Your writing is always an absolute joy, no matter the topic, but this one in particular slaps.
Thank you for giving these eyes (oh so weary from rolling to glare at the inside of my own skull rather than peruse any more vomitous paragraphs from phonically dessicated AI pieces) a reason to remember that great storytellers are still out there.
Also, I’m not a finance person, I don’t trade, I just have my spider-legs resting on many gossamer web-threads, listening through tiny hairs for vibrations that show the shape of the wind.
Slinging words, Gina and all, there’s still a place for ya! Alas when iPhones could suddenly take amazing photos all the photographers wondered. Is there no more a way for me to wend my way in the world and make my statement in a noisy over-pictured world…
I think after a bit of settling, we will all the more eagerly savor the carefully, humanly selected nuances of a delightful word string.
Do I detect hints of melon? Tobacco and a smooth finish, these grapes clearly grew on a southern slope where a precocious butterfly once landed, muses the connoisseur knowingly.
NVDA will be the first company to go from $5.7 trillion (or whatever it tops out at) to $0. The reason is that the only path to profitability for the industry is to reduce the power cost of compute Either by massive price rises from token-based billing, or they throttle throughput so the tokenmaxxers don't kill them, or they make the models way cheaper to run. Any way you slice it, it means those 70% growth rates in capacity aren't needed. They fall to 50%, then 30%, then 10%, then they're negative. And the hyperscalers and neoclouds that bought 100s of GW of their chips they haven't installed yet are going bankrupt and these chips are coming back on the market at 5 cents on the dollar. Now the annual chip sales are 5% of what they once were and they find the've stuffed the pipeline into the 22nd century. Plus they're contractually obligated to the neoclouds to buy $30 billion a year of capacity they don't need now. Even though they're all owned by the banks, a contract is a contract and they are debtors in possession of that contract. Sales have fallen 95% and that's in the $millions and not the $billions. No company could survive this. So they don't. They write the stock down to $0 and then emerge under new ownership selling about as many AI chips as they do video cards they were once known for. They let Jensen keep wearing the leather jacket.
I don't know about the $0 call, but like you said below too: they'll fall heavily - if they're ever allowed to fall that is. Because the stock can stay at $200 for example, while at the very same time the M2 is tenfold. Effectively: 90% reduction. Nominally: the same.
My more detailed take on the matter is with the links in the article.
I don't know what will prick the bubble, I just know it is one.
I don't know when it will happen, I just know it will.
I was there in 2000 and I can say this for a fact. Very few of the companies that went bust did so because they were run badly. They went bust because the bubble created way too many of them for any of them to be profitable. The only way to not burn infinite money was for 80-90% to go out of business so 10-20% could survive. And the survivors still fell 90-95% by 2003, but after that the pie started growing much faster and many fewer pieces were needed.
NVDA will still have some value down the road for whatever reason. Long-dead Memorex (“Is it live, or is it Memorex?”) was recently purchased by Finest Brands for an undisclosed sum to get the brand rights, its IP portfolio, and various licenses related to it.
"But here’s what the cheerleaders skip, because the song dies if they don’t. China’s not ahead. It’s good enough. And good enough is the colder one, way colder, because ahead you can chase and catch, good enough just quietly walks off with your business and never looks back."
This is the exact same formula that the CCP/Crony Capitalists employed to move almost the entirety of Western Industrial capacity to Chyna from the 1990s forward.
Chyna wasn't ahead, in fact they were far behind, but they were good enough & The Free West was cannibalized to bring them up-to-speed; Of course Crony Capitalists + investors profited wildly from P/E gains, among other *temporary* games.
"If you're paying $12, $13, $14 an hour for factory workers and you can move your factory South of the border, pay a dollar an hour for labor... have no health care... have no environmental controls, no pollution controls and no retirement... there will be a giant sucking sound going south." ~ Ross Perot 1992
A small, unexpected friction coefficient in a machine that needs to be perfectly lubed to sustain its own exponential acceleration. It's going to make wonders in how plans change.
In 10 days Github/CoPilot moves from "premium requests" a.k.a. number of prompts, to pay-per-token. Cost is going up ~10x. Profitability inches closer, but is probably not there yet.
No much time left before cracks open in the weakest gear. We'll see.
Thank you very much for your writing. My English is barely intermediate but today I really enjoyed your article; The content and the form. Even when I have not understood all the meanings. My native language is Spanish. Years ago I used to enjoy the rhythm and harmony of the Spanish writer Francisco Umbral. Today somehow you, your writing remind me of him. It has been a real pleasure, thank you very much!
