Such a weird sentence. The correct causality should be: It's valued in the hundreds of billions because the investors expect a 1300% revenue growth.
Another example is how Isaac Newton lost money on some other bubble as well: https://www.smithsonianmag.com/smart-news/market-crash-cost-... [ The market crash which cost newton fortune]
So even if NEWTON, the legendary ISAAC NEWTON could lose money in bubble and was left holding umbrellas when there was no rain.
From the book Intelligent investor, I want to get a quote so here it goes (opened the book from my shelf, the page number is 13)
The great physicist muttered that he "could calculate the motions of the heavenly bodies, but not hte madness of the people"
This quote seems soo applicable in today's world, I am gonna create a parent comment about it as well.
Also, For the rest of Newton's life, he forbade anyone to speak the words "South Sea" in his pressence.
Newton lost more than $3 Million in today's money because of the south sea company bubble.
Intelligent investor is a great book though.
[1] eg he wrote more than a million words on alchemy during his lifetime https://webapp1.dlib.indiana.edu/newton/project/about.do
That covers everyone. Especially and including the rationalists. Part of being highly intelligent is being a bit weird because the habits and beliefs of ordinary people are those you'd expect of people with ordinary intelligence.
Anyone involved in small-time investing should be considering that they aren't rational when setting their strategy. Larger investment houses do what they can but even then every so often will suffer from group-think episodes.
The norms of "rational" science hadn't really been established yet. There wasn't really a clear line drawn between alchemy and what we would consider chemistry today.
Newton was also definitely in favour of an empirical/axiomatic basis for science in general. If you read principia he proves almost everything[1] and of course he famously deformed his own eyeballs with wooden gadgets to do his experiments in optics.
[1] In fact pretty much the one thing he doesn't prove is the calculus, which Alex Kontorovich once said in a lecture on youtube that he has a pet theory that the reason that Newton never published the calculus was not the one everyone says about his rivalry with Hooke etc but that he wanted a rigorous proof first (which of course didn't come about until much later with Cauchy, Weierstrass, Dedekind etc for normal calculus and the 1960s for non-standard analysis to prove Newton's fluxions rigorously).
The moral of that story is that being a legend or smart doesn’t count for much in investing.
The marginal investor does.
That is a cost of capital estimate of 40%.
Which points to investors not believing the company will be that profitable.
I am not saying investors don't think they will be profitable just they certainly don't believe that profitable.
If the S-curve levels off below that level OpenAI will be an unsuccessful company.
I'm three of them and I never spent a cent on any llms, I doubt I'm the only one
> Must be nice to pull numbers out of one's ass with zero consequence.
Seems accurate?
What they are saying is if Microsoft ends up buying the rest of their shares then i.e. Microsoft's total revenue by 2030 will be more than $280 billion.
He is counting on hundreds of husbands: https://xkcd.com/605/
they'll probably fix it just like they did fix strawberry
their estimates will drop by ~20x which will be their max
as underdog in the race they'll grab fraction of even that
where are they planning to get that much money from? by showing adverts for 14h before you can prompt?
Garbage in, garbage out, same as before.
The tools are good! The main bottleneck right now is better scaffolding so that they can be thoroughly adopted and so that the agents can QA their own work.
I see no particular reason not to think that software engineering as we know it will be massively disrupted in the next few years, and probably other industries close behind.
And even say the latter strategy works - ads are driven by consumption. If you believe 100% openAI's vision of these tools replacing huge swaths of the workforce reasonably quickly, who will be left to consume? It's all nonsense, and the numbers are nonsense if you spend any real time considering it. The fact SoftBank is a major investor should be a dead giveaway.
Evidence? I’m sure someone will argue, but I think it’s generally accepted that inference can be done profitably at this point. The cost for equivalent capability is also plummeting.
Here you go: https://www.wsj.com/livecoverage/stock-market-today-dow-sp-5...
There's no such thing as "profitable inference". A company is either profitable or it isn't.
Let's for a second assume all the labs somehow manage to form a secret OPEC-style cartel that agrees to slow training to a halt, and nobody notices or investigates. This is already hard to imagine with the amount of scrutiny they're under and given that China views this as a military priority. But let's pretend they manage it. These firms also have lots of other costs:
• Staffing and comp! That's huge!
