A weekly thread to discuss financial matters - from personal all the way up to global.
Ground Rules
- Remember that we're all just Internet randos. Don't bet your life savings on a hot tip from this thread.
- Keep culture war in the culture war thread. Yes, global events may impact our personal finances, but that does not mean we have to incessantly harp on culture war aspects here. If you are going to discuss it, please stick to the practical impacts of it on an individual level.
- Be kind. Remember that everyone here comes from different circumstances. We all have different resources available and different risk tolerances.
- Don't let the perfect be the enemy of the good. Better is better. Celebrate people when they take a step up and work to move their finances in the right direction. Don't flame out because they haven't followed what you consider the optimal path. Everybody has to start somewhere.
Jump in the discussion.
No email address required.
Notes -
This weeks roundup:
The Fed has decided to hold at 3.75-4.00%, despite inflation rates remaining over target. Two votes dissented. Official communication indicates that more data is needed before another increase, while critics argue that the hold is due to political pressure around the upcoming election this week.
The Fed seems like they're stuck in a tight spot, if you believe the thesis that inflation is currently being driven by policy choices that are independent of interest rates.
[https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm](Anthropic has disclosed its S-1 to the public). Reporting has been highly variedso far, with something akin to Scott Adams' "two screens" metaphor happening. Bulls are focusing entirely on operating income and claiming that accounting losses on convertible notes are immaterial. Bears are focusing on a noteworthy lack of Q3 data.
Personally, I am not particularly impressed. But I'm very interested in bullish takes that don't involve the phrases "recursive self improvement" or "super intelligence".
Tech stocks stumbled this week as investors have grown anxious about cash flow. Google and Amazon, in particular, saw dips. Whether this anxiety persists remains to be seen.
The bulls are correct here, IMO. Anyone focusing on the "losses" seems to be missing the plot.
For me, the S1 confirmed the bull case. Anthropic has, cumulatively, spent just $9 billion to achieve its market position. Everything else has been covered from cash flows. As a result they have $120 billion in cash on their balance sheet. The IPO will give them another $100 billion.
That money is going to continue to enter the ecosystem, including into cloud providers.
I predict that, this time next year, Google and Amazon will be up a lot and it will have seen obvious to buy today.
This seems correct to me - their cash on hand really makes past losses a non-issue, as far as I can tell - but I'm not sure it follows that the bulls are right. Isn't the fact that Anthropic is on the hook for checks notes something in half a trillion in future spending regardless of revenue or use much more important than past losses?
It seems to me that investing is betting on Anthropic's compute build-out paying off. Even after the IPO, they're going to be hundreds of billions in (effective) debt even if they turn around and throw all of their cash on hand and newly-raised IPO cash (as per your numbers).
I don't think that it's impossible that bet pays off, but the trajectory doesn't look ironclad to me, between Anthropic's growth slowdown and the falling token price. And if Anthropic runs into trouble meeting those commitments, they could have a hard time taking out loans to meet them due to a projected poor credit rating.
Anthropic has a long history of one-upping OpenAI, so I definitely am not counting them out. It just seems like, based on what we know so far, a position where a lot can go wrong very quickly.
Ultimately, it's entirely a bet on capabilities. Will Anthropic (and OAI) develop models in the next year or two with enough intelligence to replace or supplement a couple trillion dollars worth of human intellectual labor? Then they're solid. If not? Shareholders will be left holding a bag of shit.
Falling token prices are a sign of strength, not of weakness: OpenAI and Anthropic have large margins on inference, and those margins are likely growing, not shrinking. This is due to improved engineering significantly reducing the cost of inference every quarter; open model competition is a minor element at most, and ANT and OAI are not currently in a race to the bottom with each other in charged cost. The lower token cost (and, more importantly, cost per task completed) increases the size of the market for intelligence. Is it better to sell X tokens at $Y profit/token, or 10X tokens at $0.5Y profit/token? (Of course, it's mathematically possible for it to drop to $0 profit/token or less, but that's not the world we're actually in, at least right now.)
Anthropic's slowdown is more concerning for the bull case, but two points there: 1) it's less a slowdown in the overall market and more OAI showing stronger competition in the coding space and 2) coding is a relatively small part of the overall economy, and AI's impacts outside of it have been limited at most so far. It returns us to the question of whether capabilities will advance enough in the next year to start making a real impact in other sectors. Will it? Impossible to say until it does or doesn't, but it's not a linear process: like coding, it would have no meaningful impact until it reaches some threshold of capability, and then rapidly explode.
I don’t think open model competition will necessarily stay minor forever. It’s also perhaps more lucrative for surviving independent or semi-independent datacenter operators. Frontier models will be key to things like cybersecurity, its much less clear that running accounts payable for a construction company or handling automated help chat for a video game publisher or writing a mean review of a hotel you just visited on TripAdvisor based on your voice note requires them, and it’s mundane tasks like these that will be the majority of inference tasks if not tokens in relatively short order, if they aren’t already. Even for relatively powerful agentic stacks, it’s my understanding that a lot of consumer and enterprise tools now use cheap open models to triage and then call agents that might themselves call, where deemed necessary, frontier models. Most customers increasingly won’t notice the difference. I also think the position of the datacenter operators is getting increasingly dire and the market seems to agree. Coreweave and Oracle debt with spreads of 400-600+ bps to the treasury curve is ominous, even new issues at like 9% interest, but even if these vendors / providers survive, the margin to pay the coupon has to come from somewhere and with both GPUs and energy completely commoditized, that side of the business essentially relies on compute demand for inference as not only a bottleneck, but one that will experience zero slack as a result of falling token costs, open models and so on.
To believe in the data center narrative (and you have to even if you don’t own Coreweave or Oracle because every other big tech and frontier lab is guaranteeing a lot of that debt one way or another) you have to believe that if, say, inference and training becomes 50% more efficient overnight, that slack is going to be immediately taken up by new demand without any price pressure or panic.
It's not that frontier models are able to magically somehow do a better job on mundane tasks: "what is 1+1?" is saturated by both open and proprietary, and even economically meaningful tasks like coding can be handled reasonably well by open models today. And the tasks that can be handled by both will expand, with open weights likely remaining 3-6 months behind.
But my claim here is that it's significantly cheaper for the frontier labs to serve those than it is for some host of open weights to. The only reason open weights are cheaper sometimes for the customer is because frontier labs don't care to compete for them: even they (especially they) are compute limited, and their success is dependent on them developing better models that can handle all economic tasks, not by them being the choice of inference provider for every mundane task in 2026 Q4. (I don't think even offering Codex or Claude Code as products matters that much economically for them, except in that they're proofs of concept for automating economically valuable work and help with internal work.)
Fully agreed here, which is where the question of model capabilities come in. If within the next year or two frontier labs introduce models that automates most white collar tasks, then there will be no slack. The timing here is critical: this also will have to occur on that short timeline. Successfully automating accounts receivable for a construction company, but in 2029 instead of 2027, is not going to cut it, at least from the perspective of a neocloud. Nvidia is incentivized to backstop them, which could buy some time, but even its powers aren't infinite.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link