In today’s issue:
- China tanks the US market in a weekend
- What does this mean for AI stocks?
- Glow duck finds his man
Meet DeepSeek… the Chinese AI that just wiped a trillion dollars or so off the US market.
This is AI from China. And as it proudly claims,
Just as a word of caution, if you had qualms about handing over your data to Big Tech in the US, probably worth checking out the DeepSeek Privacy Policy. But if you’re not as keen on reading it as I am, here’s two key points to consider:
- When you use our Services, we may collect your text or audio input, prompt, uploaded files, feedback, chat history, or other content that you provide to our model and Services.
- We collect certain device and network connection information when you access the Service. This information includes your device model, operating system, keystroke patterns or rhythms, IP address, and system language.
The thing that weirds me out there is keystroke patterns or rhythms. For what it’s worth, I checked out the same policies on ChatGPT, and it also collects network and devices information, but there’s nothing to be found about keystroke patterns or rhythms.
Take that as you will, but that does feel a little extra pervasive.
Anyway, if you don’t want to use it, but you’re technical of mind, head here to check out all its API documents. But if that’s not your thing, which for most people it’s not, here’s the lowdown on why this Chinese AI to “rival OpenAI” just wiped trillions off the US markets.
- DeepSeek is going all in on claiming it only cost them $5.5 million to train and reach this latest breakthrough in its AI. The CEO has moved from posting deep technical papers and information to memes. $5.5 million in H100 GPUs equates to (at most) a couple of hundred of the Nvidia chips. And claims it has utilised the even lesser powerful A100 GPUs. Of course, why on earth would any of us have any reason to think DeepSeek might be lying…? It’s not like China has lied to the world ever before, is it…
- Of course, taking this extraordinary claim on face value would mean the hundreds of billions of dollars the likes of Meta, Alphabet, Microsoft, Nvidia (and others) have spent on scaling AI was an utter waste of capital and time.
- It also means the “fears” from Wall St about the payback and return on investment for all this huge AI capital expenditure (capex) are coming true all at once and that US companies will never be able to get a commercial return on AI.
- In turn, because China released an AI that’s as good as US AI on a much tighter budget and far less computer – all tech stocks decided to tank lower, being led by Nvidia which shed 17% of its value in one swift move. The daily volume in Nvidia alone was around $100 billion. It melted off over $500 billion in market cap. That smashed everything (except Apple).
So, what does this really mean for the “Mag 7” stocks, Nvidia, Microsoft, Meta, Alphabet, Amazon, Tesla and Apple?
Well for Apple, not that much. For once it pays to be a laggard with this stuff. As Apple really isn’t that big in AI (yet) this doesn’t really impact them. In fact, to give you an example of how little this matters to Apple, and also how bad Apple’s AI currently is, just look at this:
Actually, come to think of it… long term it doesn’t really impact any of the Mag 7. Thinking about it even more, China has delivered the greatest gift to AI investors since Japan’s unwinding of the carry trade.
China just made AI stocks (and anything even related to AI stocks – like bitcoin miners and SMR nuclear tech companies) anywhere from 10% to 20% cheaper.
For me, right now is another great buying opportunity for companies that will lead AI over the next decade.
How on earth could I possibly say this?
Well, I’m going to explain with a meme.
Also, I don’t trust a word that’s coming out of DeepSeek right now.
$5.5 million my backside. There are multiple sources saying the reality is DeepSeek was likely trained on anywhere from 50,000 to 200,000 H100 GPUs. That’s not wildly different from the big US AI players.
And let me reinforce this point: if AI does get cheaper, then what’s going to happen is that it will come at us faster than first expected. Furthermore, Meta, Amazon, xAI, all the US AI companies, will simply be able to ramp up the output of their AI on the same hardware they have now and exponentially scale it up with plans they’ve got to add even more computer to their data centres.
In short, should you be worried about any of this? In my view, not one bit. This is a timeline accelerator, meaning demand will lift, Nvidia will still remain the main player, and companies like Meta, Amazon, Alphabet and Microsoft won’t scale back the investment – they’ll keep ramping it up but just getting way more bang for their buck.
Boomers & Busters 💰
AI and AI-related stocks moving and shaking up the markets this week. (All performance data below over the rolling week.) [Figures correct at time of writing.]
Boom 📈
- Lantern Pharma (NASDAQ:LTRN) up 24%
- Brainchip (ASX:BRN) up 18%
- Predictive Oncology (NASDAQ:POAI) up 14%
Bust 📉
- Vertiv Holdings (NYSE:VRT) down 24%
- Nvidia (NASDAQ:NVDA) down 14%
- Broadcom (NASDAQ:AVGO) down 15%
From the hive mind 🧠
- President Trump has his say on DeepSeek, and he’s not critical (at least not publicly) but affirming it’s a wakeup call to US companies to be competitive. Although I suspect that behind closed doors he wants to know how the Nvidia chips got around the export controls.
