Nothing written herein should be construed as financial or medical advice.

Monday, November 16, 2026

The bubble is bursting

(Source: GreatFactsNow)



From a Mastodon thread by Gabriele Svelto, who describes himself as:

Old school nerd
Knows things about computers that would drive you insane
Hacks on Firefox at Mozilla




The news that there's an oversupply of Nvidia GPUs convinced me that the AI bubble is bursting and it'll burst *hard*. So here's a thread where I go over the financial, technological and hardware implications of this. Strap in because this is going to be long and probably a bit rambling too.

Let's start with the oversupply and what that means for everybody involved. Whether it's traditional players like Google or new entries like CoreWeave every single company in the tech sector has been buying an increasing amount of hardware at increasingly steep prices. GPUs, memory, servers, power supplies, and the prices went up and up and up. Now when a company does that, it looks neutral on their balance sheet, because the money they spend is compensated by now having an asset of the same value. Then over the following years they write down that value in what is called depreciation, this is visible as an expense on their balance.

If Nvidia GPUs are in significant oversupply as was reported then it means the prices will come down, as there's nobody willing to buy that stuff anymore. And this poses an immediate and very tangible problem for those who bought them: they have to write down the loss of value for the stuff they already own. That's hundreds of billions of dollars of value evaporating, because that value only existed if these things were really scarce, which they aren't. Oops.

Which brings to another very thorny topic: mark-to-market accounting. Let's say you buy something worth 1 billion $, say DRAM, that's 1 billion $ of assets on your balance. But let's say that the next year the price of that stuff goes up by 100%, what are you going to do? The prudent choice is to leave that asset at the value you paid it for, in case it comes down later. And then there's the choice of changing the value of that asset to account for the increase in price, that's mark-to-market accounting. Enron became famous for this, I guess you can see why.

I don't know if hyperscalers did it, but given everybody involved in the AI bubble has been reckless and dishonest my guess is that it's highly probable they did. It's also considered a generally accepted accounting practice (GAAP) these days. Anyway, if they did it, it means that when the prices of this mountain of hardware come down the companies owning it will have to do gigantic write-downs on their balance sheets, booking huge losses, much larger than if they hadn't.

But wait, it's worse than this because none of these companies had the money needed to buy this hardware. They either used venture capitals or took out loans. Hundreds of billions of loans. And what does a company use as the collateral for a loan? Its own assets, right? Like GPUs. And memory. And the datacenters they're in. Which all of a sudden are losing value. And what do you call a loan whose collateral is insufficient to cover it? An underwater loan. And what does it cause? A margin call: the debtor must either deposit cash or provide a different collateral to compensate. I guess you already know where this is going.

Every single company in the tech space has negative cash flow. That is they're spending more money than they're making. A lot of it. So how 'll they answer the margin call? If this is starting to sound a lot like the subprime mortgage crisis it's because it very much is, but with datacenters instead of houses.

This is where we have to thank Ed Zitron for his excellent reporting on the finances of these companies and his remarkable work in connecting the dots whenever some CEO talked bullshit to investors while the rest of the press was looking the other way. This is also why the Gamers Nexus investigations about GPU smuggling were so important: the demand could not slow down, by all means necessary, so a little illegal smuggling on the side is OK, say a few tens of billions of it.

And this is when we start talking about the tech itself and why it is unsustainable. Every single company in this space is losing money, every AI startup, every AI division in an existing company, everybody. They're losing money because the costs of running an LLM - inference - are large, and it's unclear if anybody has been profitable there. But even if they are, nobody has been able to cover the escalating costs of training LLMs, which is of course a prerequisite to running them.

Training LLMs requires a gigantic amount of hardware, brute force calculations, storage, memory and power. Enormous amounts of data whose scraping is not being paid by the companies involved, it's a cost they've spread over the entire internet. As models get larger training gets more expensive. But even as companies sink more money on training LLMs have already reached a plateau in terms of functionality, and in some cases are degrading in a process called model collapse.

