AI, hyperscalers & credit default swaps, a 2008 love story
 

This is something of a lightbulb theory I had last night, so not well-researched, thought I'd post to get input and discussion.

I posted a bit ago asking where the trillions of investment $ for AI and hyperscalers are coming from, it turns out much of it is private credit, based on a hope that AI will one day be monetized on an XX or XXX% return scale.

But the AI bubble can pop, will pop, just like the dotcom and housing bubbles did causing massive economic disruption and socialized losses with consumers paying higher electric bills to build datacenters and then soften the losses of the ultra-wealthy when the promise doesn't work out.

Have the risks of CDS's been fixed since 2008? Or exacerbated?

Are they public to know who has bought what?

Can interconnections be determined to forecast and model failure?

Other than AI delivering only on a small fraction of the forecasted promise, what other triggers could ignite a worldwide meltdown à la 2008?

I’ve been out of this space for a long time now, but I’m pretty sure that notional CDS volume is way lower now than it was in ‘08. During the subprime crises the notional volume on credit default swaps dwarfed the size of the underlying bonds - so many were used for pure speculation rather than hedging. There have been several regulatory changes since then too which I think has helped. Data is hard to come by unfortunately since these are private contracts - there is some data available but it by no means covers the true size of the market.

As for AI, I definitely agree there will be a crash at some point. It’s just hard to imagine that there will be returns commensurate with the massive amount of investment that is flooding into this space right now.

Racers in my sphere who rent space at the facility have confirmed that they’ve been told to pack their things and be out by December 31. The same source indicated that adjacent properties have also been purchased. As to what’s next, that’s not entirely clear. The announcement’s headline gives credence to the rumors that it’s been sold to a developer, though as yet there’s been no confirmation of the other prevailing theme—that a data center will take over the facility.

https://www.hagerty.com/media/opinion/rip-pittrace-another-amazing-track-closes/


The original Beaverrun was really fun, I got down to the1:0s there, so almost a fast guy. The expansion made it more funner ;) almost Barber like.
Spent a few Sunday nights making the long drive home with a few wood plaques on the dash.
Sad to see it go. Rumors are it was a ridiculous amount of money.

.


.

AI companies are borrowing more money to invest more in AI.


Scott Adams touched on this during one of his recent podcasts. His opinion is that AI companies are so heavily invested in each other, they are too big to fail. If one goes down, the others have to support it.

https://substackcdn.com/image/fetch/...4_663x800.jpeg

1 hour ago

Meta Plans to Cut 600 Jobs at A.I. Superintelligence Labs

The cuts will not affect Meta’s newest A.I. hires, who are in some cases being paid up to hundreds of millions of dollars. The layoffs are focused on correcting an earlier hiring spree.

IDK why links are posted to paywalls. Just to frustrate people or to show that you are a subscriber to the NY Times?

Starting to think AI has gone too far already:

Kohler's Toilet Camera Analyzes Your Poop


A stinkin' camera in my toilet bowl.....

Sorry to dump on your thread Shaun......

I'm sorry Milt, I guess I exceeded my 10 free posts of NYT articles this month. When this has happened in the past, people have nicely PM'd, emailed, texted, posted asking if I could include the article in the post since they can't see it. Here you go. I wish I could copy and paste the Comments for the article, really all articles. NYT Comments are one of the best reasons to subscribe.

Meta Cuts 600 Jobs at A.I. Superintelligence Labs

The layoffs do not affect Meta’s newest A.I. hires, who are in some cases being paid up to hundreds of millions of dollars. The cuts were focused on correcting an earlier hiring spree.

Meta said on Wednesday that it cut approximately 600 jobs in its artificial intelligence division, according to a memo sent to employees that was relayed to The New York Times, as the company seeks to keep pace with competitors in the furious contest over the technology.

The layoffs were in Meta’s so-called Superintelligence Labs, which is the umbrella name for the company’s A.I. efforts. The division has around 3,000 employees, though the exact number of workers was unclear.

Mark Zuckerberg, Meta’s chief executive, has been on a hiring spree to stack his company with top A.I. researchers, including a new chief A.I. officer, Alexandr Wang, earlier this year. The cuts on Wednesday did not affect these newest hires, who have been empowered to develop “superintelligence,” or artificial intelligence that exceeds the human brain.

Instead, the job cuts were aimed at cleaning up the organizational bloat that resulted from three years of building up Meta’s A.I. efforts too quickly, two people with knowledge of the matter said. The layoffs aimed to help Meta develop A.I. products more rapidly, they said.

