back
93 comments
> If token costs converge toward zero for most AI use cases...

In the real world, token costs seem to be going up, as early stage pricing at a loss gives way to pricing that generates revenue.

Compute costs might go down a little over the next five years, but there's nothing coming along in hardware that leads to huge reductions in price. NVidia says don't expect better price/performance before 2030.

The models keep getting bigger, and people put loops around them which iterate, burning tokens.

Where is this cost reduction coming from?

I've switched to non-SOTA models, which deliver comparable value at a fraction of the cost. A full day of coding with Deepseek is approx $1 in tokens, and at least for my use cases the quality is equivalent to Claude.
If you don't mind sharing... what's your preferred vendor for Deepseek
Hm. I must be holding it wrong. I hit $20 easily in 5 hours with deepseek in opencode.
I don't expect any of the third party openrouter providers sell tokens at a loss. Agreed that increasing model size could drive token prices up however so far there's been a very strong trend in the opposite direction with smaller models becoming increasingly capable thanks to advances in theory and implementation.

Edit: A glaring omission on my part there is that growth of aggregate industry demand for tokens has the potential to outpace increases in supply provided by new datacenters buildouts. So tokens certainly could go up depending on how things play out.

The following two things can be true at the same time:

- Frontier state-of-the-art performance keeps getting more expensive (and better).

- Any fixed performance level is becoming cheaper.

(And a third: if you still want to see improvements instead of a fixed level, you can trail the frontier a bit and still see price some reductions over time.)

This website https://tokenpriceindex.com/ tracks Token Cost.

You are right - tokens are going up currently.

That chart is a blended metric though. I don't think anyone is using the latest and greatest frontier models (which seem to get more expensive with each generation) for mundane tasks that are solvable with existing models. It seems unreasonable to average deepseek v4 flash and fable pricing.
Don't index on tokens. Costs for useful unit of work had gone down in my experience, and that's what actually matters.

Counting tokens is like counting lines of code.

Article is authored by a private credit firm who assigned absurd valuations to AI infra (see below as an even super recent example) - what is their intention by writing this (is it due to the AI fear narratives around software investments they also hold that drive ppl to withdraw their holdings in Apollo?)

From last month: https://peinsights.substack.com/p/apollo-and-blackstone-clos...

The premise is flawed.

"The first chart below shows that so far there are no signs of profit margins rising outside the tech sector. This is ultimately what we are waiting for, because the value of AI companies today rests entirely on the promise that margins in the S&P 493 will eventually climb."

This is absolutely not necessary. The bull case is that AI will bring great efficiencies. The surplus profits from those efficiencies could easily be competed away by firms who have adopted AI. Those firms who do not adopt AI will have their margis crushed.

> The surplus profits from those efficiencies could easily be …

… usurped by the tech companies?

So then your argument would be that we could see a bifurcation in the SP493 where those who adopt AI see increasing margins and those who do not have their margins crushed. What's funny is that in that scenario, the aggregate market might look zero sum.
What does this look like for any given company? Which margins will you be crushed by for not adopting?
Labor, obviously. That's where all the money in a business goes: paying pesky human employees.

If your employees can suddenly magically do more work with the same pay, that's free money (for you). You can pay fewer employees, or pay them less by threatening to replace them with the magic robot.

The magical thinking version of this is that your productivity gains magically translate into more customers and more sales for the same input cost and labor. The free money is really free because you're a magical special snowflake company and every consumer will want your brand of magic machine outputs and not the other guy's. Where does all this money come from? Do those extra customers even exist? Who cares!

Hypothetical:

Pepsi starts using AI in some magical way that allows them to increase their margins. This allows them to reduce prices while increasing profits. Price-sensitive customers switch from Coca Cola products to Pepsi products. Coca Cola loses some market share, reducing economies of scale, and reducing margins, thus reducing profits. As the cycle repeats, Pepsi moves to dominate the market, and Coca Cola is slowly squeezed down.

With enough competition, the surplus will go to consumers (and workers).
Well those efficiency gains have to show up somewhere. It would imply that consumers / customers of these companies are receiving cheaper or higher value services / goods.

Thats at odds with current inflation trends to say the least.

