Chronicles of fat years [2024-2027 Edition]

Maybe some people invest in TQQQ :smile:

Soon, soon. These are rookie numbers.

2 Likes

I’m down 20% on the day on my Broadcom position but still up 0.08% on my stockpicking portfolio. I really don’t know what to do now?

Should I ask DeepSeek?

3 Likes

You’ll be fine, dividends will be fine, after all, everybody needs soap and toothpaste.

1 Like

But the Joneses just recently bought a red Porsche … how can I ever keep up with them with just soap and toothpaste companies?

1 Like

This.

I understand why this is bad news for NVIDIA.

I also see that it shines LLM incumbents in a bad light (incumbent is anyway kind of a weird thing to say since the whole race only got kicked off a couple of years ago).

But this is definitely big positive for everyone else, since it means that whatever value AI will create across the economy in the coming years, we now believe it can be done at much less cost than we thought a few days ago.

i.e. there is likely more value going to be created, but less of it will be captured by AI companies IMO

6 Likes

Ask them how much in debt they are. And Porsche? Only car-ignorants would buy Hitler’s squashed VW Beetle, or Range Rover, or Jaaaaags edit or Mercedes instead of BMW, or cough any Tesla cough. (Not worrying I’d offend anyone here, good cars are not “mustachian”).

1 Like

I hoped this would be the needle that starts the burst of the AI bubble :crossed_fingers:

(Only slightly) orthogonal: I subscribe (as part of my job*) to DTCC’s updates on their operations – the Depository Trust & Clearing Corporation, an American financial market infrastructure company that provides clearing, settlement and other services, and I’ve been getting at least hourly (partial) service outage and service delay notification messages basically since Saturday evening.

Normally (and it happens very seldomly, like maybe once or twice per year?), this is them saying they have issues with their IT services. More often, maybe about every other month or so, this signals – they of course don’t say that – stress in their clearing and settlement processes, e.g. when a couple hundreds of billions worth of USD want to get out of certain companies and into other companies (or other securities).
As they are the intermediary counterparty between two actors transacting on, say, Broadcom, they are the intermediary counter party risk to each side, and they (I think) take extra care for managing their risk, as the trade only settles a day after trade date (so everything is on their books for a day, the securities received from the party who sold it, the cash received from the party who bought it. Discuss.)

The good news is that most trades still clear and eventually settle, but the interesting news is that apparently the DeepSeek rooted trades probably threw a couple of buckets of sand and a few wrenches into the trading mechanics of the largest engine processing the biggest portion of security transactions in the world.

Luckily, blockchain will solve this soon any day now ... ;-)

* Not sure if anyone can subscribe, or only as a participating party settling with DTCC.

3 Likes

Apparently both Meta and Nvidia already acknowledged that the model is really good and might even be better than Meta’s next release already. Also Nv said it was made by complying with teh export restrictions.

So if they say it, i would believe it.

3 Likes

Thats great insight, thanks for sharing. If so, we probably face a few fundamental changes the next few days…

The question is now… what can you do with their model - combined with Meta‘s / Googles / Open AI‘s computing power. If the model scales (more computing power => better results), we will im the next 12-18 months see the next evolution of AI (still no general AI but probably truly usefull LLMs). If the model doesnt scale or just doesnt need that much computing power to solve the LOM challenges at hand, we will see a massive asset write-off from an entire industry and probably the end of Mag7 dominance in the Stock market.

What to do: if you ask me, life with a genuine useful LLM is a a scary thought. So lets hope that it doesnt scale. Assuming this, its time to move out of Nasdaq / S&P and into either ExUS or better Mid Caps or Equal weights.

1 Like

The implications of DeepSeek depend heavily on how you expect scaling laws to evolve over the next few years. If returns to scaling continue, DeepSeek just highlights significant efficiency potential. Companies will likely adopt these insights and invest more in training, post-training, and inference compute, driving even better-performing models. In this case NVIDIA will likely be fine, or even benefit from DeepSeek.

In general: People here are too excited about financial opportunities of AI while ignoring the main issue – companies with 100 of billions of funding are actively pursuing AGI and ultimately superintelligence. None of them have convincing plans for controlling or aligning these systems. They feel trapped in a AI race – and more and more looks like it will be a race to the bottom. I’d guess there’s a 50%+ probability we’ll see AI agents surpass PhDs across almost every domain in 2-5 years, yet states and societies and labs are woefully unprepared for what this will mean.

2 Likes

Also, the market reaction is quite plausibly irrational given the information that is/isn’t floating around.

  • Most commentators don’t seem to understand the difference between DeepSeek v3 and DeepSeek R1.
  • Almost all of the discussion about DeepSeek training costs seems to be confusing training costs for R1 with training costs for ‘v3 & R1’ but you can’t have R1 without training v3. (This is not to say that it isn’t significantly cheaper, but the information situation around this is just so bad.)
  • It’s seldomly mentioned that a substantial fraction of DeepSeek’s success is likely due to catch-up growth.
  • GDM released Gemini 2 Flash which performs only slightly worse the DeepSeek R1 at lower costs. Release was more than 1 month ago, without any effects on market.
1 Like

To be honest, i really wonder sometimes how the people afford all the stuff.
Many people i know live in huge flats near zurich lake, costing over 4 or 5k. Fat cars, luxury holidays, etc.
We earn certainly above average and are not cutting back on spending a lot. But we would never think of spending 5k+ on housing.

Maybe we earn less or these people really dont put any money away. I dont know…

2 Likes

Debt. I know a guy who is my age, makes 80k euros in Greece (which is a VERY comfortable salary) and is 500k in debt. Has a detached house (luxury in Athens) and a pool (uber luxury), and a Range Rover (so he can join the “elite” Range Rover club, probably full of Joneses). He also works like a machine and is hyper aggressive/competitive.

I couldn’t live like this, this debt-fuelled luxury would choke me.

5 Likes

Sometimes you might not have too much choice if you were looking for flats at the wrong time, or have kids so need a bigger place.

I think it’s all about how you allocate your money.
Some focus on present and some focus on future

I have this feeling with some of my friends who don’t earn well (craftsman, retail employees, etc.). As soon as they get a pay rise, they either invest the money in expensive car or in a bigger flat. These are exactly the people for whom even an expensive 3a insurance policy with a monthly bill and a lifetime commitment is worthwhile. Otherwise they wouldn’t even think of saving.

1 Like

Thank god Switzerland has the 2nd pillar, else we would have a gigantic catastrophe eventually. Like Germany will soon have.

At least 2nd pillar secures a bare minumum for most.

1 Like

Doesn’t Germany have a very similar pillar system? I know Netherlands has, and pretty sure the UK has as well. It’s not too bad if someone doesn’t want to RE I think.