Thank you for this discussion. I’ve no background in IT/AI but your presentation was generally quite understandable. I also like that you gave a seemingly balanced view of the many political issues. I’m so glad we have Substack as a resource for exposure to information and critical analysis.
I don’t either. Some Investors suspected it was overvalued and used DeepSeek as an excuse to get out. I think it is premature to react to threats about tariffs on Taiwan since so many Silicon Valley people who backed Trump will try to dissuade him from doing so. Plus those tariffs need approval from Congress, who represent areas with the tech companies who rely on those chips.
Nvidia will continue to grow. Wall Street seems to be trembling at every movement of a leaf in the tea. 🍵
beyond laughable to think you'll outcompete a nation with 5x the people by withholding a few chip designs. that shit is just so fucking dumb.
>elect a retarded president while the CCCP remains competent as ever
fighting the good fight america! this is the way to victory!
nothing else matters in comparison lmao.
you want to "beat" china? or whatever?
shoot the fucker. otherwise, shut up.
they're winning because they aren't stupid. low fucking bar, but we sure can limbo.
e: i didn't vote this year, personally. tards vs sickos, lol. it might've turned out better this way, though - tards are more easily wrangled. like lenny!
If they lack people with the expertise to replicate those advanced chip designs and they lack the manufacturing facilities to create them, having 5x as many people doesn't help. Indonesia has roughly the same population as the USA and they don't compete on par with Silicon Valley. India has a larger population than China but an arguably inferior chip and AI sector. Of course, China probably does have the expertise necessary to replicate the advances of other countries, given time and money. But the exact future remains uncertain
It affected the price because it was older Nvidia hardware. That makes their newest offerings look potentially overpriced/overvalued.
Also, OpenAI was charging $200 month for their latest model, and this Chinese company only wanted a tiny fraction of that. That absolutely calls into question how much these AI companies are going to be able to charge and earn.
I mean, I don't know (and neither does anyone), but it seems quite plausible that DeepSeek affected Nvidia prices. If the price runs up because everyone thinks you need the latest Nvidia hardware in vast quantities and then someone comes along and proves you can do it with a limited number of functionally-older hardware, then that price run up was not justified.
I'm not sure that rational explanations for stock market moves are all that rational, but I appreciate the skepticism that anybody really knows what's going on. I suppose the case for thinking that DeepSeek’s success IS bad news for Nvidia is less about Nvidia's chips and more about the capacity of its biggest customers to continue to pay so much for them. If it takes so little time and resources to match the biggest foundation models, then how does that change the advantage Nvidia has enjoyed lately?
I learned from Matt Levine's newsletter that Liang Wenfeng manages a hedge fund in his spare time, which is a fun fact that Levine makes the most of.
I don't think the chip embargo can explain the crash. If it did, then we wouldn't have seen eg selloffs for power companies that planned to supply electricity to AI data centers. The dumb explanation (no one noticed that DeepSeek existed until a couple days ago) still seems strongly favored.
It's not obvious to me how the stocks of power companies would be impacted by a 100 percent tariff on Taiwanese chips. Data centers are a fairly small share of power company revenues, and power company rates are regulated in ways that don't necessarily make additional demand all that profitable for them. And also you could tell a story where DeepSeek's efficiency innovations reduce projected demand for electric power. So I'm not sure if the lack of movement of power company stocks tells us anything interesting.
No, what I mean is that a bunch of power company stocks actually declined yesterday because everyone decided that deepseek meant US AI was over. I just don't think there's a non-dumb explanation that fits.
Datacentres are currently 3% of the world's global power usage, which is enormous. It's predicted to rise to 10% by 2030 solely on the back of A.I. investment.
If something were to change these already announced plans, like suddenly not needing all those datacentres, it'd affect power companies almost as much as the companies building the centres themselves. It'd affect the 80 new gas-fired power stations currently planned in the U.S. to cater to future data centre needs for instance, and potentially the tech start-ups currently building tiny modular nuclear reactors specifically for powering data centres too (first of which is due to switch on in 2030).
Additionally yeah, power companies are extremely regulated... that effects what they can do to manipulate the market and how much they can charge consumers. But they are still primary producers - so long as the cost of production is less than the regulated sell price, more demand = more revenue, so long as you have capacity to supply it.
The power and water use of datacentres globally right now (and predicted to skyrocket) is quite underestimated.
Datacentres are currently 3% of the world's global power usage, which is enormous. It's predicted to rise to 10% by 2030 solely on the back of A.I. investment.
If something were to change these already announced plans, like suddenly not needing all those datacentres, it'd affect power companies almost as much as the companies building the centres themselves. It'd affect the 80 new gas-fired power stations currently planned in the U.S. to cater to future data centre needs for instance, and potentially the tech start-ups currently building tiny modular nuclear reactors specifically for powering data centres too (first of which is due to switch on in 2030).
