I'm confused how the following scenario is consistent with meeting the resolution criteria. The resolution criteria imply at least 50% decrease in AI company revenues:
Remmelt thinks there will likely be a crash by 2029, since AI companies are burning too much cash on data centers to run products undergoing commodification. He thinks it’s most plausible though that the crash happens on the investment side, and that model subscription revenues could end up being mostly maintained.
It was difficult to come up with operationalizations for investment metrics. Remmelt wanted things like AI & ML VC deal activity being <30% and Anthropic valuation <$30B. I wanted things like % of YC startups being AI-related and didn't want some large investment round in SpaceX or Stripe dominate funds raised.
like AI & ML VC deal activity being <30% and Anthropic valuation <$30B
My preference was for the former metric (based on AI PitchBook-NVCA Venture Monitor), and another metric based on some threshold for the absolute amount Anthropic or OpenAI got in investments in a next round (which Marcus reasonably pointed out could be triggered if the company just decided to do a some extra top-up round).
I was okay with using Marcus’ Anthropic valuation metric with the threshold set higher, and combined with another possible metric. My worry was that Anthropic execs would not allow their valuation to be lowered unless they were absolutely forced to offer shares at a lower price; a bit like homeowners holding on to their house during a downturn unless their mortgage forces them to sell.
I kinda liked the YCombinator option in principle, but I guessed that applicants for the summer 2025 program would already start to get selected around now, so that would not pick up on a later crash. Also, YC feels like the center of the AI hype to me, so I worried that they’d be last to give way (Marcus thought staff have their hand on the pulse and could change decisions fast, and therefore that made YC more of a leading indicator).
Good question.
Marcus and I did a lot of back and forth on potential criteria. I started by suggesting metrics that capture a decline in investments into AI companies. Marcus though was reasonably trying to avoid things that can be interest rate/tariff/broad market driven.
So the criteria we have here are a result of compromise.
The revenue criteria are rather indirect for capturing my view on things. I think if OpenAI and Anthropic each continue to make $5+ billion yearly losses (along with losses by other model developers) that would result in investors declining to invest, which in turn would lead to reduction in investments into AI data centers (and a corresponding reduction of Nvidia’s revenues). I also think that OpenAI and Anthropic are facing competition with cheaper/free models that for users often function ‘good enough’, and that particularly during a larger US recession, people will be motivated to cancel their ‘luxury’ subscriptions. Though people can get locked into using personalised chatbots.
I did ask for high betting odds, so there’s a trade-off here!
I adjusted my guesstimate of winning down to a quarter.
I now guess it's more like 1/8 chance (meaning that from my perspective Marcus will win this bet on expectation). It is pretty hard to imagine so many paying customers going away, particularly as revenues have been growing in the last year.
Marcus has thought this one through carefully, and I'm naturally sticking to the commitment. If we end up seeing a crash down the line, I invite all of you to consider with me how to make maximum use of that opportunity!
I still think a crash is fairly likely, but also that if there is a large slump in investment across the industry that most customers could end up continuing to pay for subscriptions.
The main problem I see is that OpenAI and Anthropic are losing money on products they are selling, which are facing commodification (i.e. downward pressure on prices). But unless investments run dry soon, they can continue for some years and eventually find ways to lock in customers (e.g. through personalisation) and monetisation (e.g. personalised ads).
I'd certainly bet on higher US unemployment by 2027, in part due to AI automation.
Hi David,
What is your median guess for the mean monthly unemployment rate in the US in 2027? It was 4.0 % in 2024.
Median, above 6%, but I'd take a bet with reasonable odds for >8% before end of 2027.
What do you think about the following bet?
You expect to gain money from this bet if your probability of the mean monthly unemployment rate in the US in 2027 being higher than 8 % exceeds 1/3 (= 1/(10*10^3/(5*10^3) + 1)). Feel free to suggest updates.
This is not an official proposal. I would want to think about it a little more, and then publish a post about it.
This is equal to ("rate in January 2027" + "rate in February 2027" + ... + "rate in December 2027")/12.
I'd do that, but for $1000 / $2000.
Cool! I will try to set up a draft post over the next 2 weeks.
Here is the announcement of the bet.
We made a bet on the AI market crashing by the end of 2026.
The odds are $5k:$25k with an implied probability of 16.67%. If there is no crash, Remmelt pays Marcus $5,000. If there is a crash, Marcus pays Remmelt $25,000.
An AI market crash will be defined, for the purposes of this bet, as at least 2 out of 3 of the following criteria being met.
Here are our criteria:
It’s hard to come up with criteria for what constitutes an AI market crash, as many operationalizations face confounding factors that don’t constitute a crash. These criteria were chosen since these are three of the most prominent AI-related companies and have public and verifiable revenue reporting.
Remmelt thinks that there is roughly a quarter chance of winning the bet.
Marcus thinks there is a ~4% chance of Remmelt winning the bet, mostly losing due to idiosyncrasies like one of these companies collapsing due to fraud or internal turmoil.
We both think that AI developments will be more gradual than “AGI 2027” purports.
So what’s the difference in our views?
Marcus thinks there won’t be a large crash because AI products have found product market fit, will continue to improve gradually over the next couple of years, and users will continue to pay for AI products. He holds a wide probability space for how the next couple of years will play out, with a small possibility of large changes or a winter, but the bulk of the probability mass is on “staying the course” with gradual improvements and new releases over the next 1.5 years.
Remmelt thinks there will likely be a crash by 2029, since AI companies are burning too much cash on data centers to run products undergoing commodification. He thinks it’s most plausible though that the crash happens on the investment side, and that model subscription revenues could end up being mostly maintained.
Remmelt is treating this bet as a hedge – if the market crashes, he intends to give winnings to an exceptional movement builder to organise communities to resist weakened AI companies.
Marcus is approaching this bet as an experienced market trader and forecaster, looking to profit and likely donate winnings.
This covers Nvidia revenue items currently under the ‘Data Center’ category, even if renamed or moved to another category.
I like that the amounts are reasonably large :)
Why? What advantage does this have over a lower stakes bet of $50:$250?
Small bets allow you to get cheap signaling. Larger amounts force people to be at least a little more serious. FWIW, I offered Remmelt to do more.
Fair, although it's more the size of the bet relative to the individual's income and resources that is a signal of seriousness. Using raw bet size overestimates the seriousness of people with more income and resources relative to (e.g.,) students, people in developing countries, or people with more modest means.
Right, I don’t have a high income, and also have things in my personal life to take care of.