Oracle is spending billions on AI data centers as cash flow turns negative

Alfonso Maruccia

Posts: 2,610   +978
Staff
Connecting the dots: Will Oracle be the first major company to falter amid the looming AI bubble? Analysts are scrambling to interpret what is really happening at the cloud giant, as co-founder Larry Ellison emphasizes that AI coding tools are actually strengthening Oracle's SaaS business.

Despite operating in a different league than traditional Big Tech giants, Oracle has heavily invested in the AI-driven future of computing and cloud services. The Austin-based company has committed billions to build massive data centers, but recent rumors suggest that this ambitious strategy could be at risk, signaling potential trouble rather than a new golden era for the enterprise SaaS leader.

According to unnamed sources familiar with the matter, OpenAI has reportedly decided to halt its planned expansion of the Stargate data center in Abilene, Texas. The AI startup had partnered with Oracle, which is still investing heavily in building the facility and installing hardware. Initial plans called for even larger and more powerful centers to support what was described as the "unlimited" demand for chatbot workloads.

Credit: App Economy Insights

The sources claim that OpenAI's decision stems from the Abilene site's reliance on Blackwell-based AI accelerators. Nvidia recently introduced the Vera Rubin architecture, which can allegedly train models up to five times faster and reduce inference costs by a factor of 10. OpenAI is now expected to secure access to Vera Rubin accelerators, leaving Oracle tied to a slower Blackwell-based facility.

This shift could create a major challenge for Oracle, as OpenAI seeks higher performance at lower costs, potentially threatening the company's long-term investments in AI infrastructure. For a debt-laden database giant, the timing could not be more critical.

The market has long questioned the sustainability of Oracle's AI spending. The company has reportedly committed $100 billion to build and expand its new hyperscale infrastructure, much of it financed through debt. Unlike Google, Amazon, or Microsoft, Oracle lacks similarly massive revenue streams to fund what some critics call the AI data center "fantasy."

Oracle's cash flow has turned negative, as the company has already spent more than it has earned from its enterprise-focused business. Sources say the company is also planning to lay off thousands of employees, accelerating its transformation from a traditional software licensing firm into a cloud infrastructure provider for large-scale AI operations.

Oracle claims that AI technologies and "vibe coding" tools are making development teams smaller, more agile, and more productive. The company highlighted this efficiency while reporting strong quarterly results that exceeded Wall Street expectations. Overall revenue increased 22% in the third quarter, with cloud revenues up 44% year-over-year.

The company now expects to close the fiscal year with $90 billion in revenue, above analysts' forecasts of $86.6 billion. Co-founder Larry Ellison emphasized that vibe coding is helping Oracle build a complete ecosystem from scratch, suggesting that fears of a "SaaS apocalypse" will not impact the company.

Yet despite the upbeat statements, Oracle's negative free cash flow for the past 12 months reached $13.18 billion, indicating that the Austin-based giant is still burning cash even as it touts its AI-driven efficiencies.

Permalink to story:

 
This is pure speculation… I’m sure some AI-bashing trolls will start posting about how the “bubble has popped”…

Negative cash flow of 13 billion sounds terrible but… they’ve spent over 100 billion in infrastructure… that means their revenue is still pretty awesome. Once the initial infrastructure has been built, the negative should turn to positive - at least, I’m sure that’s what Oracle and their investors are thinking.
 
That's a perfect example of on-demand "journalism".

Extremely biased, concocted from "recent rumors".."unnamed sources".. and a pile of reportedly..allegedly..potentially...
 
This is pure speculation… I’m sure some AI-bashing trolls will start posting about how the “bubble has popped”…

Negative cash flow of 13 billion sounds terrible but… they’ve spent over 100 billion in infrastructure… that means their revenue is still pretty awesome. Once the initial infrastructure has been built, the negative should turn to positive - at least, I’m sure that’s what Oracle and their investors are thinking.
Except that the real expensive part, the GPUs, have short lifespans and will need to be continually swapped out every 3-5 years, while the 30 year debt taken on to fund their purchase still hangs there.
 
Watching Larry Ellison's ill-gotten empire collapse when this bubble pops will be second only to Sam Altman in terms of schadenfreude for me. Couldn't happen to more deserving people.