That was quite the rant! Not being in the game anymore I haven't been paying attention to these story stocks, but just out of curiosity I took a look at NVDA's chart:
See anything interesting? Look at the volume since the breakout in early 2024. Where is it? Next draw the LT trend line from the 2022 low and tell me what you see. If this were a mining stock I'd be shorting it. $2.19 to $1.60. Moving the decimal point gives you some perspective, but given what it is, or pretends to be, and how so much is riding on that alone, I'd just give it a miss. Of course we all know that bears are always early, right?
"And now we have Mar 26 -> May 26 -> ?end of May revisit?"
Interesting. I wouldn't have called either of those a trend line. The way I was taught is that you need at least 3 points to establish a trend line, and in both cases I only see 2. What I also see, when looking a bit closer, is a trend line from Mar 25 up to around Sept. 25, followed by a breakdown into a weak H&S top. As I said, if this was a mining stock I'd be suspicious about the rise from that point on weak volume - like someone was trying to keep it from breaking down. That's the point at which shorts would jump in, which is why I projected 160, although depending on the progress being made by the co. I might place a buy at that level. So again as a mining stock, this chart is typical of a co. that got ahead of itself price wise and is looking to consolidate, allowing for other factors such as re-financing and drill results of course.
This is all speculative on my part of course. There are far too many variables in play with NVDA to make a reliable call, which is why I tended to avoid story stocks and trade companies that were easier to analyze, at least for me. I know most investors avoid junior explorers for several good reasons, but those reasons exist across all sectors really, they’re just easier to spot in the juniors.
It is said that a lot of their GPUs are gathering dust in warehouses. Waiting for datacenters to be build. Which are waiting for electricity infrastructure (both powerlines and stations) to be build. Which cannot wait for copper to be dug up (as prices are rising a lot)...
Holy Crap - reminds me of FCEL and BLDP in the day - and PLUG as well even though I never went there....talk about flux - and at some point after losing when it went down for good - look at the long-term charts I realized - no thanks casino. I ain't interested. You know most don't even have the ability to invest so if the whiz kids lose it all on a ponzi scheme I'll hardly shed a tear for them....hard lessons are sometimes the best ones.
I mean consider the stock "Gamestop" - what a freaking casino joke that is - enough to make one lost funds look up and proclaim - never again will I participate in that rigged place where jokers abound and losers are always at the ready.
An angle I had not considered much previously: NVDA et al’s role in a sudden onslaught of Ai-assisted breakthrough in Quantum computing and corresponding holes and zero day / compromised systems / encryption standards… all potentially creating a sudden slew of self-growing demand.
A sci fi moment to be sure, but no doubt a question of “when” not whether at this point.
Recent attention shows it may be sooner in years than I anticipated? But will it rescue the party? 🎊
The mechanism underneath the sag has a name in options pricing and it applies here. When the market prices perfection as the baseline, beat-and-raise becomes the minimum threshold for the stock to hold. Every perfect quarter gets absorbed instantly because the next perfect quarter is already priced. The stock can only move on surprise, and when your baseline is perfection the only available surprise is negative. Four flawless quarters and a red tape is what it looks like when a company has priced itself into a corner where the only remaining asymmetry is downside.
The Korea parallel is the part that makes this structural rather than anecdotal. leveraged retail holding two factories in a trench coat pretending to be an economy is the downstream consequence of the same narrative playing out one layer removed. The train sweeper borrowing 150% to hold memory chips is the last link in a supply chain that starts with Nvidia's guidance and ends with someone who has never read a 10-Q pricing complete collapse as an acceptable cost of admission.
I read this post about how all of the AI providers are raising their prices through token pricing. To me, this is another data point or a bit of reality creeping in about the sustainability of the AI bubble.
Yep. Next thing you'll see is that the free tiers will get severely limited. That's where they're losing money. NOT on the paying customers (well, mostly not that is).
Sadly, I'm late to this bar session ... but want to say just this, just once: this article is why I subscribed.
(No, I've got no money in Nvidia - and as for that cold thing in the cellar, I suggest people ponder, for a moment, the size of the Chinese population, working living, buying, selling ....)
Your writing is always an absolute joy, no matter the topic, but this one in particular slaps.
Thank you for giving these eyes (oh so weary from rolling to glare at the inside of my own skull rather than peruse any more vomitous paragraphs from phonically dessicated AI pieces) a reason to remember that great storytellers are still out there.
Also, I’m not a finance person, I don’t trade, I just have my spider-legs resting on many gossamer web-threads, listening through tiny hairs for vibrations that show the shape of the wind.
I’m glad yours is a thread I follow.
Just commenting to say how much I agree with you, and, yes, it is a joy, to read well written well thought out prose.
Thank you No 1
@Gina Chick your writing aint half bad either.
Here, here!
Slinging words, Gina and all, there’s still a place for ya! Alas when iPhones could suddenly take amazing photos all the photographers wondered. Is there no more a way for me to wend my way in the world and make my statement in a noisy over-pictured world…
I think after a bit of settling, we will all the more eagerly savor the carefully, humanly selected nuances of a delightful word string.