“By reducing the size of our team, fewer conversations will be required to make a decision, and each person will be more load-bearing and have more scope and impact,” Mr. Wang wrote in the memo circulated to employees.

The cuts, which were earlier reported by Axios, came at an intensely competitive time for Meta, which has spent the past three years dealing with the rapid onset of A.I. After ChatGPT burst onto the scene in 2022, OpenAI, Google and Microsoft hired furiously to build the next generation of A.I. chatbots and other products.

Meta, which owns Facebook, Instagram and WhatsApp, struggled to keep up with the pack. After early success developing its open-source A.I. model, called Llama, its progress stagnated. The company went on a fresh hiring spree and made strategic errors, leading to product development issues over the past 18 months.

After a rocky first half of this year, Mr. Zuckerberg moved to restart the A.I. efforts. In June, he invested $14.3 billion in ScaleAI, an artificial intelligence start-up that was co-founded by Mr. Wang. Mr. Zuckerberg then brought ScaleAI’s top talent to Meta’s Superintelligence Labs, including Mr. Wang.

Mr. Zuckerberg has since also spent billions recruiting top researchers from other A.I. labs and companies, including OpenAI, Google and Microsoft. Meta has dangled pay packages to some that number well into the hundreds of millions of dollars.

In August, Mr. Zuckerberg split Meta Superintelligence into four groups. One was called FAIR, which was focused on A.I. research; a second was working on superintelligence, another on products and a fourth on infrastructure, such as data centers and other A.I. hardware.

After that restructuring, employees in the FAIR division scrambled to join Mr. Wang’s team, two people with knowledge of the matter said. The core team that Mr. Wang leads is made up of outside hires from companies like OpenAI and Google, though he has more recently brought on a few dozen A.I. researchers from other parts of Meta with specific expertise, they said.

The planned cuts will affect employees at FAIR, the product division and the infrastructure group, according to Mr. Wang’s memo. Employees who were laid off received emails by 10 a.m. Eastern time, and the company plans to try to find other positions internally for those affected.

No cuts were made to TBD, the team building superintelligence and managing Meta’s large language models, which drive chatbots and other A.I. products, the people with knowledge of the situation said. The company is still hiring A.I. researchers in the TBD unit, which is managed by Mr. Wang, the people said.

Meta executives have emphasized that the cuts do not mean they are retrenching on A.I. efforts, and that superintelligence remains among Mr. Zuckerberg’s top priorities for the company.

In a sign of the escalating competition in A.I., Meta on Saturday also said it would cut off access to non-Meta chatbots like OpenAI’s ChatGPT on WhatsApp beginning next year. That means WhatsApp’s three billion users will no longer be able to use ChatGPT in the messaging app.

In a statement, a Meta spokeswoman said the company made the change because OpenAI and other companies were using the business messaging feature beyond the intended scope of customer service. OpenAI disputed Meta’s assertion.

On Wednesday, Kevin Weil, OpenAI’s vice president of science, posted on social media that it was “hard to believe Meta is shutting off 1-800-CHATGPT, which has many millions of happy users. If you’re one of them, you can migrate to our app, website, and browser to preserve your conversations.”

No worries but really two different threads. The impending financial meltdown vs AI affecting our lives with should/shouldn't subtexts

Thank you for posting. Agree but a part of my question which may not be answerable is determining CDS volume. Given this story, with more to come, will CDS volume go up? What is the tipping point? Will we see complicated made from whole cloth financial instruments like the synthetic CDOs of 2008 fame enter the arena? Someone in the private credit world must be modeling a way to socialize the losses and creating financial instruments to affect that outcome if things so south. Someone else has to be modeling that model on Wall Street. At least I hope so.

I think that’s going to precipitate the bubble.

If history is any guide, the answer to all of the above is yes. Volume will go up, dealers will respond to demand for protection by writing more swap contracts, dealers will also probably devise more complicated structured products around the AI space (think CDO’s, high leverage) and ultimately there will be another subprime type implosion. Wall St has a short memory - sooner or later they make the same mistakes all over again.

I know very little about this stuff but one thing I have seen is that at some point, AI power usage will be such a drain on the grid that bad stuff can happen. Am I off base on this?

Bad stuff is already happening. Take a look at all the extra turbines spinning up (illegally) for Musk's AI power station down in TN when there are already air quality issues. The ones that use the regular grid are forcing home owners rates to rise considerably.

The only company making loads of money off of this AI stuff is nVidia selling the hardware, though I see AMD is trying to get in on that.

Who are these people?

Alex I'll take intelligence agencies for a thousand. Second guess crypto.
CBDC is on it's way when the public will accept it.

Companies are quietly removing their 'carbon neutral' pledges.