Rendered at 07:02:20 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
chr15m 6 hours ago [-]
"Inference" is just software running. It has always cost money to run software, it's just that it is generally too cheap to matter. If a client makes a regular API call to your server, you pay for that compute, probably in the form of a flat hosting fee. If too many calls come in and workload goes up, you pay for a more expensive hosting tier to handle it (or do dynamic scaling which is per-unit of compute).
Right now the "hosting" cost for inference is per-unit because it's new and expensive, but that won't last.
There is a lot of inefficiency right now keeping prices elevated. That will change very fast and soon paying for inference will likely resemble paying for hosting your app.
The bigger problem for SaaS is that the floor has risen - people can build their own solutions for things that they used to buy SaaS for. So the industry needs to level up and solve harder problems.
margalabargala 5 hours ago [-]
> There is a lot of inefficiency right now keeping prices elevated. That will change very fast and soon paying for inference will likely resemble paying for hosting your app.
I don't know about "very fast" or "soon" unless you're speaking in geological terms.
SOTA models like Kimi 3 require thousands of GB of RAM/VRAM to run at speeds that are real-time useful. Manufacturing the memory necessary for that quantity to be available at app-hosting prices will take decades. Software efficiency solutions might drop needed memory by an order of magnitude in that time...but a tenth of an enormous amount is still pretty darn big so won't get us there "soon".
j16sdiz 2 hours ago [-]
> That will change very fast and soon paying for inference will likely resemble paying for hosting your app.
The article have talked about this:
> You cannot, however, build a business model today on a cost breakthrough that has not happened yet. Even if costs do fall, Jevons paradox[5] tends to kick in: historically, when a resource gets cheaper, we find more uses for it rather than consuming less. Cheaper inference does not get banked as margin improvement; it gets consumed by deeper integration, more calls per interaction, more autonomous agents running in the background. The cost per call drops but the calls per user multiply.
euazOn 6 hours ago [-]
Exactly my point in another comment. Just to illustrate this further: a rough ballpark of how the cost of intelligence fell since 2022 could be about 1000x, and continues to fall. Unfortunately, it’s really hard to measure.
It’s so cheap that companies choose to spend more on AI inference (more reasoning, more capabilities, longer context), not less - see Jevons paradox.
zmmmmm 6 hours ago [-]
I don't think it's at all certain this won't land back on the same unit economics as the old way. The cost of serving a user doesn't have to be free - it never has been - it just has to not be the dominating factor in your costs. I'm guessing there are still quite a lot of per-user costs that aren't easily visible. Like how many of your users are logging support requests, or suing you, or demanding bug fixes or custom integrations or a myriad of other things. And how much are you having to invest in security updates, regulatory compliance, marketing etc. Not to mention, users are getting well acclimatised to the idea of quotas and paying for increased limits.
smalltorch 14 hours ago [-]
I think part of the new equation may also become; "Why even pay for the SaaS in the first place if you can just forge the service exactly how you want it?" The benefits of unlimited access to the tool you forge are still there, its just a lot easier to make whatever tool you want.
Are there any examples of products containing ai inference that are successful? Products that are beyond just direct access to frontier LLM's, I mean.
anal_reactor 30 minutes ago [-]
If my manager prefers to pay for SaaS while I sit on my ass rather than have me develop an in-house solution, how do LLMs drastically change the equation?
esafak 6 hours ago [-]
As if making all the decisions that go into that product, hosting it, and maintaining it are free! Now do that for all the SaaS you subscribe to. Would you even get any real work done?
nostrademons 5 hours ago [-]
There wouldn't be a product. Consumers would just use the underlying foundational models directly to solve their problem. ChatGPT and Claude and Gemini are the products, not your business.
Products are only viable when you do a lot of work to solve a problem, which is then shared by many potential customers. It makes sense to amortize the high costs of solving the problem across all the different customers to reap economies of scale. Businesses then pay for product design, hosting, and maintenance because those costs can also be amortized, and they are cheaper than a bespoke solution for each customer. But if the bespoke solution becomes cheaper than that, because it's generated by an LLM that doesn't need to be paid a living wage, there's no reason to have the product in the first place. Just solve your damn problem and let other people solve theirs.
This is an underrated factor in the market structure of the AI bubble going on now. Anecdotally, we're not seeing new AI-based products other than the foundational models and coding assistants gain traction. Why? Because AI makes it so easy to customize the solution to your particular needs that everybody is just solving their particular needs directly. It's the opposite of the Internet boom, where the network greatly expanded the potential market, reduced the cost of reaching them, and made it economical to spend large amounts of money building a software product that had a TAM of billions. The AI boom instead enables extremely cheap customization, which shrinks the market to a single customer who uses AI to directly solve their problem rather than building a product that's generally applicable.
esafak 3 hours ago [-]
> Consumers would just use the underlying foundational models directly to solve their problem. ChatGPT and Claude and Gemini are the products, not your business.