Let's talk statistics for a minute to understand model collapse. Say you have a bunch of measurements of something, and you want to build a mathematical function that will tell you what values appear in the measurements and how often they appear, this is what is called a distribution. A distribution can be used to reproduce the material it was created from, but the output will not be the same because something was lost in the process of making it. The less frequent samples, the stuff that lives at the fringes of your data set, the details.

Now let's say you use your distribution to generate a new bunch of data: it will look a lot like the inputs you used to make it, but it won't be the same. And if you try to make a new distribution out of that data it will be distorted. You'll lose more information and for every iteration the outputs will look different, they will be more uniform, they'll contain fewer details, like when you recompress a JPEG file over and over and over.

And this is exactly what happens when you train an LLM on its outputs, because an LLM is at its core a distribution of human language. It's also the reason why LLM vendors started putting invisible watermarks in LLM outputs. They need to filter out the slop they make from their training set. But even then, some slop will make its way in, and even the slop that gets filtered out is still an additional cost: you scraped it, you processed it to check for your watermark, and then you threw it away.

As the internet gets sloppified the chances of keeping slop out diminish, and the training costs increase. And don't forget that all this training has been done without consent, attribution or compensation of the people making the training materials. This space will be regulated eventually, and when it is, that will also make the training costs increase further.

Which leads us to the hardware itself: this is the point where people will tell you that right now things are looking bad, but the prices will come down because the hardware will improve, and so you'll be able to make powerful local models and whatnot. But here's the thing: hardware improvements are overin fact, they've been for a while. Moore's Law, Dennard scaling, every single factor which led to faster and cheaper hardware has ceased, years ago.

There's no magic hardware improvement around the corner, quite the contrary. Every new silicon node comes with smaller and smaller increases in density and performance over the previous one, but at significant increase in costs. A fabrication facility to produce state-of-the-art hardware takes months to go from a silicon wafer to a working chip, using machines that cost billions.

In other words the only way for a modern high-performance silicon process to be profitable is to produce a large amount of chips that will be sold at high price. See how we're coming back to the artificial scarcity where this all started? To keep the semiconductor market going we need more demand and AI represented the largest demand of silicon ever, but none of that demand was real. So when the prices crash the demand evaporates and the foundries will be left with a lot of idling capacity.

This will cause a huge amount of capital that was poured into the tech sector to shift elsewhere and this is a good thing. Climate change is weighing down on us and on our future as a species, all the resources we have should be concentrated in fighting it, not chasing a technological fad inflated by a bunch of scammers and mega-rich buffoons having mid-life crises on commercial social media.

I'll wrap this up with an appeal to the people who have been using LLMs extensively in their work, often becoming dependent on them: this is the time to come off this train before it's too late, though it might already be. Nobody is making any money with these things, but they made plenty of debts, so you'll be left holding the bag for the entire tech sector since you're the only ones paying for them. Over and out.  [Lightly edited]


Anthropic IPO

By Jeremy Banx 



Sunday, November 15, 2026

When will emissions peak?

Emissions from land transport have probably already peaked because of the jump in EV/PHEV sales.

Let's have a look at emissions in electricity generation.  Together, these make up roughly 50% of emissions.

The chart below shows the percentage of electricity demand that is provided by wind and solar.   The chart comes from EMBER's electricity data explorer.



It is an exponential curve, i.e., it rises more and more steeply over time.  So what does the increase in wind and solar look like every year?  You can see that in the chart below.  It is also an exponential curve, but more variable than the chart of the percentage levels, above, as you would expect from a first difference.

In 2025, the percentage of wind and solar in total electricity generation rose 2.24%. Since 2021, global electricity demand has grown by ~ 3% per annum.  If wind and solar together increase their percentage of total electricity production by more than 3%, then the output from other sources of electricity must by definition decline. 