Even aside from inflation, the prospect of efficiency-borne gains meaningfully benefiting the consumer rather than fattening corporate profit margins, frankly, seems like magical thinking. I’ve seen no evidence that our current corporate culture is capable of it (for any longer than it takes to dominate some market.)
Maybe there's an argument that a lack of rising profit margins in non-tech companies is a bad sign for AI, but this article doesn't make it. Why can't we have a red-queen's race where non-tech companies are implementing AI, but it's not increasing the total profits of those sectors, just meeting rising customer demands/fighting over the share of existing profits? (Never mind that if you look at that chart, profit margins aren't static to begin with, so you can't isolate AI impact from normal fluctuation).

Now, on the first order point, I agree that non-tech companies seem to be taking longer to see results from AI, even if the argument was bad.

I work on SaaS for the logistics space, and I feel like prior to the end of 2025, almost all the discussion about AI for logistics was vaporware, starting this year, companies are actually trying to deploy agents, and we'll start finding out what the ROI is later this year or next.

> Why can't we have a red-queen's race where non-tech companies are implementing AI, but it's not increasing the total profits of those sectors, just meeting rising customer demands/fighting over the share of existing profits?

But then if this happens - all of the stock market has risen in the promise of AI. If AI eats profits instead of grows them, then the economy shrinks right? So maybe that’s worse? That there is no productivity increase?

> If AI eats profits instead of grows them, then the economy shrinks right?

No, why? The economy is bigger than company profits. Eg there's workers' wages and customer surplus and investments etc.

> I agree that non-tech companies seem to be taking longer to see results from AI

Rewriting your app in rust won't increase your revenue, it will cost you more in terms of tokens and risk increasing.

One thing I can't square: if the cost to build an application goes to zero, we should see a proliferation of apps, especially from the AI labs.

The fact that we aren't seeing an app explosion (I think) is evidence that building applications people will pay for is significantly more complex than just prompting claude/codex/etc

I am absolutely seeing an explosion in apps. The reason you might not see them is because the app explosion is entirely custom and in house.

I talked with a friend last week, who has never coded before in his life, who built an absolutely incredible fit-for-purpose app for his own job. He gave me a demo and it blew my mind. It will never go beyond his walls, and he will never buy SaaS that only kinda fits what he needs.

I see things like this happening. The proliferation isn't public because why sell it? Just build the thing to make your domain job easier and save thousands per month cancelling SaaS subs.

The ROI of AI is starting to show, but it isn't in terms of growth or selling new things - it's reducing spend across the board on software and tools.

From their profile, this person makes a living selling AI programming products, by the way. Who could have guessed. There's a pattern to be noticed, even.
I’m seeing this too. I compare it to spreadsheets in terms of getting broad application building tools to the layperson
Anecdotally, Claude Code has prompted an explosion of open source projects and prototypes from self-starters. A lot of these are just hobby projects, but some of them genuinely fill a niche that was previously too complicated or unviable to develop otherwise.

Some of them have half baked financial models, but nobody will invest dollars backing a SaaS offering that could easily be replicated, or that could be made redundant tomorrow.

I made an app for myself and the local MTB community for keeping track of rain and soil moisture for nearby trails so it's easier to decide when a trail will likely be open. Much more reliable than waiting for the official (volunteer led) organization to update the status. I never would have made it without an LLM to speed things along.

A good friend of mine helped his mom keep track of Meals On Wheels (or a similar volunteer org) orders, deliveries, cancellations, etc. They were managing all of this via paper before.

I compiled a list of online recipes. Then I had an LLM typeset them for me into a printable PDF and build a companion website with links to the original recipes and complete ingredient lists for shipping. had the LLM encode links for the companion site into QR codes so the printed copy of the cookbook would bring me immediately to a shopping list, making trying a new recipe soooo much less daunting.

There are so many little things like this that you can make that just take too much effort to justify otherwise. I have other ideas for personal projects that I'll probably get to some day.

Speaking from my side of the industry (the gaming industry), we are seeing a massive increase in the number of games people are making. Above the growth that was already there.

The distinction is that the games being made are garbage, and I mean worse than shovelware garbage. It's actively made things much harder as someone that fancies himself an indie game curator because you gotta dig through more and more games to find stuff with actual people behind it.

For a long time nobody knew how to monetize OSS outside of a few Linux vendors.

There's a crapload of new repos and Github and similar things. And a lot of it is "hobby utility" stuff like you'd find everywhere pre-mobile/pre-app-store but kinda dried up a bit with the browserfication+phone-ificiation of everything. Everything had to turn into an app + an online service.