How am I going to use ChatGPT instead of, say, Figma or an office suite with my coworkers?? It's no use for it to spin up a clone for me if my coworkers can't collaborate. And if they can, what it will serve is a product.
How can it be cheaper for every company to re-invent the wheel?
> Products are only viable when you do a lot of work to solve a problem, which is then shared by many potential customers.
Yes, that's what businesses pay services for.
DangitBobby 3 hours ago [-]
> How can it be cheaper for every company to re-invent the wheel?
Companies not matching their prices to current reality, mostly.
esafak 2 hours ago [-]
There are already free, open source clones of numerous things but companies still pay for commercial versions. Do you think LLMs that you have to pay for are going to change that?
cwmoore 6 hours ago [-]
So many pivot opportunities, we’ll live remotely from office.
coconido 14 hours ago [-]
[dead]
roncesvalles 6 hours ago [-]
>Users began expecting something fundamentally different from software: not just tools that store and retrieve, but products that reason, generate, and respond.
Not really.
>Every inference call costs money.
Not really, either. If you buy your own GPU, rack it, and run an open model, there is no unit cost. This is just expensive hosting infra. You also pay unit costs for SaaS that your software uses (things like SMS etc).
euazOn 6 hours ago [-]
> If you buy your own GPU, rack it, and run an open model, there is no unit cost.
No. There is economic opportunity cost (borrowing), energy cost, infra cost, depreciation / risk of failure with each unit of work, bandwidth, maintenance, and lots more. Small, but not zero, and often overlooked - especially the opportunity cost.
nostrademons 5 hours ago [-]
These are basically all fixed costs, not unit costs. You buy the GPU once and use it for as many calls as you have traffic for, and depreciate it over a fixed lifetime. You have to power it regardless of whether it's fully utilized or not. You have to maintain it by virtue of owning it, not really based on how many queries it has served. Bandwidth is the only one that really scales as a unit cost.
Open question whether this model is actually more economical than using the cloud AI service. The whole reason the industry moved to cloud computing in the first place was because computing had very high fixed costs, and the more these could be amortized over a fully-loaded query stream, the lower the unit costs.
djsjajah 4 hours ago [-]
You have to power it all the time, but the amount of power it uses while it’s on will change by up to a few orders of magnitude depending on the gpu. It’s not uncommon for a gpu to be pulling just a couple of watts at idles and several hundred at full tilt.
So the only way it’s a fixed cost is if you don’t pay for power. If you only consider the cost of the power, it might still be cheaper paying for an api.
euazOn 5 hours ago [-]
Good points. Speaking from experience, it’s really hard to make it more economical than using a cloud AI service - even if you utilize the GPU to its fullest. There are, of course, other benefits, such as privacy/control/compliance/security, which should be the real reasons to do this, not cost.
blackjack_ 3 hours ago [-]
You sweet sweet summer child.
Spoken like a guy who has never had to maintain bare metal infrastructure ops at scale. These things break, need re-imaged, have parts that break, have to be configured (now you need provisioning pipelines, monitoring, alarms, etc), have to be maintained when something goes wrong (swapping hardware and software, debugging the alarms into actually figuring out which bits are broken and/or misconfigured), have to be catalogued, have to be planned for, deprecated, and the finances accounted for through complicated accounting to show investors the Capex at quarterly meetings.
Then you have to make fleet decisions on how much of each type of server you will want to buy, expanded storage, how long you will support each generation of server, when you will order new hardware, how to order new hardware, lag for real world installs, hiring actual humans to fly around to all of your datacenters and do the actual installations / maintainance / etc. Then you will have to do contracts with individual datacenter operators for margins, electricity rates, hosting contracts, white glove ops hourly rates, etc.
Businesses that own their own hardware tend to have a lot of employees whose jobs are maintaining it and running the business side of that.
noosphr 4 hours ago [-]
You do realize that a fully utilised GPU uses about 10 to 50 times more power than an idling one?
roncesvalles 3 hours ago [-]
How is this different from hosting infra or just SaaS in general?
TFA contends that there is some fundamental shift in the economics of software, but it doesn't look to be very different from either a new SaaS dependency or racking new hardware.