• User subsidies to allow flat rate plans
• Support (including abuse control and handling the escalations from their support bots)
• Marketing
• Legal fees and data licensing
• Corporate/enterprise sales, which is expensive as hell even though it's often worth it
• Debt servicing (!!)
• Generating returns for investors
Inferencing margins have to cover all of those, even if progress stops tomorrow and the RoI to investors has to be likewise very large, so margins can't be trivial. Yet what these firms have said about their margins is very ambiguous. As they're arriving at this statement by excluding major cost components like training, it's not clear what they think the cost of inferencing actually is. Are they excluding other things too like hw depreciation and upgrades? Are they excluding the cost of the corporate sales/support infrastructure around the inferencing?
But also, "profitable inference" IS a thing! "Gross margin" is important and meaningful, even if a company has other obligations that mean it's overall not profitable.
You mention "debt servicing", but OpenAI has no debt. All the money they have raised is equity not debt.
Have any of you tried re-producing an identical output, given an identical set of inputs? It simply doesn't happen. Its like a lottery.
This lack of reproducibility is a huge problem and limits how far the thing can go.
However, we can start by claiming that non-determinism is not necessarily a bad thing - non-greedy token sampling helps prevent certain degenerate/repetitive states and tends to produce overall higher quality responses [0]. I would also observe that part of the yin-yang of working with the agents is letting go of the idea that one is working with a "compiler" and thinking of it more as a promising but fallible collaborator.
With that out of the way, what leads to non-determinism? The classic explanation is the sampling strategy used to select the next token from the LLM. As mentioned above, there are incentives to use a non-zero temperature for this, which means that most LLM APIs are intentionally non-deterministic by default. And, even at temperature zero LLMs are not 100% deterministic [1]. But it's usually pretty close; I am running a local LLM as we speak with greedy sampling and the result is predictably the same each time.
Proprietary reasoning models are another layer of abstraction that may not even offer temperature as knob anymore[2]. I think Claude still offers it, but it doesn't guarantee 100% determinism at temperature 0 either. [3]
Finally, an agentic tool loop may encounter different results from run to run via tool calls -- it's pretty hard to force a truly reproducible environment from run to run.
So, yeah, at best you could get something that is "mostly" deterministic if you coded up your own coding agent that focused on using models that support temperature and always forced it to zero, while carefully ensuring that your environment has not changed from run to run. And this would, unfortunately, probably produce worse output than a non-deterministic model.
[0] https://arxiv.org/abs/2007.14966 [1] https://thinkingmachines.ai/blog/defeating-nondeterminism-in... [2] https://learn.microsoft.com/en-us/azure/ai-foundry/openai/ho... [3] https://platform.claude.com/docs/en/about-claude/glossary
This world of extremes is annoying for people who have the ability to think more broadly and see a world where deterministic systems and non-deterministic systems can work together, where it makes sense.
Doesn't mean the tools aren't useful — it means we're probably measuring the wrong thing. "Prompt engineering" was always a dead end that obscured the deeper question: the structure an AI operates within — persistent context, feedback loops, behavioral constraints — matters more than the model or the prompts you feed it. The real intelligence might be in the harness, not the horse.
And scaffolding does matter a lot, but mostly because the models just got a lot better and the corresponding scaffolding for long running tasks hasn't really caught up yet.
They are working on making a new measurement approach that will be more accurate.
OpenAI and Anthropic aren't building companies that aim to be API endpoints or chatbots forever, their vision is clearly: you will do everything through them.
The gamble is that this change is going to reach deeper into every business it touches than Microsoft Office ever did, and that this will happen extremely quickly. The way things are headed I increasingly think that's not a terrible bet.
For context, that is more than the annual revenue of all but 3 tech companies in the world (Nvidia, Apple, Google), and about the same as Microsoft.
OpenAI meanwhile is projected to make $20 billion in 2026. So a casual 1300% revenue growth in under 4 years for a company that is already valued in the hundreds of billions.
Must be nice to pull numbers out of one's ass with zero consequence.