Dear Folks,

For some time, as I have mentioned, I have wondered whether or not most people will be willing to pay for the true cost (cost per token) of the AI services they actually use.

Now it seems that Microsoft, at least one major player, is making the shift to begin charging users for the real cost of the AI services each customer uses.

According to most information we read, at this time, most AI companies are hugely subsidizing the cost of the AI services they provide to their customers. How much longer can customers expect a free lunch?

How much would you pay for AI?

I pay per month for AI, as part of my Google Drive subscription.

But I have always asked myself whether my monthly subscription payment actually is enough to cover the GEMINI AI Services that I use. I haven’t a clue, because Google has never told me the real costs of providing the Gemini AI Services, which are also embedded in Google Search, and much more, such as Google DOCS, and Google SHEETS.

Seems to me that this transition in order to have customers actually pay for the cost of what they use each month, or by tokens used, will be a game changer….

 Do you think this is correct? Or, do you believe in Free Lunch for Everybody?

Here are two excellent vids from ED which speak to both the cost issue, as well as whether people even like AI as much as Sammy-Boy mistakenly believes.

Happy viewing….

Best regards,

Gamma

Two things come to mind.

If they could really be charging at scale for AI, they would already be doing it instead of effectively subsidizing the cost of public usage. I think the general model will stay the same with all of them, a free access tier and a paid access tier with higher capability. They could of course keep reducing the free access tier until it becomes crippled and close to 0% usable, but the space is too competitive for that to be a sustainable long term strategy.

You see the same pattern across other digital services. Webmail, cloud storage, password managers, antivirus and VPNs all charge, but many have free versions, and in many cases those free tiers are still perfectly adequate for me and a large number of people out there decades later.

The only thing I personally pay for is a VPN, and that is very cheap, around $3 a month. I also pay for web hosting, but that is still relatively inexpensive and necessary for what I do. Beyond that I do not really pay for most digital services, no AI subscriptions, no cloud storage (except $1 a month to Apple to backup my iPhone data), and no music or movie streaming.

I do pay for an outgoing mail server, but that is only about $15 a year, mainly because standard SMTP setups do not retain sent mail by default. I also pay for a couple of VoIP phone numbers in different countries for making overseas calls and receiving SMS messages, again very inexpensive. In total, I am probably around $175-$200 a year for everything combined, maybe less, I never really added it all up.

When you compare that to AI subscriptions, many of them are priced around $250 a year for a pro tier. For me that is hard to justify when that is what I am paying now for all my needed digital services combined. AI is very useful, but not 100% reliable. If they were significantly more accurate and cheaper, I would consider a pro plan, possibly with Gemini. ChatGPT has a cheaper GO plan at about $8 a month, which is more reasonable, but I find it less accurate overall compared to Gemini, Grok and Claude. I put ChatGPT in last place and why I wont pay for it, though its text editing is good and image generation has just recently improved, but I would not pay for an AI service purely for image or text editing features, it needs to be good at everything, especially giving accurate analysis.

So as things stand, I am not really a buyer at current price levels. If Gemini came down in price I might reconsider, but I do not expect that to change soon, so I expect to stay where I am with not paying for AI services.

56 minutes ago, FriscoKid said:

If they could really be charging at scale for AI, they would already be doing it instead of effectively subsidizing the cost of public usage. I think the general model will stay the same with all of them, a free access tier and a paid access tier with higher capability. They could of course keep reducing the free access tier until it becomes crippled and close to 0% usable, but the space is too competitive for that to be a sustainable long term strategy.

I perfectly understand what you are saying.

However, it seems the reality is even worse that you say:

It does seem that NOBODY is paying for what they are using.

The people who are paying USD200 per month are actually using maybe USD2000 worth of tokens.

This seems crazy and not sustainable.

Therefore, how will these companies be able to provide what they are providing and in the way they are providing it...without going broke.

If it is true that Microsoft Pilot, for example, has shifted to charging by "Tokens Used", then will this work?

Will users actually be willing to pay ten times as much for the same service, just because they must pay for real cost?

The first video explains this conundrum.

Edited April 27Apr 27 by GammaGlobulin

I watched a bit of the first one. It got boring pretty quick for me. The underlying message is basically the same in all of these videos. The AI business model is unsustainable. There is a lot of money going in and very little coming out. You can tell the same story in a million different ways, but it keeps coming back to that same point. It will collapse, probably sooner rather than later. OpenAI will likely be one of the first big losers and then the rest of the dominoes will fall. Six to twelve months, eighteen at most, before the fresh money going in dries up. You can watch these kinds of videos all day, but it is the same repeating narrative and it is already a well known phenomenon. Nobody is going to be surprised when it all collapses, so I see little point in continuing to watch the same repeating story.