4 hours ago [-]
robocat 4 hours ago [-]
> opportunity cost
We need a better word for this because the things you didn't do are not a cost
dwattttt 4 hours ago [-]
The things you would have done, but now can't do, are a loss to you
gofreddygo 2 hours ago [-]
> Build the product once, distribute it to a million users for roughly the same cost as distributing it to one. Every incremental customer flowed largely to the bottom line.
common misconception about software unit economics. With enterprise software (one that costs real $$$) there always more costs attached post shipping. Before client/server it was support. Then it was security and the constant threat of cyber attacks.
Distributing software is nothing like distributing books.
matchagaucho 2 hours ago [-]
Fortunately, SaaS vendors have already conditioned users to accept usage limits within seat-based plans: “Upgrade to Pro for 50 GB of storage.”
How do we make subscribers become equally comfortable paying for AI usage?
Tokens, credits, inference calls?
SwellJoe 6 hours ago [-]
This is one reason I've been trying to figure out tasks (and products based on those tasks) that can be pushed to the edge, either via small specialized models or small general purpose open models. I suspect the same desire to keep unit costs low is part of why Google is falling behind on the "frontier", but seemingly at the lead, or near it, on models that run on-device. I think they're just focused on making models for tasks that don't require boiling the ocean.
But, it's a hard problem. The models that run locally on normal computers/phones are pretty terrible compared to the frontier, without specialization and fine-tuning. And, even with specialization and fine-tuning, often a high-end general purpose model is going to do a better job and people don't need a bunch of local tools installed to do their various tasks.
skinfaxi 6 hours ago [-]
> And, even with specialization and fine-tuning, often a high-end general purpose model is going to do a better job and people don't need a bunch of local tools installed to do their various tasks.
I think this is the critical point that would be interesting to see if it holds. Technology seemingly tends towards increased specialization.
antonvs 5 hours ago [-]
> Technology seemingly tends towards increased specialization.
The bitter lesson says the exact opposite.
euazOn 7 hours ago [-]
> Meeting that expectation means making LLM calls, and LLM calls cost money.
Of course, and so does everything in the software world. The point is getting the cost so low that it’s basically free. The new DS V4 Flash or the smaller Qwen3.6 models are still really expensive compared to what we were used to in the economics of software, but it’s not unreasonable to expect these costs to continue falling down.
Rough chatgpt estimate says 3-5 orders of magnitude of difference compared to a typical user interaction with a SPA (db/cache lookup, CDN…)
throwaway27448 6 hours ago [-]
> and so does everything in the software world.
Well, no. Copying is free, or so near free it makes zero sense to charge. LLMs are just papering over the damage caused by profit.
euazOn 6 hours ago [-]
Yeah, my point is that LLMs can reach that point too. Especially if you do it clientside. Copying was also way more expensive back in the 60s (accounting just for machine time and electricity, not storage cost, about 100 million times more expensive than today). Everything has a cost.
6 hours ago [-]
throwaway27448 6 hours ago [-]
LLMs will never be as cheap or reliable as copying.
euazOn 5 hours ago [-]
Definitely, of course. But I think the logic of the author’s article is based on there being a huge difference between the two, or rather a high cost of inference in absolute terms. And that can change and we have seen it change. Which breaks the entire premise of the article going forward, no?
Inference will be always more expensive than db operations or copying, sure. But how much more expensive is the question.
mullingitover 5 hours ago [-]
If I'm any kind of indicator of where Youtube users are headed, their AI chatbots in the video pages are going to kill their business model.
There are so many videos with hooks/teasers/'you won't believe what we discovered!!1', and now I just pause the video in the first second, ask "what's the tldr" and get the value from the video without a single ad impression (and likely racking up far more opex for Youtube than if I just streamed the video).
Spooky23 6 hours ago [-]
It depends on the solution. If AI is generating value, you can charge for the value.
Most SaaS already works this way. M365 or Adobe Creative Cloud are great examples. They value it like a life insurance policy and find ways to make you sticky. It’s easier to just buy it.
The first round of AI products suck because they are not well defined. Copilot only makes sense if you do shit in office and SharePoint isn’t a dumpster fire. In my large O365 environment the bottom 50% of users use less storage than the top 2%. So why would i buy copilot for my janitor?
When M365 E9 reconciles invoices automatically with Excel, I’ll pay $150/mo and fire a bunch of people.
carlosjobim 6 hours ago [-]
"Users began expecting something fundamentally different from software: not just tools that store and retrieve, but products that reason, generate, and respond."