- You can’t tell me Zuckerberg and Meta didn’t know about DeepSeek. And if they did, and they still are going to ramp up AI to millions of chips… well, get ready for AI in your life everywhere.
- Meanwhile… Trump’s tariff programme is going to see a lot more investment in the US. Particularly when it comes to making AI chips.
Weirdest AI image of the day
The AI videos are getting really good and also increasingly random and hilarious. Here’s one I found quite amusing.
ChatGPT’s random quote of the day
Computers are like Old Testament gods; lots of rules and no mercy.
– Joseph Campbell
Thanks for reading, see you next time.
Sam Volkering
Contributing Editor, Fortune & Freedom
Desert Island Challenge
Bill Bonner, writing from Baltimore, Maryland
Where’s the Big Gain? Is there one?
For every patch of ice on the sidewalk there’s a tort lawyer dreaming of a beach house. And for every big sell-off, there must be a big run-up somewhere else. Or not?
The press is focused on the techs, as we did yesterday. They seem extremely vulnerable… especially now that China has introduced a low-cost competitor. You could probably make money by ‘shorting’ them.
But that is for gamblers. What we want to know is: where is the Big, long-term Gain? We are looking at the biggest credit bubble in financial history; the risk is that it will blow up. Where can we rest comfortably… safely… profitably?
Early this year, a colleague, Porter Stansberry, put a similar question to a group of analysts. The idea was first proposed as the ‘Desert Island Challenge’ by Warren Buffett in 1969. Buffett invited prominent investors to the Colony Hotel in Palm Beach, Florida.
Each was asked to name a stock to hold for ten years… as if he were shipwrecked on a desert island. Most of them picked the bright, shining stars of the 1960s ‘Nifty Fifty’ market – Coca Cola, Polaroid, and IBM were favorites.
Buffett himself chose Dow Jones & Co, publisher of the Wall Street Journal.
But then came the stagflation of the ‘70s. The go-go market of the ‘60s had raised up stock prices… and then, the bleak ‘70s pushed them back down again. Stock prices went nowhere, while inflation reduced real values by more than 50%.
Even the shrewdest investors in America couldn’t beat the Primary Trend. Most of the choices were losers.
This time, Stansberry’s participants are once again betting on proven winners, expecting them to remain winners. Hershey, PSH (Pershing Square, Bill Ackman’s company), Chubb, CME, and Philip Morris were among the choices. Philip Morris, by the way, is the most profitable company in US history… with a very long record of both capital gains and dividends, stretching back to its incorporation in New York in 1902.
But this time at least two of the big-league analysts are betting against the stock market. They chose Bitcoin… reckoning that the crypto currency would do better than America’s most successful wealth-producing companies.
Who will be right?
Most likely, none of them. All are betting that the future will be much like the recent past. Which is why you have to know where you are in order to get where you need to go.
Led by the US, in the late 20th century, the whole world moved away from gold-backed money. This enabled people to borrow far more than they had before. Central banks conveniently lowered interest rates… making it easier to repay… raising capital values… and making borrowing more attractive than ever.
The US itself provides a good illustration. In 2000, total US debt was $5.6 trillion. That year, the federal government paid $350 billion in interest. Over the next 25 years, debt rose nearly seven times. But the interest paid on the debt only increased to $510 billion in 2020 – not even 2 times. Meanwhile, the Fed Funds rate went from 6% in 2000 down to under 1% in 2020… effectively masking the burden of so much extra debt.
Globally, total debt increased in harmony with the US. Worldwide, total government debt is running about $100 trillion… with total public and private debt over $300 trillion.
Last week, we explored how asset values (supported by debt) were no longer tethered to real output. A company that produces ten autos a year might make a 10% profit on each one… and be worth, say, ten times profits… or the equivalent of 100 autos.
Then the Fed lowers the interest rate, and suddenly, the company is worth 200 autos. But there are still only ten autos coming off the assembly line each year. Half of the company’s stock market value is phony. And now, all up and down the street, real estate, stock and bond assets have been boosted by central banks’ interest rate meddling… not by increases in earnings or real goods or services.
Charlie Bilello:
With a CAPE Ratio of 37.8, US equity valuations at the start of Trump’s second term are higher than the start of any other presidential term in history. Which is another way of saying that expectations today are extremely high. Historically, that has meant below average future returns for stocks when looking out 10 years.
This is where we are. US government debt grew by $2.2 trillion last year, while the interest rose to $1.2 trillion. Such a debt build-up (and corresponding increase in credit-supported asset prices) can’t continue for long. Which means, the future must be different from the recent past… and the investments that did well over the last ten years are likely to do badly over the next ten.
Meanwhile, lenders and investors hold trillions of dollars’ worth of assets – with no corresponding real-world wealth.
The risk of the Big Loss is obvious; those assets could fall in price.
But what’s on the flip side? What happens when $100 trillion of phantom asset values disappears?
Tune in tomorrow.
Bill Bonner
Contributing Editor, Fortune & Freedom
For more from Bill Bonner, visit www.bonnerprivateresearch.com