Chronicles of 2026 - the next chapter

Alright. So consensus is that rising interest rates (and eventually a potential AI bubble burst) is bad for both bonds and equities

Most of our portfolios are constructed to benefit from AI driven growth.But let’s say there is a crash similar to Dotcom, what do we believe will be least impacted in terms of equity segment or asset class. Asking except for CHF cash.

Options

  • Gold
  • Dev ex US equities
  • EM equities
  • Govt bonds
  • Crypto
  • REITs

Value stocks and especially small cap value stocks should do well in such a scenario. They have done really well throughout the Dotcom bust.

We also have seen a glimpse of it the past months in this rising interest rate environment and the AI hiccup.

A fund like AVWS or the US versions AVUV/AVDV for small cap value + profitability. Or going for the full value + prof. suite over all caps in AVGV.

Cheap + profitable companies should weather the storm well enough.

2 Likes

My positioning is to not fight the crash and take an 80%+ fall if it happens.

My preferred diversification assets would have been gold and selected Dev ex-US government bonds (with Swiss as my preference) or Swiss or European corporate bonds as a secondary choice.

1 Like

Conventional wisdom:

  • rising interest rates across the board inherently mean rising debt / bond yields across the board, with the prices of those instruments moving in the inverse fashion, i.e. down
  • rising interest rates mean debt instruments getting more attractive (higher yield) compared to equities, on a relative basis
  • rising interest rates also mean higher borrowing costs for companies needing debt, putting a burden on earnings, hence equity prices
  • lastly, short term (central bank / e.g. Fed) rates rising, driving short-term government yields above long term government yields (= inverted yield curve) often forecasts a recession, usual delay of 4-6 quarters
1 Like

I’d say commodities (including gold).

I interview of Trump, when asked what can US do to keep long term interest rates low. Trump made a comment the biggest intervention will be our military.

It was seen as a joke in media but I think it was reference to Mar-o-Lago accords where a plan to use Tariffs and Military leverage (to force other nations to buy long term US bonds) was discussed

Is this something still on cards ?

1 Like

Bond yields rising. I wonder if this is the beginning of a change in the credit environment?

We don’t just have pressure from Iran and government spending, but also debt demand from AI build-out is huge and should also raise interest rates.

We could see an interesting dynamic where the US government slows down the datacenter build-out to keep a lid on interest rates.

I imagine we finally see credit spreads widen after being compressed for so long along with the term premium rising.

Ordinarily you’d expect this to be bad for stocks and gold, but in the context of letting the economy rip and the debasement trade, simple relationships may not pan out as expected.

1 Like

haha like there were any of them :slight_smile:

3 Likes

I just can’t reconcile these pessimistic views with the fact that all CEOs in the space are saying that there’s not enough compute for the demand. One of the explanations is that OpenAI and Anthropic are consuming it all, but at least Anthropic from what i know has reached profitability (not sure at which level, if gross profit or else). So there is demand for these things.

I believe openweight models will be a big thing. But running locally is hard, so people will still need a lot of compute from hyperscalers. Leave Switzerland aside, where i can’t find a decent offering with swiss servers and decent models among hyperscalers.

My guess? this is not black or white, I think that the truth is in the middle, with probably some overcapacity in the way. I am much more worried about replacement costs and impairements when this race to build new datacenters will fade. People are paying 4x HBM prices and 2.5x hard disks, not to mention cooling systems, land, energy supply etc. It seems a race to get to massive profits before reality shows up.

I’m with you here. AI should be very profitable and we are just at the early days of demand where we are early along the adoption curve. Most big companies haven’t yet figured out how to get the most from AI.

I think one fly in the ointment is whether the open source models can get good enough to impact the growth of the likes of Anthropic, but even if so, the profit simply shifts to these open source providers and the hyperscalers that ultimately host the model. The value will still be there, the only difference is who takes the lion share of that value.

I’m fairly optimistic, right now I think it will be:

  • End customer still gets the most value
  • The hardware vendors get a lot of value now, but I expect this will be competed away to come back to more commoditized levels, but that still could take years if hardware bottlenecks keep supply way below demand
  • Hyperscalers I expect will have worse-than-cloud profits in the long term
  • AI Labs will be a mixed bag. I think it depends if a lab can have some secret sauce. If so, then it will be an Apple like situation where the best lab captures a big portion of the token revenue even though it provides a minority of the token volume.
  • Currently, we see a bullwhip effect as demand signal cascades along the supply chain
1 Like

I am sure AI is promising technology. But I think it’s like internet. Who cares who provides internet services and infrastructure. We care about what we do with internet.

So it seems like massive infrastructure is built at a very high price. China is able to produce similar models and results in 1/10 of costs. So why are US companies spending so much is not clear to me

2 Likes

The moat is less the model but the harness and the platform with models optimized for them or even proprietarely integrated.
E.g once you start running your business with claude teams/enterprise you wont switch that easily even if cheaper models are availble.
Governance security and integration being substantial factors.
Also US vendors cab decrease inference cost with destilled model and inference optimization techniques.
While they had a rough start, i believe Microsoft is well positioned to capture significant adoption due to their reach on the users desktop and established enterprise ecosystem.
If it will pay off for all of them investing gazillions? Probably not. But the price is huge for the winners.

It is like Apple and Android. They are mostly comparable, but consumers are willing to pay a premium.

In the world of AI, there are price insensitive buyers who will pay a premium for the very best as the cost is nothing compared to the value it brings.

This is why I hypothesize there could be a similar market split as the Apple/Android world where the premium offering takes he lion’s share of the profits. However, it is less sustainable as the lock-in value of a model is much lower than the Apple ecosystem (friction of switching models is much lower than switching away from Apple ecosystem).

But I don’t see this as a stable equilibrium as eventually #2 will catch up. One scenario where it could happen is if the #1 leader gains a lead and reaches a point where the AI can recursively self-improve itself and the leader has a compute advantage so that AI improves itself recursively so that no #2 competitor can catch up due to the head start. Then we are into the take-off/singularity scenario.

I would be very careful in strictly comparing this to the internet and railroads. A balanced view would consider the totally different world it is today compared to the 90s, companies are flexible, adjust continously. See 2022 just to not go back too far.

My guess? compute cost will go down but not as fast. Depreciation will be a moderate issue. Anthropic and OpenAI will be forced to release very cheap models to not lose the race with China. The question is, can they still capture 80%+ of the market share where maybe only 10% is premium models and they can still make good inference margin on the cheaper ones?

Now would be a great time to own a big piece of land already connected to electricity and serviced. I would blast it to basically all neoclouds and hyperscalers and auction it… Pretty sure that’s a lifetime deal.

Only in FR. Ask AI to translate.

I did not look at the original source / stats, but I find it very interesting. I thought Inflation (including behavioral) was eating the saving rate in the population ex-mustachian.

If it‘s average, it‘s not that meaningful. Median is the relevant metric.

We have a shit ton of super rich in CH, that heavily skew average statistics. Those are not representative of the average Joe citizen.

My translation tool doesn‘t show me anything making a differentiation between that.

5 Likes

Good point, thank you.

1 Like

Who would have thought that one can apply tariffs (correction end trade with) to other countries because FED is not reducing interest rates

6 Likes

We are reaching levels of dumb, that I didn’t know were possible.

5 Likes

That’s ok guys, nobody lives forever.

2 Likes