This month I spent $50 on OpenRouter credits. I find it very useful for coding assistance (not coding per se). It's $50 that I never spent before on this kind of service. That's what we call growth - it quite literally came out of nowhere.
And I'd argue the timer started in 2023.
Reminds me of this:
>I went through this Ford engine plant about three years ago, when they first opened it.
> There are acres and acres of machines, and here and there you will find a worker standing at a master switchboard, just watching, green and yellow lights blinking off and on, which tell the worker what is happening in the machine.
>One of the management people, with a slightly gleeful tone in his voice said to me, “How are you going to collect union dues from all these machines?”
>And I replied, “You know, that is not what’s bothering me. I’m troubled by the problem of how to sell automobiles to these machines
- Walter Reuther, Nov. 1956 https://quoteinvestigator.com/2011/11/16/robots-buy-cars
Trust me, there are many other people like me in the world and the enterprises are spending even more.
There is more money flowing into the overall AI ecosystem (by far) than the money Nvidia puts in.
The idea that Nvidia is artificially creating the whole demand is laughable and doesn't add up.
Power tools for knowledge workers, which is what we are getting, isn't enough to save it.
> Trust me, there are many other people like me in the world and the enterprises are spending even more.
Awesome, let's do some math here.
I'll do both $1k/month and $3k/month, please follow along. To make the math even simpler to follow, I'll actually reduce your amounts. ~$800/month will get us about $10k/year and ~1600/month will mean about $20k/year, makes for easier divisions.
Current total investment into AI is at least: https://isaiprofitable.com/ -> $1.8tn.
So $1 800 000 000 000.
From what we know about current AI tech, about ever increasing hardware prices, about ever increasing electricity prices, about the ever increasing DC construction prices, AI companies need to invest a fair chunk of money each year to keep the whole thing going, let's be SUPER conservative and put that amount at $200bn per year.
So:
1 800 000 000 000 + 200 000 000 000 = $2tn next year.
Then at least another 200 000 000 000 per year = $0.2tn/year.
So $2tn next year divided by $10k/year means that means that they will need 200 million yearly subscriptions to recover the money already invested. At $20 k/year would mean 100 million yearly subscriptions. Spread over 5 years that would mean 40 million yearly subscriptions and 20 million yearly subscriptions.
Then for each year, just to cover the costs, at $10k/year 20 million yearly subscriptions would be needed, and at $20k/year 10 million yearly subscriptions are needed.
So that's 60 million yearly subscriptions and 30 million yearly subscriptions.
I used Claude to extract some numbers. The total global addressable workforce that makes more than $80k per year (where an employer would dare spend $10k/$20k per year on AI) is about 100 million people. The total private population that has $10k/20k per year in disposable income is about 500 million people (excluding China, since they will for sure not use Western AIs en masse).
So that's about 600 million users (just stacking private users + how much companies would pay for their workers).
So just to break even each year, 10% of those would need to pay those crazy high subscriptions, and 5% the extra crazy high subscriptions.
For private users if they get 1% of that rate, it means that a lot of private individuals have fallen on their collective heads.
For enterprises, nobody's going to increase their salary expenses from $80k to $90k-$100k for benefits that we can't even quantity, let alone guarantee an upside of 10-25%. That kind of budget will be allocated for people making $150k or above, which makes the total global addressable workforce something like 50 million, most likely less.
I want to have what you're smoking.
Yes, demand is there. Demand to prop up how much we're investing. NO WAY. At actual prices and actual LLM productivity gains, we should probably be investing 20% of what we're investing.
A lot of people will be wiped because of Nvidia and friends. Even worse, a lot of regular people will suffer because we've distorted our societies so much due to this hype train.
If AI makes these workers even 1% more productive, that is $500 - 700 billion value annually. At an ongoing annual $0.5T return, a $2T investment (also over the next few years, note) doesn't seem too bad!
Then consider that actual studies from all the way back in 2024, i.e. the era of spicy autocomplete, before agents landed on the scene, put the productivity boosts much higher, like 30% or more. (Interestingly, this is corroborated by survey based data from the St. Lous Fed: https://www.genaiadoptiontracker.com/) Even assuming a conservative average boost of 10%, that is $5 - 7T value annually.
Add how many ever grains of salt you want to those numbers, the investment is nowhere near as out of whack to the potential revenues as people fear. This is why everybody from Big Tech to VCs to entire nation states are desperately scrambling to get in on the action.
My point is that their math is wrong.
1. There's no way China will let any of the Western frontier labs in, so that's probably 1/3 out of those $50-70tn that they'll never touch.
2. It turns out that LLMs are more of a commodity than expected because the basic tech is basically "Attention is all you need" plus a few things everyone has access to (mixture of experts, caching, batching, etc). So yes, it's a "winner take most" market, but there will likely be a healthy base of cheap models so the "collection" (price gouging) part of the cycle (or enshittification) will be hard to execute.
3. Either hardware remains expensive, in which case every N years entire DCs have to be rebuilt and then Capex needs to flood in - think highway systems being rebuilt, but instead of every 20-30 years for highways, here we'd be talking every 5-7 years.
4. Or hardware becomes cheap in which case cheap LLM hosters are competitive and problem #2 is even worse. Or the nightmare scenario for all of these investment scenarios, local LLMs become viable for most people.
I agree LLMs are already a commodity market, definitely at the non-frontier model level, but I don't think it affects monetization prospects much. After all, server compute is a commodity and yet cloud businesses have been exploding even before AI.
And compute is exactly why it won't be a "winner takes most" market. It is clear now that compute capacity is and will likely remain the biggest moat. Looking at Claude Code is instructive; arguably it was the better product, but it kept going down so much that Codex and other competitors have gained on it. Similarly, China could have the best models, but its access to hardware is deliberately limited by geopolitics, so it's likely their threat will be manageable for a while yet.
A key part of the success of AI companies will be in securing hardware and operating that infra cost-effectively via economies of scale. Hardware will remain expensive for a long time yet because all the hyper-scalers and neo-clouds are severely crunched, and all the fabs (mostly TSMC) are already at capacity even as demand keeps exploding.
And for better or worse, most of that supply will still flow through Nvidia, despite attempts from competitors like TPUs and NPUs, for the simple reason that Nvidia has the monopoly profits to outbid everyone else on the real chokepoint, which is fab capacity.