Sure why not

No

Imagine paying for something that gets it wrong so often. Peak comedy.

2 minutes ago, Nemises said:

Imagine paying for something that gets it wrong so often. Peak comedy.

So what do it get wrong ? What is the main problem ?

AI is a great tool to widen your horizon and digg deeper, and if you are sceptical, just continue searching with available sources. But if you believe everything is wrong with this world, of course you will see wrong everywhere.

2 minutes ago, Hummin said:

So what do it get wrong ? What is the main problem ?

AI is a great tool to widen your horizon and digg deeper, and if you are sceptical, just continue searching with available sources. But if you believe everything is wrong with this world, of course you will see wrong everywhere.

Be careful asking it if something's edible 🙃. Refer Celsius's post above.

8 minutes ago, Hummin said:

AI is a great tool to widen your horizon and digg deeper, and if you are sceptical, just continue searching with available sources. But if you believe everything is wrong with this world, of course you will see wrong everywhere.

Why buy the cow when the prompt box gives away free milk? 🥛🚀

7 minutes ago, Nemises said:

Be careful asking it if something's edible 🙃. Refer Celsius's post above.

Even if AI is a great tool, it also comes with consequences. Darwin’s law never ceases to exist.

I already pay for AI, and I find it great for digging deeper into history, science, and philosophy. It is also a very useful tool for people with reading and writing difficulties, and for explaining things when there is something you do not understand.

Even when you disagree with it, you can ask for sources and make AI dig deeper to get closer to the truth. But yes, sometimes it makes mistakes.

It can also help you build small apps, edit pictures and videos, and turn ideas into something practical. For the younger generation, learning AI will almost be a must. For us in middle age, it is probably better to jump on it before it feels too late.

I have seen my own parents struggle with online banking and official services because everything became too complicated too fast. The same thing can happen again with AI. The longer you wait, the more you delay something that will probably become part of everyday life anyway.

It is also great for writing memoirs from your own life. You can write down what you remember, go back later, add more memories, and build everything neatly, chapter by chapter.

12 minutes ago, Hummin said:

Even if AI is a great tool, it also comes with consequences. Darwin’s law never ceases to exist.

I already pay for AI, and I find it great for digging deeper into history, science, and philosophy. It is also a very useful tool for people with reading and writing difficulties, and for explaining things when there is something you do not understand.

Even when you disagree with it, you can ask for sources and make AI dig deeper to get closer to the truth. But yes, sometimes it makes mistakes.

It can also help you build small apps, edit pictures and videos, and turn ideas into something practical. For the younger generation, learning AI will almost be a must. For us in middle age, it is probably better to jump on it before it feels too late.

I have seen my own parents struggle with online banking and official services because everything became too complicated too fast. The same thing can happen again with AI. The longer you wait, the more you delay something that will probably become part of everyday life anyway.

It is also great for writing memoirs from your own life. You can write down what you remember, go back later, add more memories, and build everything neatly, chapter by chapter.

Agreed. But the topic's about paying for it. As I already said - why buy the cow when the milk's free.

2 minutes ago, Nemises said:

Agreed. But the topic's about paying for it. As I already said - why buy the cow when the milk's free

It is the options that come with paying, the same as with everything else.

If you want better quality, more functions, and more reliable access to all services, you usually have to pay. Free versions are useful, but they come with limitations.

8 minutes ago, Hummin said:

It is the options that come with paying, the same as with everything else.

If you want better quality, more functions, and more reliable access to all services, you usually have to pay. Free versions are useful, but they come with limitations.

True, but for most people, paying for AI is like buying a Ferrari just to drive to the grocery store. If you're not a power user hitting those limits every day, then $30 a month (or whatever) is a total waste. The free versions are already so good now that most of us wouldn't even notice the 'premium' features if we had them.

24 minutes ago, Nemises said:

True, but for most people, paying for AI is like buying a Ferrari just to drive to the grocery store. If you're not a power user hitting those limits every day, then $30 a month (or whatever) is a total waste. The free versions are already so good now that most of us wouldn't even notice the 'premium' features if we had them.

Sorry for the long answer, but this is also how you can work with AI. You can use it to clean up your thoughts, question your own claims, ask for better sources, and make an argument clearer.

And the discussion is real. What do we actually miss when we only use free apps, free AI, free news, and free information? And just as important: who controls the information we see?

The same applies here. We have to trust that moderators and owners do their job with discretion, and that they do not abuse our information or control the discussion unfairly. Free access is useful, but it often comes with limits, and sometimes those limits are not only about money. Sometimes they are about control, quality, access, and trust.

