@sgued mark to market accounting has been part of GAAP for a long time, it's perfectly legal. But it can be abused which makes it dangerous. I don't know if they're actually doing it with RAM, but there's been some very creative accounting in the AI sector, so IMHO it's a possibility.
gabrielesvelto@mas.to
Posts
-
I was thinking about the skyrocketing memory prices and something dawned on me: could it be a way to prop up AI investments by artificially covering the huge loans that have been taken out to fund them? -
I was thinking about the skyrocketing memory prices and something dawned on me: could it be a way to prop up AI investments by artificially covering the huge loans that have been taken out to fund them?Oh, and BTW, in general a company shouldn't adjust the value of their depreciating assets upwards, even if the market value of those goes up, because it's dangerous to do so. But companies do it all the time since the '90s, it's called "mark to market" accounting. And guess what company became notorious for doing it to artificially inflate its assets? Enron. Enough said. 6/6
-
I was thinking about the skyrocketing memory prices and something dawned on me: could it be a way to prop up AI investments by artificially covering the huge loans that have been taken out to fund them?Of course I can't prove a cause-and-effect relationship here, but the timing is awfully convenient for the tech market. There's no downside in the short term, but in the medium term this is all smoke and mirrors. At these prices demand for memory will crater, inventories will climb and at some point the prices will have to come down, and any financial bet that has been made on them will unravel. 5/6
-
I was thinking about the skyrocketing memory prices and something dawned on me: could it be a way to prop up AI investments by artificially covering the huge loans that have been taken out to fund them?And these datacenters are packed with RAM too. And RAM prices went up by 5 times in a matter of weeks, against a market that was predicted to *decline* in volume this year. That is awfully convenient if you have a lot of this stuff on your balance sheets: it's an asset. Even more convenient when at the same time you've taken on a lot of debt, a liability. 4/6
-
I was thinking about the skyrocketing memory prices and something dawned on me: could it be a way to prop up AI investments by artificially covering the huge loans that have been taken out to fund them?When was the last time you bought a piece of gear and the price went up? Never. After a year the value of a new graphics card is already measurably less than when you bought it. And these companies are amortizing the value of these things over six years which is a very long time. They'll be worth nothing in less than three. Unless the price of new hardware goes up a lot. Like multiple times. Like what just happened with memory. 3/6
-
I was thinking about the skyrocketing memory prices and something dawned on me: could it be a way to prop up AI investments by artificially covering the huge loans that have been taken out to fund them?As you know several companies have taken on a gigantic amount of debt to build datacenters - Oracle is a good example. This debt is in the form of loans where the collateral is the datacenter itself, or rather what's inside of it. A building can't be worth billions unless it's packed with hardware that costs billions, so the actual collateral is the hardware... But hardware depreciates *fast* so it's a terrible collateral. 2/6
-
I was thinking about the skyrocketing memory prices and something dawned on me: could it be a way to prop up AI investments by artificially covering the huge loans that have been taken out to fund them?I was thinking about the skyrocketing memory prices and something dawned on me: could it be a way to prop up AI investments by artificially covering the huge loans that have been taken out to fund them? Could it be a way to artificially inflate the value of datacenters these companies have on their balance sheets? I don't want to sound paranoid but bear with me for a second. 🧵 1/6
-
A new argument that is cropping up among LLM apologists is that when things go wrong it's not because of the tools but it's because of the people.A new argument that is cropping up among LLM apologists is that when things go wrong it's not because of the tools but it's because of the people.
That's literally the core argument of pro-gun rights advocates in the US, the country with the highest number of gun-related deaths in the world.
If the tool's primary use is to cause other people harm then it's 100% the tool, not the people.