100%. Just listen to Larry Ellison speak, and you’ll understand he’s just a patsy for those above and beyond him, because dumbasses like him don’t make it in the real world. He’s a front for those that are designing our track and trace control grids. A little known fact, Ellison owns most of your EMRs (electronic medical record), so he’s got the inside (health) scoop on everybody.
 
This is pure speculation… I’m sure some AI-bashing trolls will start posting about how the “bubble has popped”…

Negative cash flow of 13 billion sounds terrible but… they’ve spent over 100 billion in infrastructure… that means their revenue is still pretty awesome. Once the initial infrastructure has been built, the negative should turn to positive - at least, I’m sure that’s what Oracle and their investors are thinking.
No, this $100 billion debt is not "speculation". It is called a fact.

Revenue is neither "awesome" nor "horrible", it's just a meaningless number in a sheet without the expenses counterpart, coupled together also known as profit.

Walmart's annual revenue is $700 billion, yet their annual profit is... $20 billion. Quite a difference, don't you think? The only thing revenue tells you about is volume. But volume doesn't imply making money in any shape or form whatsoever. In fact, if you're selling at a loss, the bigger the volume, the more losses you incur.

And Oracle is currently selling at a loss. Their profits are currently negative, which is a far cry from this "awesomeness" you just hallucinated.
 
Last edited:
I don't really have my finger on the pulse of this subject, being in my 70's and not really giving a flying **** about AI - but reading the above, it seems to me that the super-duper overblown Stargate centre is already out of date...over the hill, so to speak, before it's even completed. No doubt it will still eventually be completed, but the efficiency will be.............well, basically crap, compared to other centers which may plan to use those newer Nvidia accelerators. You have to give Nvidia their fair dues -they have some classy people working there.
 
Except that the real expensive part, the GPUs, have short lifespans and will need to be continually swapped out every 3-5 years, while the 30 year debt taken on to fund their purchase still hangs there.
Money spent - 100 billion
Money earned - 90 billion

3-5 years from now, money earned = 270-450bill (minimum, as it will almost certainly increase)

Debt gone in a year or 2… plenty of cash to spend in 3-5 years on new GPUs…
No, this $100 billion debt is not "speculation". It is called a fact.

Revenue is neither "awesome" nor "horrible", it's just a meaningless number in a sheet without the expenses counterpart, coupled together also known as profit.

Walmart's annual revenue is $700 billion, yet their annual profit is... $20 billion. Quite a difference, don't you think? The only thing revenue tells you about is volume. But volume doesn't imply making money in any shape or form whatsoever. In fact, if you're selling at a loss, the bigger the volume, the more losses you incur.

And Oracle is currently selling at a loss. Their profits are currently negative, which is a far cry from this "awesomeness" you just hallucinated.
Revenue is important when expenditures are not an ongoing cost. While they will be in the red for the short term, clearly they are banking on being profitable as soon as those expenditures are finished…

Why do people naturally assume that CEOs of multi-billion dollar corporations are stupid? Does it make you sleep better at night maybe?
 
Money spent - 100 billion
Money earned - 90 billion

3-5 years from now, money earned = 270-450bill (minimum, as it will almost certainly increase)

Debt gone in a year or 2… plenty of cash to spend in 3-5 years on new GPUs…
$100B to build infrastructure
$20B/year to run it ... it makes $100B in 5 years
If Oracle had to borrow those $100B, then there will be interest.
And do not forget to fund some R&D. Otherwise you wont get anything to sell for profit.

In the end Oracle will be happy to get $50B to replace GPUs after 5 years.
 
Bottom line is, the investment is not sustainable. And you know it when your expenses is consistently more than what you are earning. All these big techs are cutting headcount left right center to try and keep the bubble inflated, but there is a limit to how many heads you can cut until it impacts the business. AI cannot reliably be used in many industries without someone actually validating and correcting the output, so that means you will always need someone there. For Oracle, their credit rating is not looking too good with debts swelling over the past couple of years. So if they cannot get financing easily or affordably and not able to turn a profit in foreseeable future, the only way is for some companies like Nvidia or Softbank to throw more money to burn.
 
Back