3w ago

The Algorithm That Detected a $610 Billion Fraud: How Machine Intelligence Exposed the AI Industry’s Circular Financing Scheme

The current situation bears structural similarities to three major accounting frauds: Enron (2001), WorldCom (2002), and Lucent Technologies (2000).
Lucent, once America’s largest telecommunications equipment manufacturer, grew revenue through vendor financing arrangements. The company lent money to telecom carriers to purchase Lucent equipment, booking the equipment sales as revenue while the loans appeared as receivables. When carriers couldn’t repay, Lucent took $8.7 billion in writeoffs.
Lucent’s DSO peaked at 64 days before the fraud became public. Nvidia’s current 53-day DSO remains below that threshold but exceeds its historical baseline by the same percentage that preceded Lucent’s collapse.
Enron used Special Purpose Entities to hide debt and inflate revenue. These entities existed as legally separate companies but were economically controlled by Enron. The structure created artificial revenue through transactions with entities Enron itself funded.
The xAI SPV structure mirrors this approach. Nvidia provides equity capital to an entity that exists primarily to purchase Nvidia products. The transaction appears as an arms-length sale in Nvidia’s accounting, but economically, Nvidia is funding its own revenue.

Technology @lemmy.world

Thomas2024 @feddit.org

1w ago

substack.com /inbox/post/179453867

30 comments

I mean… a few people have been pointing out this shit doesn’t make sense for a few months now.
- I know. The content of the article is a useful dissection of the funky accounting going on, but it's stuck with a beside-the-point headline, unfortunately.
- But as today was a machine they listened, because nowadays we trust AI more than people.
- The stock market is wholly price controlled by the biggest players via dark pools, algorithms, AI and HFT and that there will not be an appreciable crash ever again because of this. So it's going to be very interesting as it continues to make less and less sense and we continue to get ever-higher numbers coming from Wall Street.
  Edit: downvoted every time man, people really want to believe in the stock market (or maybe they just want to believe it can go down in flames lol). Well that's cool, the big players I'm sure are glad for your existence. But you'll all come around.
  Edit: Oh would you look at that, another week, another up by several thousand points. Come back and hit me up when it crashes, boys!
  
  That would require really large numbers of the most selfish people ever to work together as a group and not break ranks even when self-interest dictates they should.
  
  That would be way too smart for reality.
  
  Nah, as soon as somebody starts missing payments things are gonna deleverage fast
- This has been said ever since cryptocurrency became a thing.
“Vibe Revenue” is my new favorite term.
- Vibe revenue courtesy of their vibe business plan to acquire more vibe customers. 😎
Bubble burst 2026!
- Please make it 2025 Santa, that's all I ask.
  