The power and water use of datacentres globally right now (and predicted to skyrocket) is quite underestimated.
This is honestly very sloppy reporting. Just look at nuclear energy stocks dropping 30% and you can see that deepseek is what caused the market crash, not TSMC tariffs. Larping as a stock analyst to explain a market crash when that's clearly not your forte is a bad idea.
Taken from a friends post ‘𝗡𝘃𝗶𝗱𝗶𝗮'𝘀 sharp decline … (along with the broader market selloff), was 𝗻𝗼𝘁 𝗰𝗮𝘂𝘀𝗲𝗱 by #Deepseek. Instead, it was driven by a combination of significant other factors of which, fair enough... deepseek was a minor factor.
𝗠𝗮𝗰𝗿𝗼𝗲𝗰𝗼𝗻𝗼𝗺𝗶𝗰 𝗙𝗮𝗰𝘁𝗼𝗿𝘀 - the Yen carry trade unwind, (triggered by Japan's interest rate hike). This forced investors to pull back from risk-heavy positions where #Nvidia is a key target for these liquidity-driven trades hence why it was hit hard. This selloff then smashed through the broader tech sector and even spilled over into crypto. Also, 𝗚𝗲𝗼𝗽𝗼𝗹𝗶𝘁𝗶𝗰𝗮𝗹 𝗧𝗲𝗻𝘀𝗶𝗼𝗻𝘀 as Trump’s proposed to impose tariffs on semiconductors imported from Taiwan added another layer of complexity bevause Taiwan Semiconductor Manufacturing Company (TSMC) is critical to the supply chains of U.S. tech giants like Nvidia and Broadcom.’
people were looking for a reason that Nvidia stock was going to crash, and there was an initial impression that deepseek didn't use Nvidia. a lot of people are just looking at it like it's a money printer, while simultaneously being aware that money printers aren't actually a thing.
Thank you for this discussion. I’ve no background in IT/AI but your presentation was generally quite understandable. I also like that you gave a seemingly balanced view of the many political issues. I’m so glad we have Substack as a resource for exposure to information and critical analysis.
I don’t either. Some Investors suspected it was overvalued and used DeepSeek as an excuse to get out. I think it is premature to react to threats about tariffs on Taiwan since so many Silicon Valley people who backed Trump will try to dissuade him from doing so. Plus those tariffs need approval from Congress, who represent areas with the tech companies who rely on those chips.
Nvidia will continue to grow. Wall Street seems to be trembling at every movement of a leaf in the tea. 🍵
beyond laughable to think you'll outcompete a nation with 5x the people by withholding a few chip designs. that shit is just so fucking dumb.
>elect a retarded president while the CCCP remains competent as ever
fighting the good fight america! this is the way to victory!
nothing else matters in comparison lmao.
you want to "beat" china? or whatever?
shoot the fucker. otherwise, shut up.
they're winning because they aren't stupid. low fucking bar, but we sure can limbo.
e: i didn't vote this year, personally. tards vs sickos, lol. it might've turned out better this way, though - tards are more easily wrangled. like lenny!
The CCP is hardly competent.
You elected Trump then.
Blame yourself.
If they lack people with the expertise to replicate those advanced chip designs and they lack the manufacturing facilities to create them, having 5x as many people doesn't help. Indonesia has roughly the same population as the USA and they don't compete on par with Silicon Valley. India has a larger population than China but an arguably inferior chip and AI sector. Of course, China probably does have the expertise necessary to replicate the advances of other countries, given time and money. But the exact future remains uncertain
It affected the price because it was older Nvidia hardware. That makes their newest offerings look potentially overpriced/overvalued.
Also, OpenAI was charging $200 month for their latest model, and this Chinese company only wanted a tiny fraction of that. That absolutely calls into question how much these AI companies are going to be able to charge and earn.
I mean, I don't know (and neither does anyone), but it seems quite plausible that DeepSeek affected Nvidia prices. If the price runs up because everyone thinks you need the latest Nvidia hardware in vast quantities and then someone comes along and proves you can do it with a limited number of functionally-older hardware, then that price run up was not justified.
I'm not sure that rational explanations for stock market moves are all that rational, but I appreciate the skepticism that anybody really knows what's going on. I suppose the case for thinking that DeepSeek’s success IS bad news for Nvidia is less about Nvidia's chips and more about the capacity of its biggest customers to continue to pay so much for them. If it takes so little time and resources to match the biggest foundation models, then how does that change the advantage Nvidia has enjoyed lately?
I learned from Matt Levine's newsletter that Liang Wenfeng manages a hedge fund in his spare time, which is a fun fact that Levine makes the most of.
