Which means Microsoft (9% of the Nasdaq 100), is using its cash and services to power an AI company which touched $1.6billion in ARR for 2023 but is raising money at a $100 billion valuation.. (raising money at >50X non-profitable sales)
I don't think $100b for OpenAI is that farfetched. You have way less transformative companies being valued north of $100b.
After playing with ChatGPT (using GPT4) and GPT4 API, I'm quite convinced that only 2 things are stopping LLMs from being ubiquitous right now and one of them is simply inference cost. The other is GPT4 being having many intentional guardrails.
That's not to mention GPT5 and 6 and so on coming out.
You can argue that OpenAI might not hold the lead. That's fine. But the author also argued against Nvidia. Nvidia doesn't care if OpenAI loses the lead because the next LLM lab also uses Nvidia.
I'm not a stock picking expert. But I'm investing like I'm going to lose my job as a software dev. If I lose my job, my stocks will surely pay for my food. If I don't lose my job, then it will continue to pay for my food. It's a hedge for myself.
The problem is that other companies were able to rapidly replicate and sometimes surpass most of what OpenAI have. It costs millions of dollars to do it, but that’s less than the billions of OpenAIs market cap. In terms of product / ecosystem / IP moat, I would argue the incumbent tech companies are in a better place to deploy these models. Will you use OpenAI when AWS can serve a model that’s good enough and they have already gone through procurement and billing? Will you visit the OpenAI page if Apple are able to integrate chat into your OS?
The tech is real but OpenAI wins as part of Microsoft IMO.
I think the expectation is that as companies begin drawing moats around their data, it will be harder for new players to enter the market. As these models progress, it will just get more expensive to compete, so only the big players will have a shot. OpenAI might only have a slight advantage at the moment, but they have proven they can do it, they have the talent necessary to continue to innovate, and some great backers, including the government through Larry Summers. All they need is a boatload of money, and that's exactly what they are getting through their huge valuation.
I have no idea if that reasoning is sound, but it is at least plausible. I've compared most of the big LLMs, and OpenAI still has a really solid lead imo., mostly due to tooling. But I also suspect they are very close to GPT5, but will wait to release it until they have serious competition, eg. through Gemini.
Yes. This article demonstrates the problem of trying to reason by simple earning analysis alone. You can have to consider the dynamics of the businesses involved.
OpenAI being worth $100B doesn’t seem ridiculous at all to me. There is a lot of risk investing in such a stock, but given it has gone from nowhere to over $1B in revenue in such a short time, and it has a very strong brand position, is partnered with one of the best run tech companies, has a unique training pipeline, has fantastic connections to important institutions, is the first mover, and has an array of impressive products, you would expect that 1B to easily reach 10B, and reaching revenue of 100B doesn’t seem far fetched.
Taking another analysis: if they manage to charge $50 per business user, which isn’t too far from current charges and not that different from say what Microsoft costs, and is a small amount to recoup in productivity gains, then they need 150 million or so users. Office has double that. And of course we can see many other revenue sources, like education.
Their biggest challenge will be competition, and that makes it highly uncertain. It’s a very _risky_ stock, but the potential upside is very high. It is not a bubble if the price is driven by expectations of earnings growth.
I do consider the dynamics of the businesses involved - when I Write about a single company I consider it very deeply, when I write about a basket of stocks, I adjust my message to provide for that. I have written tons about differentiating between business models, and have spoken about that on my recent podcast interview with Value Hive.
To get to the punch, Open AI being valued at $100bln may not be ridiculous to you, but it is to me. What is your risk/reward? Abysmal. Even if you hit your targets, no biggie.
What insider information/expertise do you have to say OpenAI valued at $100b is ridiculous?
Do you have their financial information? Growth rate? Roadmap? Customer pipeline? Have you tested GPT5? Are you a leading AI expert? Have you worked at OpenAI before?
I'm not saying that you're wrong. It's just silly to say its valuation is ridiculous when it's a private company.
We know Open AI ARR, competitive landscape, the reason why they raised the capital at that valuation (MSFT's goals). We also know that they DO NOT make money, which isn't great considering that there's a bunch of other competitors ramping up to compete with Open AI.
We also know that this is an incredibly dynamic sector/business, which means there's a high chance Open AI may not even get close to achieving its objectives.
Roadmap is one thing, how you value it is another. No, far from a leading AI expert, but I don't need to be a leading AI expert to think.
Do I need to work at Open AI to make an assessment of the company's current valuation? Come on now...if you've been an investor for long enough you know that employees very rarely have a good understanding of the company they work at.