  It will burst the moment Democrats get back in power and not one second before, by design
It didn’t really take ai to figure that one out.
GPU spot pricing on third-party compute marketplaces provides additional evidence. Vast.ai and RunPod.io, platforms where users rent GPU compute capacity by the hour, show H100 spot prices declining from $3.20 per hour in August 2025 to $2.12 per hour as of November 20, 2025. This 34% price decline directly contradicts claims of insatiable demand.
This is interesting… especially to someone interested in cheap rentals.
The cynical part of me figures a GPU “crash” would result in a lot of hardware liquidation (eg Nvidia repossessing GPUs and throwing them out) and not affect prices much. Heck, prices may get even worse with fewer providers bothering to host GPU instances… but maybe not?
I would love for a mega crash to send H100 prices underground, and knock down other GPUs in a cascade.
H100 spot prices declining from $3.20 per hour in August 2025 to $2.12 per hour as of November 20, 2025
This is probably the key metric to lead to a crash. The US empire is all in on NVIDIA other AI oligarchy, and western chip industry is all in on making only datacenter AI chips. Computers, and soon phones, will become grossly overpriced and unbuyable. Even if TSMC makes only and all of the "AI brains chips", there's not enough memory production to package them all into GPUs, and then actually less TSMC production/profits even though it is positioned as a winner no matter which hardware company succeeds. This can also crash economy and consumer inflation.
The actual worst circular financing AI deal is not mentioned, and involves an older traditional scam. If you buy 1 share in a private company for $1000, then the company boasts a PR where its new value is $1000/share. VC firms like Softbank have always done this to boost their accounting value, because they bought millions of share at $100 or less, and small amounts of new shares makes whole stake appear to succeed.
Anthropic recently made a circular deal with NVDA and MSFT investing $15B in Anthropic, but at over double its valuation. This gives Anthropic shareholders over $150B in "value". Circularly, Anthropic will be paying $30B to the above 2 companies.
xAI and StarX also made scam new valuations for each other early this year. They just swapped shares among themselves, while imagining a new $ amount for each where accounting permits the imaginary to be real.
The “Vibe Revenue” Admission
On November 14, 2025, at the Web Summit conference in Lisbon, multiple AI company CEOs acknowledged this dynamic in public for the first time.
Brian Chesky, CEO of Airbnb, stated: “There’s a lot of vibe revenue in AI. Companies are talking about billions in pipeline that may never materialize.”
Vinod Khosla, venture capitalist and prominent AI investor, told the audience: “Ninety-five percent of AI startups will fail. The question is which five percent become Google.”
Sam Altman, CEO of OpenAI, said: “We’re in uncharted territory. Nobody knows if this scales to AGI or hits a wall at GPT-5.”
These admissions carry weight because they contradict the growth narratives supporting current valuations. OpenAI, valued at $157 billion in its most recent funding round, reported $3.7 billion in revenue for 2025 according to The Information. The company simultaneously disclosed operating expenses of $13 billion, resulting in a $9.3 billion annual cash burn.
- Good grief!
Oh, the irony!
It's an interesting read. But it doesn't account for external investments pushing up the total capital flow. Entities like the Saudis, Dubai, and Qatar, as well as many sovereign funds have a lot of cash on hand to pump into the system and keep the party rolling.
I wonder if the short-selling models account for these anomalies.
- This bubble will pop because the profits won't come close to what is needed. They're betting the house on developing agi with models that are completely unsuitable for that purpose. Even their redefined agi they most likely won't reach. Basing agi on llm is fundamentally unsound, therefore can't live up to the hype and won't get the record adaptation required for profitability.
  The arabs can pump all their fossil fuel money into this and keep the development going for much longer than it should, but even they will want to see returns at one point.
  
  I wish you were right, but ponder the last point - do they really want returns ? For the oligarchy it’s just about maintaining power, arabs bought crazily expensive us fighter jets for years with zero 'returns', except political favours. In this case datacenters also support fossil demand expansion narratives. Has anybody done the math of the scale of those 'reserves', compared to the bubble ?
- It's a weird investment for Zionist dictatorship allies. Unclear why their economy needs massive datacenters, but like the US, Skynet is appealing population control, and governments will be leading consumers of datacenters.
  Still, there is a shortage of AI chips (we are told... only because government is the only real customer), and diverting them to where power can be built, makes both AI chips more expensive for US, and buyers unable to compete on operating (power) costs.
  Same issue with "profit from China" policies. H20s are great inference cards and cheap and should be purchasable in US outside of Chinese ebay sellers. This week announcement that NVIDIA will be allowed to sell H200s to China, is just giving all US money to NVIDIA policy.
Wow that... Didn't take AI at all.
People have been pointing this out for months.
It's true then? Ai will make people lose their (ai) jobs?
If you can't envision a robust and healthy economy based on the ability to create an endless stream of pictures of comically obese men and videos of same, I don't want to participate in that economy.
This is the future. We no longer need a connection to physical reality, as long as bits flow, a few people can create money from that.
The rest can 3D print what they need.
Site doesn't load

30 comments

The “Vibe Revenue” Admission