The Breakdown - Why China's DeepSeek AI Model Crashed Bitcoin

Episode Date: January 29, 2025

Everyone everywhere is talking about DeepSeek. It sent Bitcoin down under $100k temporarily and ripped trillions of the US stock market. NLW breaks down what happened and why markets reacted so strongly. Sponsored by: Ledn Need liquidity without selling your Bitcoin? For 6+ years, Ledn has been the trusted choice for Bitcoin-backed lending. With transparency, security, and trust at our core, we help you access your BTC’s wealth while HODLing. Discover what your Bitcoin can do at ledn.io/borrowing. Enjoying this content? SUBSCRIBE to the Podcast: https://pod.link/1438693620 Watch on YouTube: https://www.youtube.com/nathanielwhittemorecrypto Subscribe to the newsletter: https://breakdown.beehiiv.com/ Join the discussion: https://discord.gg/VrKRrfKCz8 Follow on Twitter: NLW: https://twitter.com/nlw Breakdown: https://twitter.com/BreakdownNLW

Welcome back to The Breakdown with me, NLW. It's a daily podcast on macro, Bitcoin, and the Big Picture Power Shifts remaking our world. What's going on, guys? It is Tuesday, January 28th, and today we are talking about deep-seek crashing Bitcoin. Before we get into that, however, if you are enjoying the breakdown, please go subscribe to it, give it a rating, give it a review, or if you want to dive deeper into the conversation, come join us on the Breakers Discord. You can find a link in the show notes or go to bit.ly slash breakdown pod. All right, friends. One thing to note, I am not assuming any overlap in my listenership between this show and the AI Daily Brief. However, I have for the last couple of days gone much more in depth on this DeepSeak issue,

which is going to form the context for today's show. I'm going to get into it assuming no previous knowledge. But if you do want to go deeper, I would suggest you go check out the AI Daily Brief. All right, but what we are talking about today is, of course, this idea that hype around Deepseek crash the Bitcoin market. What we're going to tease out, though, is if that's really all there is to the story. going back, Bitcoin plummeted on Sunday night as traders digested the release of a new cut-price AI model out of China. BTC dropped by 5% hitting a two-week low of 98,000. Bitcoin did rise quickly back above $100,000 once markets open on Monday morning, but the volatile price action has some traders spooked? The question, has the release of a cutting-edge Chinese AI model really

made your digital gold worth less? Or is this just volatile Bitcoin markets overreacting the headlines? So let's start with a brief explanation of Deepseek and why its release has markets on edge. Last Monday saw the release of a new model called R1 out of a Chinese lab called Deepseek. After slowly building momentum throughout the week, the model became a viral hit over the weekend with its chatbot app hitting number one in the Apple App Store. Functionally, this is a reasoning model, meaning it's more advanced or at least different than the GPT4O model that currently powers ChatGPT. Instead of returning an immediate answer, the model spends time thinking to refine the answer. The style of model is much better at tasks like

document analysis and research with improved accuracy and more detailed responses. It is also not the first reasoning model that's been released. OpenAI's O1 model has been available for around four months, and Google launched their own reasoning model in December. R1 has very similar performance, depending on which metric you look at. But the big deal that's moving markets is the cost. R1 is available for around one-tenth the price of O1 for developers to use to drive their apps. It's also available for free to consumers through a chatbot U.S. And this is the part that broke the internet. Chatting with R1 is the first time most people have experienced reasoning models, compared to the free version of ChachyPT, it's a massive jump-up in quality.

It makes it seem like the Chinese labs are way ahead to people who haven't paid attention to U.S. models. In terms of quality, again, R1 is only in the ballpark of cutting-edge models from U.S. labs, but the reduced cost has some real implications. In addition to serving the model at rock-bottom prices, DeepSeek claims to have trained the model with a budget of less than $6 million. For context, OpenAI's O1 training run was estimated at around a half a billion dollars, and training next-generation models could stretch into multiple billions.

Deepseek also claims to have achieved this on a tiny training cluster of downrated chips designed to comply with export controls. The company released their model open source with the technical paper, and there's reason to believe these claims. Right now, labs are furiously reverse engineering and trying it out themselves to see if it's actually possible. The model also contains a lot of efficiency gains and new strategies that make it plausible that Deepseek are offering the use at such a competitive price. Because the model is open source, it's also freely available to run locally on high-end consumer hardware, or for third parties to serve independent of Deepseek. Several U.S. companies have already made the model available through their own infrastructure completely cordoned off from Chinese providers.