Whether private or public, I have used limited information to make an assessment of the $100bln valuation, and yes for me, it's ridiculous.
I do appreciate your feedback however and constructive criticism.
>Their biggest challenge will be competition, and that makes it highly uncertain. It’s a very _risky_ stock, but the potential upside is very high. It is not a bubble if the price is driven by expectations of earnings growth.
OpenAI isn't a stock. It's a private valuation.
I actually think they'd be well above $100b if they're a public company right now. Just look at how much $MS added in market cap since ChatGPT.
Re: OpenAI, being "transformative" doesn't necessarily mean being worth a lot. Do you think OpenAI can hold their (so far) unchecked monopoly in such a hot market? Can they 10x their current customer base? If inference cost falls, wouldn't they have to lower their prices to compete with others? And as open source models continue to improve, might companies opt for local, slightly-less-powerful models that save them hundreds of thousands?
First point, being transformative doesn't mean making money. Absolutely, that is a massive layer that people seem to completely miss. (Blind spot?)
As for OpenAI's monopoly - no I don't think they can keep it. Everyone and their grandma is trying to raise funding for LLMs and chatbots. What is the bottleneck or moat keeping OpenAI immune from competition? Not much.
For the last point - yes, at some point LLMs could become local and change everything. The space is so dynamic, virtually guaranteeing that everyone will become unseated over a long enough period of time.
Personally, I don't care about OpenAI's valuation. It's a private company. Who knows what's on their roadmap? For all we know, they're closer to AGI than anyone else and GPT5 is a huge leap in performance and cost, which would make $100b a bargain.
Trying to valuate a private company with little to no information is pointless.
Ok.. and? Why would that translate to these kinds of valuations? The open source models seem pretty close behind. Close enough to wipe out a huge chunk of any potential valuation at the very least.
>Ok.. and? Why would that translate to these kinds of valuations?
I don't know. No one knows. It's just a thought. They could be on the verge of GPT5 and will continue to dominate for years to come? They have developed an extremely cost effective LLM delivery platform that open source can't compete against? They have a huge backlog of customers that they can't serve yet because of AI chip shortage?
Sure. That's why it's (arguably) a bubble: it's speculation about the future.
But the pattern of history is pretty clear on this: it's not the first movers that capture the value. In this specific case the value seems to be captured quite well by open source models too.
You are right about the inference cost. The real economic value-adding tasks (for example replacing a whole team of developers with a technical product manager plus AI code generation) are already almost feasible (enough to see what the future will bring), but severely constrained by context limitations, processing speed and ease of integration into development workflows. Since both computing power and clever algorithms can be brought to bear on it, it will prove to be a situation featuring a Moore's law with a very short doubling time.
I spent Christmas playing around with the GPT4 API. I quickly discovered many ideas on how to use it. But many of my ideas depended on much lower inference cost and much higher context sizes.
I equate this problem to the cost of loading a website in 1996 vs 2024. My wild guess is that we are probably in 1997 right now.
Therefore, I concluded that not only would LLMs be ubiquitous in the future, I should invest in companies trying to solve the inference cost problem right now.
"You can argue that OpenAI might not hold the lead. That's fine. But the author also argued against Nvidia. Nvidia doesn't care if OpenAI loses the lead because the next LLM lab also uses Nvidia"
Yes, your point is that Nvidia is overvalued because of AI hype and that its financials looks priced for future growth instead of value. I asked ChatGPT to summarize the article for me.
Listening to finance and VC podcasts it seems to me quite clear there is a bubble brewing. They are all repeating the same mantras and convincing themselves of a new gold rush.
The financial side is convinced developers are 10x as productive with copilot and AI tools. They are and will increasingly be forcing CTOs to get on the copilot bandwagon or risk getting fired. The tech companies are not likely to get fooled, but your average non-tech S&P 500 company is going to be giving MS a fortune.
AI companies are getting funded with hundreds of millions/billions with no realistic path to making money (mistral, anthropic, huggingface). The only one making money is nVidia and MS.
nVidia, MS and others are frankly committing clear fraud with round tripping money and making a killing.
When there is billions being spent on ventures with no business model we are repeating the dotcom bubble. I bet it is going to last a while yet though so plenty of money to be made before the inevitable.
I don't think copilot makes you 10x more productive. Anyone who has used it can tell you that's nonsense. 30% more productive is way more realistic, but even 30% is clearly worth it from a financial point of view.
If you're a company the only reason you should have for not buying Copilot for all your employees is legal concerns, which are definitely still not resolved. Once they are though it's going to be pure masochism to not use Copilot (or similar).