 So, how close are we to that point, where wind and solar rise by more than 3% per annum?  The percentage share of W&S has risen by 14% per annum, from 10.3% in 2021 to 17.3% in 2025. Ignoring year-to-year fluctuations, at this growth rate, 17.3% will rise to 19.7% (or by 2.4 percentage points) in 2026, and from 19.7% to 22.5% in 2027 (= plus 2.8 percentage points), and 25.7% in 2028 (= plus 3.2 percentage points). In 2028, the rise (in percentage points) will be 3.2, and emissions from electricity generation will have peaked.  Or will they?

Of course, it's not that simple.  Electricity demand will accelerate, as EVs roll out and data centres get built.  For what it's worth, I'm certain that the AI bubble is popping right now, and electricity demand for AI will collapse.  But I could be wrong.

What about EVs?  So far, all EVs on the road have increased electricity demand by 236 TWh per year, compared to total annual electricity demand of 31,925 TWh. Electricity demand is rising by ~1000 TWh each year on average.  Since there are currently roughly 1,000 million cars on the world's roads, when EVs have completely replaced all petrol cars, in, say 20 years, electricity needed for EVs will rise from 236 TWh to 2600 TWh, or about 8% of current total electricity demand.  Per year, it'll be less; for example in 2027, I estimate a rise in EV electricity demand of 47 TWh, pushing up that average of 3% growth in total demand to 3.1%.   (Don't forget that all these EV batteries will make stabilising the grid much easier, so having a higher percentage of solar becomes more feasible.)  That still means that emissions from electricity are likely to peak in 2028, but will rise by 0.3% in 2027.

Let's put it together.  

Emissions from burning oil to fuel cars and light trucks have prolly peaked.  Emissions from oil totalled 12.6 billion tonnes in 2025.  So far, EVs have reduced oil demand (and therefore oil emissions) by 0.4%; and that figure is growing by 0.2 percentage points per annum, and it will accelerate. So emissions from oil will fall by 0.2% per annum, and more each year thereafter.

Coal produced 15.7 billion tonnes of CO2 emissions in 2025.  On my numbers, emissions from coal are still rising, just, ignoring the likely global recession next year, which I think quite likely, but won't include in my numbers.

In 2027, the fall in emissions from oil will more or less offset the rise from coal.  And that will be the peak, though emissions will not plunge immediately.  It will be a rolling top, a levelling off, following by small declines which will however accelerate.  

We need to cut emissions by 3% a year, and that's not going to happen for a couple of years.  Yet we can see the turning point just ahead.  Now it is a matter of bending the curve to achieve a peak sooner and a more rapid decline thereafter.  And that is doable.

I have concentrated on emissions from electricity and road transport, because these are the easy sectors to de-carbonise; we are at or close to the peak; it seems plausible that by 2050 we will have de-carbonised both; which in turn translates to a 3% per annum compound rate of decline in emissions.  That leaves steel, cement, air travel, sea transport, and of course, the biggie, agriculture. We need to act to cut emissions in all these sectors, too.  

But there is hope.  Half of all global emissions will peak in 18 months or so; emissions in China are peaking; the US (the second-largest emitter) may be pulling its ox wagon out of the weeds and rejoining civilisation; and with electricity going green, other sectors can be de-carbonised too.





EVs exceed 31% of global car sales

 Plug-in hybrids and full EVs made up 31.3% of global car sales in September, according to EMBER's EV sales data tracker, up from 29.5% in September 2025.  EVs made up 23%, PHEVs 8.3%.    The number of EVs & PHEVs on the road has increased by 20.4% since January 2026.  

The S-curve continues to flex up. And remember: even if oil prices fall because of an end to the Iran War, EVs will still be much cheaper to run than petrol cars.  When nearly a third of new car sales are EVs, the psychological barriers to sales of EVs/PHEVs fall away.   

Source: EMBER


Birds of a feather

 



Be kind

 By Jesse Duquette

Happy 104th birthday to one of my favorite writers of all time, Kurt Vonnegut Jr.




Compromise

 By Jesse Duquette