Do I detect hints of melon? Tobacco and a smooth finish, these grapes clearly grew on a southern slope where a precocious butterfly once landed, muses the connoisseur knowingly.
Kudos, No1 and friends.
NVDA will be the first company to go from $5.7 trillion (or whatever it tops out at) to $0. The reason is that the only path to profitability for the industry is to reduce the power cost of compute Either by massive price rises from token-based billing, or they throttle throughput so the tokenmaxxers don't kill them, or they make the models way cheaper to run. Any way you slice it, it means those 70% growth rates in capacity aren't needed. They fall to 50%, then 30%, then 10%, then they're negative. And the hyperscalers and neoclouds that bought 100s of GW of their chips they haven't installed yet are going bankrupt and these chips are coming back on the market at 5 cents on the dollar. Now the annual chip sales are 5% of what they once were and they find the've stuffed the pipeline into the 22nd century. Plus they're contractually obligated to the neoclouds to buy $30 billion a year of capacity they don't need now. Even though they're all owned by the banks, a contract is a contract and they are debtors in possession of that contract. Sales have fallen 95% and that's in the $millions and not the $billions. No company could survive this. So they don't. They write the stock down to $0 and then emerge under new ownership selling about as many AI chips as they do video cards they were once known for. They let Jensen keep wearing the leather jacket.
I don't know about the $0 call, but like you said below too: they'll fall heavily - if they're ever allowed to fall that is. Because the stock can stay at $200 for example, while at the very same time the M2 is tenfold. Effectively: 90% reduction. Nominally: the same.
My more detailed take on the matter is with the links in the article.
I don't know what will prick the bubble, I just know it is one.
I don't know when it will happen, I just know it will.
I was there in 2000 and I can say this for a fact. Very few of the companies that went bust did so because they were run badly. They went bust because the bubble created way too many of them for any of them to be profitable. The only way to not burn infinite money was for 80-90% to go out of business so 10-20% could survive. And the survivors still fell 90-95% by 2003, but after that the pie started growing much faster and many fewer pieces were needed.
I was there too...
The lenders stopped lending the orders stopped coming in commited orders were cancelled, and the swirl began.
What if it's "too big to fail" and gets bailed out?
My money is on that after the inevitable "correction"
NVDA will still have some value down the road for whatever reason. Long-dead Memorex (“Is it live, or is it Memorex?”) was recently purchased by Finest Brands for an undisclosed sum to get the brand rights, its IP portfolio, and various licenses related to it.
Brilliant commentary No1
"But here’s what the cheerleaders skip, because the song dies if they don’t. China’s not ahead. It’s good enough. And good enough is the colder one, way colder, because ahead you can chase and catch, good enough just quietly walks off with your business and never looks back."
This is the exact same formula that the CCP/Crony Capitalists employed to move almost the entirety of Western Industrial capacity to Chyna from the 1990s forward.
Chyna wasn't ahead, in fact they were far behind, but they were good enough & The Free West was cannibalized to bring them up-to-speed; Of course Crony Capitalists + investors profited wildly from P/E gains, among other *temporary* games.
"If you're paying $12, $13, $14 an hour for factory workers and you can move your factory South of the border, pay a dollar an hour for labor... have no health care... have no environmental controls, no pollution controls and no retirement... there will be a giant sucking sound going south." ~ Ross Perot 1992
I think I first put that idea forth in https://no01.substack.com/p/the-chip-that-pops-the-ai-bubble. (I've held it longer, but never publicly expressed it anywhere -- I just kept my mouth shut until early last year)
Wonder if that was a warning from Ross or he was reciting a playbook.
A small, unexpected friction coefficient in a machine that needs to be perfectly lubed to sustain its own exponential acceleration. It's going to make wonders in how plans change.
In 10 days Github/CoPilot moves from "premium requests" a.k.a. number of prompts, to pay-per-token. Cost is going up ~10x. Profitability inches closer, but is probably not there yet.
No much time left before cracks open in the weakest gear. We'll see.
Dear NO1,
Thank you very much for your writing. My English is barely intermediate but today I really enjoyed your article; The content and the form. Even when I have not understood all the meanings. My native language is Spanish. Years ago I used to enjoy the rhythm and harmony of the Spanish writer Francisco Umbral. Today somehow you, your writing remind me of him. It has been a real pleasure, thank you very much!
That was quite the rant! Not being in the game anymore I haven't been paying attention to these story stocks, but just out of curiosity I took a look at NVDA's chart:
https://www.barchart.com/stocks/quotes/NVDA/interactive-chart
See anything interesting? Look at the volume since the breakout in early 2024. Where is it? Next draw the LT trend line from the 2022 low and tell me what you see. If this were a mining stock I'd be shorting it. $2.19 to $1.60. Moving the decimal point gives you some perspective, but given what it is, or pretends to be, and how so much is riding on that alone, I'd just give it a miss. Of course we all know that bears are always early, right?