Like not paying for CI, or not using Git, or not using an IDE. You can totally still do it but you're going to suffer compared to the competition.
Of course it does not, but people in finance do not understand engineering. And the C suite needs to make AI plans or wallstreet is going to punish them.
Consider that opinion on podcasts are easily on sale. It says nothing of what companies will actually do.
And I think any legal department worth their value might want to veto it since there are ample unanswered questions about who owns the copyright afterwards. Microsoft claims one thing… but their contract also puts the blame for Microsoft being wrong on the customer, so I don't think they are very confident about it.
> It says nothing of what companies will actually do.
Wallstreet rewards any company with AI plans. It is clear what the C suite will do.
There will be no legal issue except for some very specific paranoid companies. Plenty of them host their code in private github repos and manufacture in China. It is all standard practice. Generally nothing is kept that close to heart.
> The financial side is convinced developers are 10x as productive with copilot and AI tools.
The opportunity is bigger. It isn't 10x existing developers, but making everyone 3-5x as productive by turning everyone into someone able to automate a basic system. AI could tackle pretty much every knowledge job in the lowest tier of a job pyramid.
1. AI becomes so good that I will be out of a job as a software dev
2. AI becomes really good and I will 10x my productivity
3. AI hits a hard ceiling and it might only make me 2x more productive
Based on those 3 possibilities, I invest my portfolio like this:
1. 40% goes to the main layer of the AI revolution which are chip manufacturers which is mostly TSMC & Nvidia. I throw some money at Intel as well because geopolitics. If AI is ubiquitous, then we need a lot more silicon.
2. 30% goes to the next layer which is mostly Microsoft, Apple, Google, etc. These are the companies close to consumers. I don't know which one will win out so I spread it across.
3. 30% goes to S&P 500 because productivity increase will benefit all industries. Also, if AI hits the ceiling fast, at least I will have 30% of my money in S&P 500 and not AI bubble companies.
I basically invest like I'm going to lose my job. If I don't lose my job, great. If I lose my job, that means my stocks will have to pay for my food.
I think Copilot makes me 2x productive at rote stuff like writing tests and - wait for it - less productive at writing complex code. It’s just not very helpful having incorrect suggestions flash across the screen. It’s actually quite distracting. I have turned Copilot off for now.
I mentioned in another post here that I believe that one major reason LLMs aren't ubiquitous right now is inference cost.
I'm going to guess that one major reason Copilot fails at complex code is because it can't read enough of your codebase to suggest complex code. The technical reason is because inference cost is too high right now for very large context queries. Therefore, the major bottleneck is inference cost.
That's why I'm investing in TSMC because we will need a lot more silicon to reduce inference cost.
I imagine a world in which future GPT versions are at least 2x better than GPT4, context size is 100x higher, and inference cost is 100x lower.
> That's why I'm investing in TSMC because we will need a lot more silicon to reduce inference cost.
Or what we really need is another approach to either the hardware problem (more efficient chip designs that may be entirely different from existing ones), or the software problem (more efficient ML core models). The latter is obviously more likely to happen, rather than the former.
But I don't know who to bet on for solving the software problem. The hardware problem will depend on chip fabs, TSMC, Intel, Samsung. Since I can't buy Samsung stocks easily, I bet mostly on TSMC and a bit of Intel for geopolitics reasons.
There are also chip designers such as Nvidia, AMD, Apple, ARM, etc that will need to solve the inference problem. I'm already invested in Nvidia and Apple. I don't know who else to invest in right now. Many of the AI chip design companies are private and startups.
I suspect that AI and tech in general are _extremely_ well correlated. Indeed AI is just a subset of tech. Because they are so well correlated, your strategy really does not provide any hedge, you’re just all-in on tech. I’m very bullish on tech as well, but your described strategy makes me worried because I sincerely think you could lose your job and then find that the AI stocks don’t do well. Eg, consider the scenario:
1. We hit sudden limits in computational density, energy efficiency, and energy storage technology
2. We face another AI winter (which happens surprisingly often!)
I don’t know the industry you work in, but for many in tech that could lead to a layoff and then underperforming stocks. I think you can imagine an equivalent scenario for yourself, based on your industry, etc
almost everyone I know who is somewhat 'digitally-savvy' (and even some others) use GPT for various things. I, as a very early adopter, use llm's of various kinds for more and more things; my conservative estimate is that I do at leat two 'requests' an hour, but obviously ANY real interactive session balloons those requests to bigger and more.
So aside from the fact that LLM's are already used so much in such a short time, I fully expect that as UI's get better, even if everything else stays the same, more and more things will be (subtly) LLM-powered. Because a lot of the stuff I do regularly are super useful to non-techies, and it's just a matter of UX/acclimatization.
