once seen as Ellison's likely successor, but he eventually left to start Salesforce — with Ellison even investing in it, a story that became something of a legend in Silicon Valley. But Salesforce and Oracle later became fierce competitors, which led Ellison to step down from Salesforce's board. You could say Oracle's DNA has a firm grip on America's — and even the world's — enterprise software market.

In this current AI wave, American tech leaders are taking noticeably different approaches. Apple seems to be sticking with a wait-and-see posture, while the bosses at Google, Meta, Microsoft, and OpenAI are all going all-out to showcase their positions at the AI frontier. Watching some of Ellison's own talks, his style stands apart — the emphasis seems to be squarely on helping customers succeed. That positioning has found him a real niche: selling shovels to the gold miners, and his primary customer there is OpenAI.

Over the past six months, Google has released a relentless stream of AI products that are both excellent and cheap, and this offensive has driven the point home: the king of the internet still has real teeth. Google also has arguably the best AI infrastructure around — compute, cloud, algorithms, software, and users, it's got everything. The only thing it was missing was resolve, and now it doesn't lack that either.

Could Oracle Steer the TikTok Deal? A Possible Future on Oracle Cloud

OpenAI is too far behind to build out its own infrastructure fast enough now, so partnering with Oracle is the only way to complete AI infrastructure buildout at the speed needed to hold its current position. Ellison shrewdly recognized this, positioning Oracle to partner with other tech giants rather than fighting them for frontier business directly.

There have also been recent media reports suggesting Oracle could end up leading the TikTok acquisition — another tailwind. Oracle has teamed up with top-tier VC firm a16z and others to form a consortium bidding for TikTok, and if the deal closes, TikTok would run on Oracle Cloud — a genuinely massive amount of compute usage.

Let me take a brief detour on the TikTok deal itself. In practice, there are two likely outcomes going forward, and the main difference is whether the recommendation algorithm is included in the sale or not. Conventional wisdom holds that TikTok's recommendation algorithm is extremely powerful, and that if it isn't included, US TikTok would face a huge headache rebuilding it from scratch.

My own view: while the algorithm is important, if the user base is large enough and stays actively engaged, the algorithm doesn't necessarily need to be sold along with everything else — and the buyer may not even need to buy it, since the acquiring tech companies are perfectly capable of building their own.

So my guess is that the final deal likely won't include the algorithm — or it'll switch over to a new algorithm at some point down the line. If data security concerns are the whole reason TikTok is being forced to sell in the first place, then rebuilding the algorithm from scratch isn't such a strange outcome. And besides, today's AI and compute infrastructure is orders of magnitude more advanced than it used to be, which makes rebuilding all of this far easier than it once was.

This AI wave has already shown us an elephant that can dance — and even fly. Is there a bubble risk lurking? Personally, I think stock prices will have their ups and downs, but the underlying AI demand is real.

The demand for AI cloud capacity right now is real too. Meeting that demand requires real solutions, and while, in theory, all that compute could someday sit on the user's own device, no such solution is realistically on the horizon in the near-to-medium term. So AI-related industries should still be in for a good stretch of prosperity.