radiofree.asia

Blair-backing Digital ID magnate just lost £370bn in data centre disaster

Oracle is the database and cloud computing company which is owned by Larry Ellison. Ellison wasn’t that well known in the UK until recently, but he became infamous after it was revealed he backs the Tony Blair Institute.

Even worse than this, he’s not just a supporter of Tony Blair; he’s a key supporter of Blair’s worst idea — the loathed Digital ID scheme:


With all this in mind, you may be pleased to learn Oracle’s AI drive is turning into something of a car crash:


Oracle and the AI bubble

There are many technologies which are labelled ‘AI’, but the most notable in recent years has been ‘generative AI’ — i.e. the tech which drives text, image, and music generators like ChatGPT, Sora 2, and Suno.

In simple terms, these systems are predictive engines, which means they generate content by predicting what happens next. While this technology is impressive, it suffers from ‘hallucinations’, which is the technical term for ‘making stuff up’. These hallucinations are an intrinsic part of the generative process, and as of right now there seems to be no way of stopping them.

One key problem with the generative AI boom is that the hallucination problem makes this technology impractical for many work places. It’s difficult to rely on software which works well sometimes but shits the bed when you’re not expecting it. The issue creates a situation in which users have to handhold their AI tools, and this means the time AI saves is lost in other ways. This is why one report found 95% of AI pilots failed, and another found 42% of companies abandoned their AI initiatives in 2025.

The other issue is AI companies like OpenAI have been attracting investment by claiming generative AI — a predictive technology — will one day turn into general artificial intelligence, a thinking technology. Some people believe this is possible; others do not:

The bubble economy

Regardless of who’s right, the current state of play is:

  • These companies are not making significant profit from AI products.
  • There are no signs that their tech will magically turn into AGI.

Or that was the state of play, anyway. A recent leak revealed by tech journalist Ed Zitron showed that:

  • Open AI’s services cost a lot more to run that we thought.
  • Open AI’s services make a lot less money that we thought.

Exclusive: Based on documents viewed by this newsletter, OpenAI spent over $12.4 billion on inference from 2024 to September 2025. As part of its Microsoft revenue share, it sent $493.8m in 2024/$865.8m Jan-Sep 2025, implying lower revenues than previously reported.
www.wheresyoured.at/oai_docs/

[image or embed]

— Ed Zitron (@edzitron.com) 12 November 2025 at 16:30

In other words, this is all a bubble, and the only reason it hasn’t popped yet is because the US government is propping it up.

Why are they propping it up?

Because they’re worried they’ll go into recession when the crash eventually happens, as AI czar David Sacks made clear:


This may work for a while, but you can only swim against the tide for so long, as Oracle are now finding out.

Underwater

On 10 September, Oracle announced a deal with OpenAI. Their stock prices jumped up as a result, but they’ve since plummeted, as this graph from the Financial Times shows:

AI companies like OpenAI require these massive data centres so their users can mass produce godawful slop like the following:


These data centres cost billions and use significant amounts of electricity:


Additionally, it turns out the GPUs data centres are using are advancing so quickly that they need to be replaced every few years, as CITP blog reports:

One unnamed Google architect assessed that GPUs running at 60-70% utilization—standard for AI workloads—survive one to two years, with three years as a maximum. The reason: thermal and electrical stress is simply too high.

But physical failure isn’t the only concern. Technological obsolescence drives replacement cycles. Nvidia’s GB200 (“Blackwell”) chip provides 4-5x faster inference than the H100. When competitors deploy hardware with significantly better performance, three-year-old chips become economically obsolete even if they still function.

In other words, these unfathomably expensive data centres are somehow more expensive than we realised. Combine this with the fact that many of the newer ones exist solely to support a failing technology, and you can see why stock valuations are plummeting.

Oracle

We’ve no idea how long it will take for all this to come crashing down, but we can say this; do not invest in these companies right now.

We’ll add that with a company named ‘Oracle’, Larry Ellison really should have seen this coming.

Featured image via World Governments Summit

By Willem Moore

This post was originally published on Canary.