Essentially, the big picture take and the fear in the market is that R1 and the advances it represents could crater the value of big tech firms. Over the past few years, U.S. tech giants have plowed generationally high levels of investment into building AI data centers and training models. Last year's spend was hundreds of billions of dollars, and this year is looking closer to a trillion dollars. A big chunk of that is the $500 billion Stargate project that aims to build 20 high-end data centers for open AI over the next four years.

If advanced reasoning models can be trained on the shoestring budget and served at a steep discount, the implication, at least to some on Wall Street, is that data centers are massively overbuilt. The Kobayisi letter contrasted two headlines from the weekend, writing, meta, we're spending $60 billion to develop an AI data center that's almost the size of Manhattan. Deepseek, we trained our entire AI model for $6 million. Some think does not add up here. However, that's only one narrative, and there's a lot of reason to be a little bit skeptical of it. First, this is an open source model, so every AI company in the world is rushing to duplicate its results.

There's a pathway where training costs for AI models hyperdeflate this year, and U.S. labs are able to take advantage of the cut price costs for serving their next generation models. Many are suggesting the effect of this will be something that we call Javon's paradox, where price reductions in a commodity actually stimulate demand and you end up needing far more of the resource. Think about how use of cloud storage changed from the 2010s to now as the price came down. Cloud infrastructure actually boomed during that period, even though unit costs dropped. Most AI data centers are actually used to serve computing power for use of the models, with only a tiny handful of highly specialized facilities used for training. In a world where prices

plummet, big tech paradoxically needs far more AI infrastructure rather than needing to scrap facilities that are already built. There's also the question of competition. Will DeepSeat capture a monopoly on AI and put big tech firms out of business? That also seems very unlikely. R1 is currently being offered at an introductory discount, and the company seems to be struggling to keep up with demand, shutting off new signups outside of China on Monday. Next month, the price will more than double to be around a quarter of the cost of OpenAIs 01. Google is also competing heavily in this category with a model roughly the cost of R1 after the introductory discount ends. Simply put, some U.S. tech companies were already in striking distance of an R1 level service.

They just didn't have the same viral marketing moment. Still, none of that analysis stopped the stock market from freaking out to begin the week. Yesterday, Invidia was down as much as 17% dragging the NASDAQ index down 3% on the day. Over half a trillion dollars was wiped off NVIDIA's market cap, the largest single day loss in market history. total, the NASDAQ index lost more than a trillion dollars on Monday. Mike Bird, the Wall Street editor for The Economist wrote, In total market cap, the Nvidia sell-off today is a little bit bigger than if the entire listed market of Mexico went to zero. Now, I don't think it makes sense to dismiss the price action entirely because there's plenty of smart analysts sounding the alarm.

Vaser and Ling, managing director at Union Private Bank said, Deep Seek shows that it's possible to develop powerful AI models that cost less. It can potentially derail the investment case for the entire AI supply chain, which is driven by high spending from a small handful of hyperscalers. Jeffries's analyst wrote for their Monday note, concerns have immediately emerged that it could be a disruptor to the current AI business model, which relies on high-end chips and extensive computing power and hence energy. And yet, many think the price action has already played out, with Nick Carter tweeting, as concerned as I was about Nvidia yesterday, I am the same amount of concern now that

Nvidia is oversold. Fund stats, Tom Lee, said yesterday's market sell-off was the worst overreaction since COVID, adding, I'd be personally surprised if Nvidia became Betamax in the past week. That would be the kind of change that would be required to justify selling here. Invidia's chip dominance is still strong unless a new model emerges that doesn't require GPUs entirely. Markets are just generally fire, ready, aim. If you've been around Bitcoin for long enough, you've heard the term Hoddle and the name Ledden. Ledin has been the go-to leader in Bitcoin lending for over six years. They help clients unlock the liquidity of their Bitcoin, allowing them to hoddle while still accessing the wealth of their BTC. They've been battle-tested with their

focus on transparency, security, and trust, allowing them to build a proven track record of client's success and security. Letin has helped tens of thousands of clients harness the value of their digital assets, issuing more than $6.5 billion in loans over the years. But as the crypto industry says, don't trust verify. Check out Ledin's trust pilot or their reviews on social media. And to learn more about what your Bitcoin can do for you, check out ledden.com That's LEDN.io slash borrowing. Please visit leaden.io slash legal for product terms and disclosures. Product availability varies by jurisdiction. So with all of that said, what does deep seek have to do with Bitcoin? Frankly, not a whole lot as far as I can tell. Former hedge funder James LeVish tweeted,