For what it's worth, I'm not a 'fan' of what I think is a a pandora's box of sorts, culturally especially. I unironically call mine Gepetto to remind me that I might not be as much in the driver's seat as I'd like, pulling the metaphorical car strings.
okay, sorry about this turning into a glorified blogpost :)
I have 30% in S&P500 for a reason. Maybe you think it's too low?
I do want to consider your scenario:
1. If we hit a limit on chip node tech as you said, all stocks would suffer since advancements in computer chips contributes to all industries.
2. If we face another AI winter, then my job is more safe? Even if I get laid off in my current job, I assume that AI isn't good enough to completely replace me based on your scenario so I at least can find another job.
I think a professional financial planner (which I’m not) would say that 30% is too low, and also advocate holding cash and some bonds. You really have a very aggressive investment portfolio. But personally I don’t mind aggressive investment, and I feel your present allocation may be fine depending on your beliefs and your risk tolerance. Your initial post has this if-I-get-fired/if-I-don’t-get-fired analysis, which I think presents your investments as lower risk than they really are. I think your job and 70% of your investments is going to do whatever tech does. If you still feel ok with that, then I think your allocation is probably matching your beliefs and risk tolerance well.
Your initial post has this if-I-get-fired/if-I-don’t-get-fired analysis, which I think presents your investments as lower risk than they really are.
The idea is that if I don't get fired, then my job will continue to feed me. If I permanently lose my ability to make money as a software dev, then AI will have taken off and my stocks should now feed me.
As another commenter pointed out, you’re essentially betting heavily on tech, which may or may not work.
But why bother picking individual stocks if you can get lower expenses and higher rebalancing efficiency for your factor tilt by picking a tech ETF instead? You also wouldn’t have to count on Nvidia, Google and the like retaining their top positions forever.
There’s also no guarantee that S&P 500 won’t end up trailing world indices in the near future.
Probably because I don't see other public tech companies right now that can challenge the AI incumbents. None of the AI chip startups looking to challenge Nvidia are public companies. I don't see any potential AI consumer companies that can challenge AI delivery to consumers better than Apple, Microsoft, Google, Amazon can.
Many of the tech companies in the NASDAQ are themselves older, "traditional" tech companies which can be disrupted by AI as well so I don't want to buy into all tech companies.
I might switch to a tech ETF in the future. But right now, I prefer my own picks.
What if you lose your job and your stock? E.g. chipmaker neither wins nor loses, and all the other tech goes down or doesn't rise because some other, new, challenger takes the market.
You sell some to pay for your living expenses and sell some to move into fixed income for some guaranteed income. You rebalance your portfolio by your risk tolerance and your current needs.
How would this make sense if the market is a bubble? Bubbles pop by definition and you end up losing everything? Not sure what it has to do with your employment.
You can view your skills as an asset, and balance your investments accordingly.
"Lose everything" is a wild stretch. They have 30% in a moderately broad fund. They have a chunk in companies that are still doing non-ai things, then their first part is chip manufacturers. AI could disappear and people would still need computers.
Because if AI is good enough to permanently end my software dev job, it should be valued much much much higher than the stock market today.
I imagine a world where anyone has an AI powerful enough to make apps that used to require a full team of professional software developers and do so much quicker. If we get to that world, I don't see how the stock market wouldn't be significantly higher than today, even if it's in a bubble today.
Back then, for over 10 years, it grew at a pace not seen since then.
If Nadella grows Microsoft into the AI area as fast as Gates did in the Desktop area, it would become a 100 trillion dollar company within the next ten years.
Is AI a step forward for productivity as big as desktop computers were back then? I think it might be. Not sure what percentage of people became more productive via PCs. And how much more productive they became. But with AI, it seems everyone will benefit from it. And become twice as productive or so within the next 20 years.
Programming without an LLM at hand would already feel very clumsy to me these days.
>Which means Microsoft (9% of the Nasdaq 100), is using its cash and services to power an AI company which touched $1.6billion in ARR for 2023 but is raising money at a $100 billion valuation.. (raising money at >50X non-profitable sales)
I've been saying it for quite some time. The end game of ChatGPT (and similar) services is:
1- Get people used to them so much they can't function without them (without massive drop in productivity, inconvenience etc)
2- Raise the price 3 orders of magnitude
3 -Profit
This only works if open source models are so far behind to be essentially unusable as a replacement and traditional internet search becomes completely useless(it has been heading in that direction for a while by now).