Now, like OSS, freeware, and even most shareware in the 90s, most of these new projects have no path to VC-level interest.

The whole "basic business or business-process BUT ON THE INTERNET with a dash of social/web-2.0/personalization/crypto/fad-of-the-year" that recent VC firms have been pushing for the last 15+ years may be numbered.

But it's also unlikely that growing companies with big ambitions will want to base their business on vibe-coded free software for too long. It opens up too many unknowns/risks ("oh no, the disgruntled employee leveraged a misconfiguration in our in-house accounts payable system!") There will be a new middle ground model to be found.

There are a lot of specialty/niche apps showing up which are vibe coded --- tons of 3D CAD apps which are a variation/extension of OpenSCAD, a fair number of tools which work with G-code in various ways, &c.

On a commercial support forum I moderate we had to ban software announcements there were so many.

I think a) the labs are releasing very fast and b) why would they implement the long tail of app features when they can effectively sell tokens to every user to write their own version of the app, which is what is currently happening?
I think we actually are seeing an app explosion, just not a consumer app explosion.
I think there is an explosion of new apps, the problem is still distribution in marketing. If I develop a new vibe coded app, it will still take some time for it to be known. And also get good.

Also, what another commenter said that most of new apps are in-house, fully agree with that.

> especially from the AI labs

IMO this is one of the endgame for big ai LABS, they will allow and subsidize users to test and validate on their behalf and once there is a PMF they will step in.

I’ve seen a ton of new open source slop programs. Every day there’s so many “announcing my cool new app” posts. I don’t remember the rate being this high before.
There is no reason to believe the ROI runway is not long inside the tech sector either. I mean, you cannot base that on claims made by the AI sub-sector of the tech sector; of course they are going to claim nothing other than that eating their dogfood is great ROI with a short runway.
If you need immediate ROI (say, because you just invested a trillion dollars into datacenters) you may be out of luck.

And I don't think this is unusual. It took decades for previous technologies to be fully integrated into existing businesses. In the 80s you could see the IT revolution everywhere... except the productivity statistics, which didn't catch up until the 90s.

LLMs are still very new and have significant limitations (like prompt injection and high token costs) that are very likely solveable but will take time.

I don't understand why anyone insists that this needs more time. None of what we've seen in the past few years is new tech. It's more money and hardware thrown at the problem than ever before for diminishing returns.

The market has clearly spoken. Knowing what you're doing is much more valuable than just the doing. That still requires humans. This AI winter has already begun.

>None of what we've seen in the past few years is new tech.

Well that's just wrong. Reasoning models are new and very powerful. LLMs can complete open-ended tasks that require many complex steps.

We're just beginning. The bubble will pop and investors will lose a lot of money, but we're not going back into a winter. It actually works this time.

Ehh, it’s a late AI summer at best. You still need the economic leaves to finish falling off the trees before you can get to the true start of winter.
Microsoft, Amazon are all building forward deployment engineering teams - to increase AI adoption. It will take time, but it will happen.
This is more marketing hype than substance. Amazon isn’t “building” a team, it’s broadly just taking existing people and now calling them “forward deployment engineering teams.”

Also, adoption isn’t lacking because of lack of awareness. Adoption isn’t happening because the math doesn’t add up and the ROI isn’t there. Consulting pixie dust can’t fix that.

Feels like a new consulting wave which is implementation-led by AI vendors taking a large share of the services layer. The existing consulting companies (big 4 etc) will have to shift to niche advisory or heavy channel partnerships with way fewer consultants.
The slower adoption in non-tech sectors isn't just cultural lag - the integration surface is genuinely harder. Legacy ERP systems, compliance review cycles for what data the model can touch, and change management overhead all front-load costs before any efficiency shows up in margins.

I've seen this in payment/API systems: the actual model integration takes weeks, but getting legal and security sign-off on the data pipeline takes months. Non-tech companies face the same pattern but with less internal tooling to manage it.

The margin signal might also be appearing at the wrong level. Gains in these sectors often show up first as headcount flatness or throughput improvements before they hit EBITDA. Measuring at the P&L level on a 2-year horizon is probably too early and too coarse - the operational metrics are moving, the accounting just hasn't caught up yet.