if you're selling your Bitcoin because of the Deepseek news, you have no idea what you own. Matt Cole of Strive Funds thinks AI advances actually strengthen the case for Bitcoin commenting. Deepseek increases my conviction that Bitcoin must be the hurdle rate for capital deployment. AI innovation at this speed and scale will substantially disrupt valuation metrics across industries and is not currently priced in. Those who secure a war chest of Bitcoin will be the long-run winners, picking up distressed assets for pennies on the dollar. Focus on three things. AI, Bitcoin, and energy. Then again, the painful price action definitely happens, so let's dig into some potential reasons. The first and most obvious is that this was a highly correlated drawdown.

It wasn't just Nvidia that took a punch. Valuations were down across most of big tech and everything else that touches AI infrastructure. Interestingly, the things that weren't in tech didn't have a bad day. But as we've seen, crypto and AI are totally related in most investors' minds right now. It's split out right there in the fact that we have a white house czar for AI and crypto. What's more, we know from previous cycles that Bitcoin is highly correlated to the NASDAQ at times and especially to Nvidia. During the 2017 bull run, Invita was that year's best performing stock on the back of Bitcoin and Ethereum mining demand. Those days are long over, but the correlation is still backed into price history. Algorithmic trading models generally don't consider

narrative shifts they just trade on historical data. And for Bitcoin, it goes down alongside Invidia. While many have been pushing that this correlation doesn't make a ton of sense, it's difficult to dismiss in its entirety. Joe Wisenthall of the Odd Lots podcast tweeted, once again, Bitcoin isn't beating the three tech stocks in a trench code allegations. Beyond the correlation, there's also a feeling that market valuations were already stretched. Bitcoin is in year two of a very strong bull market. Tech stocks are even more frothy with price-to-earnings metrics steadily rising over the past year. Max Gokman, a senior VP at Franklin Templeton, said, Today's move showed just how precarious this market setup is. When valuations stretched to the sky,

it's easier for small trembles to make the entire market rumble. Chow Wang of Alliance Dow wrote, If the market is dumping for an idiotic reason like this, it was probably overvalued and just needed a reason to dump. Frankly, that is absolutely my base case. I think that two things happened simultaneously a while ago that have never been able to fully separate themselves. Those two things were the beginning of the hiking cycle in the end of ZERP and the launch of ChachypT. For about two years there, as the Fed was hiking rates, the one counter-narrative on Wall Street that was keeping things afloat was, of course, AI and specifically NVIDIA. Subsequent to the hiking cycle ending and the rate-cutting cycle beginning, AI didn't have to do

as much narrative work anymore, and so I think people were looking for a way to reprice those stocks. And yet, Nvidia just kept performing so well that it hasn't given people much of a chance to do that sort of repricing. What that leads to is anytime there's any sort of narrative justification for an Nvidia flip, people way overdo it. This is just the latest in a line of exactly that. So bringing it back to Bitcoin, most Bitcoiners are looking at this as a buying opportunity, betting that the bull run isn't over. CMS Holdings tweeted, You can print more Chinese knockoff AI companies, but you can't print more Bitcoin.

Standard charters Jeffrey Kendrick suggested that with the announcement of Trump's crypto policy last week, were out of the hope phase and into the disappointment or confusion phase. The next phase he wrote is to buy the dip. Jim Kramer is also pounding on the table about Bitcoin's dating on Monday show, I own Bitcoin, you should own Bitcoin. Bitcoin is a great thing to have in your portfolio. Okay, so maybe that one isn't the greatest indicator. Now, quickly to flag, Deepseek isn't the only thing going on in markets this week. It's just the easiest thing to point to. Bloomberg's Katie Greenfield wrote, 10-year treasury yields down 12 basis points this morning. Deepseek is a macro event. With respect to Katie, there's a lot of factors overlapping

to make it seem like Deepseek is a macro event when it might just be a catalyst. The Fed meeting is currently underway with the rates decision on Wednesday. The pause is almost certain to begin this month and expected to drag on into the summer. This isn't a surprise, but it also isn't a tailwind. Stocks can no longer expect Fed support to move higher, so we'll instead need to be powered by growth narratives. That means there's even less reason to get bullish here. In particular, there's very little reason to add more risk on dicey price action heading into a Fed meeting. Equity looked a little sketchy on Monday with traders happy to wait the verdict from Jerome Powell. This was also the first Fed meeting since inauguration day. Trump came into office with a stated agenda to bring