Absolutely not. Customers want systems for sales, reservations, accounting, and taking stock. That's where almost all the SaaS money is and none of it benefits from AI - and never will.
horticulturist 6 hours ago [-]
Wonder how much this cost the author to write, as it’s just AI slop…
phendrenad2 6 hours ago [-]
Customers expect more, so they'll pay more. It really isn't more complicated than that.
jrm4 4 hours ago [-]
I strongly predict this article is mostly pointless very soon.
Much as people may not want to like it, "software" as a product to buy and sell, even as a subscription, is probably going away, and will make about as much sense as "math" as a product.
We were already headed in this direction, but AI's going to rapidly accelerate this.
Ozzie-D 6 hours ago [-]
The inference cost discussion is interesting but I think it misses the more consequential shift. The real change in unit economics isn't what it costs to run the model — it's that the marginal cost of building a new feature dropped by an order of magnitude.
Previously the bottleneck was engineering time. Now a competent person with a frontier model can prototype in hours what used to take a team weeks. That compresses the cost side but it also compresses the moat. If your product can be rebuilt by a motivated person in a weekend, your pricing power evaporates regardless of your inference costs.
The SaaS companies that survive this will be the ones whose value comes from network effects, proprietary data, or integration depth — not from code complexity that used to be expensive to replicate.
cleandreams 3 hours ago [-]
I don't think the bottleneck was engineering time alone. There is context, brand, customer support, etc. I am not sure big complex products are particularly vulnerable in the way you state. However I have been thinking of a one smallish problem in a niche space that I could easily code for. There will be more of that.
Right now the "hosting" cost for inference is per-unit because it's new and expensive, but that won't last.
There is a lot of inefficiency right now keeping prices elevated. That will change very fast and soon paying for inference will likely resemble paying for hosting your app.
The bigger problem for SaaS is that the floor has risen - people can build their own solutions for things that they used to buy SaaS for. So the industry needs to level up and solve harder problems.
I don't know about "very fast" or "soon" unless you're speaking in geological terms.
SOTA models like Kimi 3 require thousands of GB of RAM/VRAM to run at speeds that are real-time useful. Manufacturing the memory necessary for that quantity to be available at app-hosting prices will take decades. Software efficiency solutions might drop needed memory by an order of magnitude in that time...but a tenth of an enormous amount is still pretty darn big so won't get us there "soon".
The article have talked about this:
> You cannot, however, build a business model today on a cost breakthrough that has not happened yet. Even if costs do fall, Jevons paradox[5] tends to kick in: historically, when a resource gets cheaper, we find more uses for it rather than consuming less. Cheaper inference does not get banked as margin improvement; it gets consumed by deeper integration, more calls per interaction, more autonomous agents running in the background. The cost per call drops but the calls per user multiply.
It’s so cheap that companies choose to spend more on AI inference (more reasoning, more capabilities, longer context), not less - see Jevons paradox.
Are there any examples of products containing ai inference that are successful? Products that are beyond just direct access to frontier LLM's, I mean.
Products are only viable when you do a lot of work to solve a problem, which is then shared by many potential customers. It makes sense to amortize the high costs of solving the problem across all the different customers to reap economies of scale. Businesses then pay for product design, hosting, and maintenance because those costs can also be amortized, and they are cheaper than a bespoke solution for each customer. But if the bespoke solution becomes cheaper than that, because it's generated by an LLM that doesn't need to be paid a living wage, there's no reason to have the product in the first place. Just solve your damn problem and let other people solve theirs.
This is an underrated factor in the market structure of the AI bubble going on now. Anecdotally, we're not seeing new AI-based products other than the foundational models and coding assistants gain traction. Why? Because AI makes it so easy to customize the solution to your particular needs that everybody is just solving their particular needs directly. It's the opposite of the Internet boom, where the network greatly expanded the potential market, reduced the cost of reaching them, and made it economical to spend large amounts of money building a software product that had a TAM of billions. The AI boom instead enables extremely cheap customization, which shrinks the market to a single customer who uses AI to directly solve their problem rather than building a product that's generally applicable.
How am I going to use ChatGPT instead of, say, Figma or an office suite with my coworkers?? It's no use for it to spin up a clone for me if my coworkers can't collaborate. And if they can, what it will serve is a product.
How can it be cheaper for every company to re-invent the wheel?
> Products are only viable when you do a lot of work to solve a problem, which is then shared by many potential customers.
Yes, that's what businesses pay services for.
Companies not matching their prices to current reality, mostly.
Not really.
>Every inference call costs money.
Not really, either. If you buy your own GPU, rack it, and run an open model, there is no unit cost. This is just expensive hosting infra. You also pay unit costs for SaaS that your software uses (things like SMS etc).