1.6 is quite aggressive. Remember that a mania can stay on longer than you can stay solvent...
However, the rate of change is accelerating.
We have the one trend you mentioned: Jan 2023 -> Nov 23 -> Apr 25. That's one.
We have the Apr 25 -> Mar 26 one
And now we have Mar 26 -> May 26 -> ?end of May revisit?
And like you said: volume is diminishing over the past 3 to 4 years.
"We have the Apr 25 -> Mar 26 one"
"And now we have Mar 26 -> May 26 -> ?end of May revisit?"
Interesting. I wouldn't have called either of those a trend line. The way I was taught is that you need at least 3 points to establish a trend line, and in both cases I only see 2. What I also see, when looking a bit closer, is a trend line from Mar 25 up to around Sept. 25, followed by a breakdown into a weak H&S top. As I said, if this was a mining stock I'd be suspicious about the rise from that point on weak volume - like someone was trying to keep it from breaking down. That's the point at which shorts would jump in, which is why I projected 160, although depending on the progress being made by the co. I might place a buy at that level. So again as a mining stock, this chart is typical of a co. that got ahead of itself price wise and is looking to consolidate, allowing for other factors such as re-financing and drill results of course.
This is all speculative on my part of course. There are far too many variables in play with NVDA to make a reliable call, which is why I tended to avoid story stocks and trade companies that were easier to analyze, at least for me. I know most investors avoid junior explorers for several good reasons, but those reasons exist across all sectors really, they’re just easier to spot in the juniors.
I expected all their GPUs purchased for unlit data centers be what was in the cellar…
It is said that a lot of their GPUs are gathering dust in warehouses. Waiting for datacenters to be build. Which are waiting for electricity infrastructure (both powerlines and stations) to be build. Which cannot wait for copper to be dug up (as prices are rising a lot)...
I brought you three-fingers of 10yr Whistle Pig Rye, neat of course.
I put a straw in the bottle, the rest is for me.
An entertaining voice. ;-)
Thanks for pointing out the Achilles heel, sword of Damocles and the bar-tab of Nvidia.
Holy Crap - reminds me of FCEL and BLDP in the day - and PLUG as well even though I never went there....talk about flux - and at some point after losing when it went down for good - look at the long-term charts I realized - no thanks casino. I ain't interested. You know most don't even have the ability to invest so if the whiz kids lose it all on a ponzi scheme I'll hardly shed a tear for them....hard lessons are sometimes the best ones.
I mean consider the stock "Gamestop" - what a freaking casino joke that is - enough to make one lost funds look up and proclaim - never again will I participate in that rigged place where jokers abound and losers are always at the ready.
An angle I had not considered much previously: NVDA et al’s role in a sudden onslaught of Ai-assisted breakthrough in Quantum computing and corresponding holes and zero day / compromised systems / encryption standards… all potentially creating a sudden slew of self-growing demand.
A sci fi moment to be sure, but no doubt a question of “when” not whether at this point.
Recent attention shows it may be sooner in years than I anticipated? But will it rescue the party? 🎊
The mechanism underneath the sag has a name in options pricing and it applies here. When the market prices perfection as the baseline, beat-and-raise becomes the minimum threshold for the stock to hold. Every perfect quarter gets absorbed instantly because the next perfect quarter is already priced. The stock can only move on surprise, and when your baseline is perfection the only available surprise is negative. Four flawless quarters and a red tape is what it looks like when a company has priced itself into a corner where the only remaining asymmetry is downside.
The Korea parallel is the part that makes this structural rather than anecdotal. leveraged retail holding two factories in a trench coat pretending to be an economy is the downstream consequence of the same narrative playing out one layer removed. The train sweeper borrowing 150% to hold memory chips is the last link in a supply chain that starts with Nvidia's guidance and ends with someone who has never read a 10-Q pricing complete collapse as an acceptable cost of admission.
https://xcancel.com/HedgieMarkets/status/2057531661785628841#m
I read this post about how all of the AI providers are raising their prices through token pricing. To me, this is another data point or a bit of reality creeping in about the sustainability of the AI bubble.
Yep. Next thing you'll see is that the free tiers will get severely limited. That's where they're losing money. NOT on the paying customers (well, mostly not that is).
Sadly, I'm late to this bar session ... but want to say just this, just once: this article is why I subscribed.
(No, I've got no money in Nvidia - and as for that cold thing in the cellar, I suggest people ponder, for a moment, the size of the Chinese population, working living, buying, selling ....)
ai slop