interest rates down. He even floated the idea of exerting influence over the Fed this week. Powell announcing a rate pause at the first meeting of the new administration might not go down so well and could reignite tensions with Trump. That's not to say a surprise is expected, but headline risk is elevated. Another technical but important factor weighing on tech stocks is the lack of buybacks. We're currently heading into earning season so buybacks are in their blackout period. There's been a lot of ink spilled about whether or not the blackout period affects stock prices, but it's reasonable to think they have an impact on the margins. It's also logical this effect would be seen most acutely during big drawdowns due to the lack of buying support. Big Tech spent over $200 billion on buybacks last year, retiring around 2% of outstanding shares. The blackout period likely isn't a major cause of the drawdown, but it's another factor adding to market

fragility. Tesla, meta, and Microsoft are all scheduled to report earnings on Wednesday following the Fed meeting. Apple and Amazon are reporting the following day, with NVIDIA and Google up next month. Each of the Mag 7 bar Apple are forecast to beat estimates. There's the potential, though, that a miss, or even just an awkward comment about Deepseek, could drive another leg of the sell-off. Lastly, going a little under the radar as a contributing factor was a rate hike from the Bank of Japan on Friday. The policy rate was raised to an 18-year high of 0.5%. Although the move was expected the yen strengthened by 0.5% against the dollar, reversing a two-week downtrend. As strange as it is to say, Japan seems to be heading into an inflation crisis. December saw 3.6%

annualized inflation on another major uptick. It was the hottest month in the last year, and dwarfed anything seen from 1998 until 2022. You'll recall that in August, the BOJ shocked markets with a surprise rate hike, contributing to the unwind of the yen carry trade and sending global markets diving into a brutal sell-off. The yen was traditionally used as a funding currency, meaning global financial firms would borrow in yen at rock-bottom rates and then convert into dollars to trade international markets. After the August unwind, it looks like there's much less leverage in this trade, and few firms are getting caught offside. Still strengthening in the yen and rising Japanese rates have a negative impact on global liquidity, even if it's marginal. After August's embarrassing event, the BOJ adopted

bed-style forward guidance to ensure markets were clear about their next move. On Monday, though, the central bank reversed this policy, stating they would go back to their traditionally fuzzy communication style. Markets have already adjusted pricing to just 46% chance of a cut in February down from 70% last week. This is the first Japanese hiking cycle in almost 20 years, and the B.OJ is understandably taking it a little low. They don't want to snuff out the inflation they've been waiting decades to see. There's also the risk of bankrupting the government, which has a world record debt-to-d-GDP ratio of 217%. Isirukato, the chief economist at Toton Research, said, because the BOJ doesn't know where exactly the neutral rate is, it would have to wait about

six months after each hike to check the health of the economy? Only after judging that the neutral rate is still distant, would it raise rates again? That slow and steady approach runs the risk of seeing inflation running out of control and forcing a swift and volatile response. Nothing has happened yet, but it's another looming black swan to watch out for. So, does the release of an efficient AI model out of China mean it's time to pack it up for this Bitcoin cycle? Zach Rines, aka Chainling God, isn't buying it, tweeting, for some reason, the Deep Spark market-related fud just isn't grabbing me. Bitcoin dumped because China made a faster, cheaper AI model, which may be bad for some specific U.S. companies, but is broadly good for everyone else. Okay, wake me up when there's some

actual fud to worry about. Preston Pish commented, Bitcoin is the solution to intense technological AI deflation, which governments will print fiat like crazy to offset the effects of. Neil Jacob thinks the stock market freak out could be an interesting case for Bitcoin, tweeting, The only thing I feel comfortable owning in these uncertain times is Bitcoin. With rapid advancements in AI and technological expansion, I think you'd have to be crazy to pick and choose individual equities. A ton of people have also been asking Deepseek for its Bitcoin prediction for this cycle. It's responding with the prediction of $180,000 to $220,000 top in October or November. Zooming out a little, Bitcoin traded below $100,000 for all of about 10 hours.

In fact, frankly, most of my listeners were probably asleep for it. As Pomp wrote, years ago, we dreamed of the day when the media would scream, Bitcoin is crashing below $100,000. Be thankful we finally arrived. That's going to do it for today's breakdown. Appreciate you listening, as always. And until next time, be safe and take care of each other. Peace.