No. There is economic opportunity cost (borrowing), energy cost, infra cost, depreciation / risk of failure with each unit of work, bandwidth, maintenance, and lots more. Small, but not zero, and often overlooked - especially the opportunity cost.
Open question whether this model is actually more economical than using the cloud AI service. The whole reason the industry moved to cloud computing in the first place was because computing had very high fixed costs, and the more these could be amortized over a fully-loaded query stream, the lower the unit costs.
So the only way it’s a fixed cost is if you don’t pay for power. If you only consider the cost of the power, it might still be cheaper paying for an api.
Spoken like a guy who has never had to maintain bare metal infrastructure ops at scale. These things break, need re-imaged, have parts that break, have to be configured (now you need provisioning pipelines, monitoring, alarms, etc), have to be maintained when something goes wrong (swapping hardware and software, debugging the alarms into actually figuring out which bits are broken and/or misconfigured), have to be catalogued, have to be planned for, deprecated, and the finances accounted for through complicated accounting to show investors the Capex at quarterly meetings.
Then you have to make fleet decisions on how much of each type of server you will want to buy, expanded storage, how long you will support each generation of server, when you will order new hardware, how to order new hardware, lag for real world installs, hiring actual humans to fly around to all of your datacenters and do the actual installations / maintainance / etc. Then you will have to do contracts with individual datacenter operators for margins, electricity rates, hosting contracts, white glove ops hourly rates, etc.
Businesses that own their own hardware tend to have a lot of employees whose jobs are maintaining it and running the business side of that.
TFA contends that there is some fundamental shift in the economics of software, but it doesn't look to be very different from either a new SaaS dependency or racking new hardware.
We need a better word for this because the things you didn't do are not a cost
common misconception about software unit economics. With enterprise software (one that costs real $$$) there always more costs attached post shipping. Before client/server it was support. Then it was security and the constant threat of cyber attacks.
Distributing software is nothing like distributing books.
How do we make subscribers become equally comfortable paying for AI usage? Tokens, credits, inference calls?
But, it's a hard problem. The models that run locally on normal computers/phones are pretty terrible compared to the frontier, without specialization and fine-tuning. And, even with specialization and fine-tuning, often a high-end general purpose model is going to do a better job and people don't need a bunch of local tools installed to do their various tasks.
I think this is the critical point that would be interesting to see if it holds. Technology seemingly tends towards increased specialization.
The bitter lesson says the exact opposite.
Of course, and so does everything in the software world. The point is getting the cost so low that it’s basically free. The new DS V4 Flash or the smaller Qwen3.6 models are still really expensive compared to what we were used to in the economics of software, but it’s not unreasonable to expect these costs to continue falling down.
Rough chatgpt estimate says 3-5 orders of magnitude of difference compared to a typical user interaction with a SPA (db/cache lookup, CDN…)
Well, no. Copying is free, or so near free it makes zero sense to charge. LLMs are just papering over the damage caused by profit.
Inference will be always more expensive than db operations or copying, sure. But how much more expensive is the question.
There are so many videos with hooks/teasers/'you won't believe what we discovered!!1', and now I just pause the video in the first second, ask "what's the tldr" and get the value from the video without a single ad impression (and likely racking up far more opex for Youtube than if I just streamed the video).
Most SaaS already works this way. M365 or Adobe Creative Cloud are great examples. They value it like a life insurance policy and find ways to make you sticky. It’s easier to just buy it.
The first round of AI products suck because they are not well defined. Copilot only makes sense if you do shit in office and SharePoint isn’t a dumpster fire. In my large O365 environment the bottom 50% of users use less storage than the top 2%. So why would i buy copilot for my janitor?
When M365 E9 reconciles invoices automatically with Excel, I’ll pay $150/mo and fire a bunch of people.
Absolutely not. Customers want systems for sales, reservations, accounting, and taking stock. That's where almost all the SaaS money is and none of it benefits from AI - and never will.
Much as people may not want to like it, "software" as a product to buy and sell, even as a subscription, is probably going away, and will make about as much sense as "math" as a product.
We were already headed in this direction, but AI's going to rapidly accelerate this.
Previously the bottleneck was engineering time. Now a competent person with a frontier model can prototype in hours what used to take a team weeks. That compresses the cost side but it also compresses the moat. If your product can be rebuilt by a motivated person in a weekend, your pricing power evaporates regardless of your inference costs.
The SaaS companies that survive this will be the ones whose value comes from network effects, proprietary data, or integration depth — not from code complexity that used to be expensive to replicate.