Current NVIDIA stock is a classic example of a gamma squeeze. If a market maker is short a call/put, he is hedging with stock in the direction of the movement. If you buy out of the money calls(i.e NVDA stock is at 500 and you are buying a 520 call), especially shorter dated ones, then the market maker will end up supporting a rally if it comes around via hedging(this is also exacerbated by the fact that retail, which is the segment of the market usually "aping" into OTMs, doesn't actually trade the vol component of options, meaning that they don't sell short stock as it goes higher).
Internet investment advice... If the writer was any good, they would quietly invest their money based on their predictions (such as shorting the "bubble" stocks) rather than provide advice for free to everybody else.
Hold until December or January. Expect volatility over the summer. There is probably about 15-20% left to go. They're expecting more than $20 billion in chip sales alone this quarter. Expect rhetoric out of both political parties over regulation, which will cause additional volatility. Expect a deal between Nvidia and Apple. Apple is behind the curve. Both companies want to control ARM.
You’re making some very bold predictions. It makes me uncomfortable to see such specific claims, backed up with no data, so soon after the bubble with crypto. You’re either a prophet or overconfident.
ycombinator makes it difficult to provide a detailed analysis. But I would recommend you look to history of other melt-ups, and compare. The one notable difference with Nvidia is they actually have a product and revenue. To verify the $20 billion in revenue, go look at their own guidance.
I guess you haven't heard. Apple is already under contract with Nvidia for the H200 and B100. I guess hell froze over. Apple will not be able to participate in the AI space w/ out Nvidia. Nvidia dropped CUDA support for MacOS, but Apple is going to have to get it back. Wait for the deal that is coming.
The problem isn't in the stock market, the problem is the excess production. Trying to boost the economy has the exact opposite effect, and makes everything even worse. You need to slow it down, and boost spending instead.
The nasdaq is 1500 companies, seems the article is about the Mag7. People are putting money back in the sector, it may not be MS or Nvidia, but the money will move to the next big thing that pops up in the nasdaq
I don't see it as a bubble at all. Investors see that generative AI has a lot of potential, and invest in the companies that are likely to benefit from it. That's how stock market is supposed to work.
Something tells me that lots of people are about to get trapped in the market believing that the AI narrative will make it keep going up forever.
It would take a very nasty surprise in the market that will put all of that at risk.
I guess the euphoric sentiment in AI is a signal to safely take your money out right now before the stock market gets upset with several companies showing signs of growing geo-political / legal risks and especially extreme over-valuation.
Given that a single company is carrying the stock market, it can take one bad event to cascade onto everyone else by surprise.
Also, everbody is invested in the same stocks. With ETF this has accelerated and I decided to find some stocks that are underrepresented in ETFs for various reasons and still have a solid, old-fashioned business model.
So far most of the use cases of AI I see in the wild has been some stupid chat bot, replacing specific job groups or magically solving a problem with AI.
All of these attempts fail miserably. It‘s Blockchain and NFTs all over again. The promises are so far fetched from current capabilities that it is not even funny.
Not saying that AI is useless, there are many impressive achievements made possible by it, but I do think it’s currently overhyped and filled with empty promises.
Depends on the use case. I would argue promising capabilities that are not there yet and can’t be developed by the supplier themselves (for example they just fine tune an existing model) in a reasonable timeframe are practically just as useless.
Other promises like predicting the future or magically solving security are pretty comparable to NFTs.
This^^ title is editorialized and possibly deliberately misleading. Article ends with:
> I will start working on a piece that will assess whether the NASDAQ (primarily) and hence passive investing as a whole is in a tech and AI-infused reflexive bubble.
The piece is now finished and about to be published in the next couple of days..
So the title is not deliberately misleading. If from what you have read and see in markets, you don't believe that the Naz is in an A.I.-led bubble, then ok.
After playing with ChatGPT (using GPT4) and GPT4 API, I'm quite convinced that only 2 things are stopping LLMs from being ubiquitous right now and one of them is simply inference cost. The other is GPT4 being having many intentional guardrails.
That's not to mention GPT5 and 6 and so on coming out.
You can argue that OpenAI might not hold the lead. That's fine. But the author also argued against Nvidia. Nvidia doesn't care if OpenAI loses the lead because the next LLM lab also uses Nvidia.
Note: I'm also invested in the market. Here's my investing strategy at the moment: https://news.ycombinator.com/item?id=39342028#39342362
I'm not a stock picking expert. But I'm investing like I'm going to lose my job as a software dev. If I lose my job, my stocks will surely pay for my food. If I don't lose my job, then it will continue to pay for my food. It's a hedge for myself.