The perils of concentration, and the opportunities in the cracks
- what the market is freaking out about, and why things are likely to calm down
“The way that organizations and organisms anticipate the future is by taking signals from the past, most of the time.” – Kevin Kelly ||
Hello, everyone! I hope you had a great weekend and that you got to touch some grass, or whatever your spiritual equivalent is – last week was exhausting, and it’s possible that was just the warmup.
Markets are hurting today, crypto included. Below, I touch on why, and why the fuss will blow over once valuations have been adjusted.
I would have liked more time to dive deeper into the market reaction and the real threat from DeepSeek, but I have to hop onto Scott Melker’s Macro Monday show in a few minutes - you can see that here.
Programming note: I have to skip publication on Thursday, apologies! Back in your inboxes on Friday.
IN THIS NEWSLETTER:
The perils of concentration
Market reaction
Something else?
Looking ahead
If you’re not a premium subscriber, I hope you’ll consider becoming one! You get ~daily commentary on markets, tokenization, regulation and other signs that crypto IS impacting the macro landscape. As well as audio, relevant links and music recommendations ‘cos why not.
WHAT I’M WATCHING:
The perils of concentration
We’re told time and time again that concentration is convenient, but risky. And yet, our collective behaviour suggests that we don’t really believe that, until the widening cracks affect us directly.
This applies to concentration in authority, as we’ve seen in elections around the world over the past 12 months. It applies to concentration in global trade and finance, as we are seeing with the move to build non-dollar settlement channels. It also applies to concentration in markets.
While all of those threads are part of what we’re seeing play out in our feeds and on our screens today, the shift in market mood is for now the flashiest and loudest.
The trigger was a sudden panic that US was at risk of losing AI dominance, and the consequences for tech stock valuations.
Last week, a Chinese startup released an AI model that was apparently trained on local chips at a fraction of the cost of US AI models. DeepSeek’s R1 was partially revealed to the public in November, but last Monday it was paired with a front-end and shared with the public. What’s more, its creator has made its code open-source, inviting anyone, anywhere to use and build on it.
The shockwaves from this are complex, but can be distilled into two main immediate concerns.
One is that the eyewatering capex plans of the major US tech companies now looks overpriced – if a Chinese startup can train a model at a fraction of the price, the ROI of the US expenditure forecasts is at best circumspect.
(tweet by Matt Turk)
Another is that US sanctions backfired. The ban on the sale of Nvidia chips to China was supposed to prevent any other country from significant AI breakthroughs. Instead, necessity is the mother of invention, and the lack of high-end chips meant that DeepSeek engineers focused on maximizing the power of what they had.
They succeeded. I confess I haven’t yet tried DeepSeek’s R1, but people who spend time using AI than I do say it’s good. In tests, it has performed as well as the much-more-expensive models from ChatGPT and other platforms.
(tweet by Marc Andreessen)
Market reaction
Keep reading with a 7-day free trial
Subscribe to Crypto is Macro Now to keep reading this post and get 7 days of free access to the full post archives.