It's starting to sound like every other Chinese product, it's just a knockoff. Microsoft thinks they just stole OpenAI's data to train it.
I don't think the chip embargo can explain the crash. If it did, then we wouldn't have seen eg selloffs for power companies that planned to supply electricity to AI data centers. The dumb explanation (no one noticed that DeepSeek existed until a couple days ago) still seems strongly favored.
It's not obvious to me how the stocks of power companies would be impacted by a 100 percent tariff on Taiwanese chips. Data centers are a fairly small share of power company revenues, and power company rates are regulated in ways that don't necessarily make additional demand all that profitable for them. And also you could tell a story where DeepSeek's efficiency innovations reduce projected demand for electric power. So I'm not sure if the lack of movement of power company stocks tells us anything interesting.
No, what I mean is that a bunch of power company stocks actually declined yesterday because everyone decided that deepseek meant US AI was over. I just don't think there's a non-dumb explanation that fits.
Datacentres are currently 3% of the world's global power usage, which is enormous. It's predicted to rise to 10% by 2030 solely on the back of A.I. investment.
If something were to change these already announced plans, like suddenly not needing all those datacentres, it'd affect power companies almost as much as the companies building the centres themselves. It'd affect the 80 new gas-fired power stations currently planned in the U.S. to cater to future data centre needs for instance, and potentially the tech start-ups currently building tiny modular nuclear reactors specifically for powering data centres too (first of which is due to switch on in 2030).
Additionally yeah, power companies are extremely regulated... that effects what they can do to manipulate the market and how much they can charge consumers. But they are still primary producers - so long as the cost of production is less than the regulated sell price, more demand = more revenue, so long as you have capacity to supply it.
The power and water use of datacentres globally right now (and predicted to skyrocket) is quite underestimated.
Because AI and the proposed data centers use a massive amount of electricity. It makes perfect sense.
Datacentres are currently 3% of the world's global power usage, which is enormous. It's predicted to rise to 10% by 2030 solely on the back of A.I. investment.
If something were to change these already announced plans, like suddenly not needing all those datacentres, it'd affect power companies almost as much as the companies building the centres themselves. It'd affect the 80 new gas-fired power stations currently planned in the U.S. to cater to future data centre needs for instance, and potentially the tech start-ups currently building tiny modular nuclear reactors specifically for powering data centres too (first of which is due to switch on in 2030).
The power and water use of datacentres globally right now (and predicted to skyrocket) is quite underestimated.
Typo: “because depend for computing”
Will fix thanks!
This is honestly very sloppy reporting. Just look at nuclear energy stocks dropping 30% and you can see that deepseek is what caused the market crash, not TSMC tariffs. Larping as a stock analyst to explain a market crash when that's clearly not your forte is a bad idea.
Taken from a friends post ‘𝗡𝘃𝗶𝗱𝗶𝗮'𝘀 sharp decline … (along with the broader market selloff), was 𝗻𝗼𝘁 𝗰𝗮𝘂𝘀𝗲𝗱 by #Deepseek. Instead, it was driven by a combination of significant other factors of which, fair enough... deepseek was a minor factor.
𝗠𝗮𝗰𝗿𝗼𝗲𝗰𝗼𝗻𝗼𝗺𝗶𝗰 𝗙𝗮𝗰𝘁𝗼𝗿𝘀 - the Yen carry trade unwind, (triggered by Japan's interest rate hike). This forced investors to pull back from risk-heavy positions where #Nvidia is a key target for these liquidity-driven trades hence why it was hit hard. This selloff then smashed through the broader tech sector and even spilled over into crypto. Also, 𝗚𝗲𝗼𝗽𝗼𝗹𝗶𝘁𝗶𝗰𝗮𝗹 𝗧𝗲𝗻𝘀𝗶𝗼𝗻𝘀 as Trump’s proposed to impose tariffs on semiconductors imported from Taiwan added another layer of complexity bevause Taiwan Semiconductor Manufacturing Company (TSMC) is critical to the supply chains of U.S. tech giants like Nvidia and Broadcom.’
We also wrote an article about deepseek and china AI ecosystem, and chatGPT relating to NVDA stocks here:
🚨 AI just got 45x cheaper—DeepSeek built a GPT-4-level model for $5.6M, and if this scales, Nvidia’s AI monopoly might not last. 🚨
https://ghginvest.substack.com/p/ai-just-got-45x-cheaperand-it-might
I like $TSM better than $NVDA, but agree with the overall sentiment
people were looking for a reason that Nvidia stock was going to crash, and there was an initial impression that deepseek didn't use Nvidia. a lot of people are just looking at it like it's a money printer, while simultaneously being aware that money printers aren't actually a thing.