However I agree with you, still there is a bit like only reading free news sources and free articles. You still get access to a lot, but not everything.

I think that is where the limits will be in the future too. The free versions of AI will be useful, but they will not give you access to all tools, all sources, or the highest quality services. The best information, the best models, and the best functions will most likely sit behind some kind of payment, subscription, or license.

So yes, the milk may be free, but not all the milk, and not always the best milk.

And just to fact check my answer:

AI says

Fact: yes, mostly.

Your point is correct: free AI will most likely be like free news, free apps, or free software. Useful, but limited. The paid versions usually give more access, better tools, higher limits, file uploads, image generation, data analysis, tasks, and other functions that free versions either limit or do not include. OpenAI’s own pricing page already shows clear differences between free and paid plans, including limits on data analysis, file uploads, image generation, tasks, and access to more advanced tools.

Also, your comparison with free news is good. A lot of high-quality information is already behind paywalls, subscriptions, licenses, or private databases. Publishers are now working on licensing systems so AI companies must pay for premium content instead of just scraping everything freely.

Yes, your point is basically right, but I would be careful with the “40% is fake news” number. That exact number is hard to prove, because it depends on what you count as fake news: completely false stories, misleading headlines, propaganda, AI images, biased framing, or just badly checked claims.

And note it corrected my claim about 40%, something I read somewhere someplace.

Still it is sad that quality news is becoming something mainly available to those who can afford access. Before, newspapers and magazines were often lying around in cafés, waiting rooms, hotels, libraries, workplaces, and public places. Even if you did not buy them yourself, you could still read them.

Now much of that has moved behind paywalls, subscriptions, and apps. For many people, the free alternative becomes social media, where there is far less editorial control and where false, emotional, or manipulated content can spread very quickly.

Fake news on social media is a serious problem, and so far the platforms have not shown that they are willing, or able, to deal with it properly. And when they do interfere, we still have to ask who decides what is true, what gets removed, and in which direction that control will move. That is a scary future in itself.

UNESCO has warned that social media has accelerated the spread of false information and hate speech, and one UNESCO survey found that 62% of digital content creators do not systematically fact-check information before sharing it.

And now we have another problem: what we see is often not even real anymore. Pictures, videos, voices, and stories can be edited or generated. People already believed too much nonsense before AI. Now the nonsense can look professional.

So what can possibly go wrong? Quite a lot. History has already shown us that people are willing to believe things without proof: religion, myths, ideology, propaganda, conspiracy theories, and political stories that divide people instead of helping society.

The danger is not only that people believe lies. The bigger danger is that they stop knowing what to trust at all. The Reuters Institute’s 2025 Digital News Report describes a media world increasingly shaped by social media, synthetic content, and misinformation, while still showing that people value trusted news brands when they can find them.

When quality information becomes expensive, and cheap information becomes emotional, fake, or manipulated, society has a real problem. People do not only need access to information. They need access to information they can trust.

1 hour ago, Hummin said:

I already pay for AI, and I find it great for digging deeper into history, science, and philosophy.

Yes.True.

BUT, you are NOT paying for the actual cost of the AI that you use.

You are paying for a subsidized platform.

We do not yet know the real cost to users.

They do not tell us how much of what we use is subsidized by a company losing money on the deal.

This cannot go on.

So, the question is: Will we pay for the real cost, the total cost, and profit margin?

10 minutes ago, Hummin said:

So yes, the milk may be free, but not all the milk, and not always the best milk.

The point is:

a. The Free Offerings are coming to an end.

b. What will people do if what was once Free now costs Money?

c. This is the question of the Topic.

d. This is the subject of the attached videos, two of them.

20 minutes ago, GammaGlobulin said:

Yes.True.

BUT, you are NOT paying for the actual cost of the AI that you use.

You are paying for a subsidized platform.

We do not yet know the real cost to users.

They do not tell us how much of what we use is subsidized by a company losing money on the deal.

This cannot go on.

So, the question is: Will we pay for the real cost, the total cost, and profit margin?

15 minutes ago, GammaGlobulin said:

The point is:

a. The Free Offerings are coming to an end.

b. What will people do if what was once Free now costs Money?

c. This is the question of the Topic.

d. This is the subject of the attached videos, two of them.

That is how most technology markets work. The real price will eventually find its level, so we do not need to speculate too much yet.

New technology will come, competition will grow, and the cost may also come down over time. Honestly, we do not know the future or what comes next.

For now, AI is available, many can use it for free, and some choose to pay for more options. If the value stays strong enough, people will continue. If it becomes too expensive, people will stop paying or move to alternatives.

In comparison:

I paid more than $15,000 for a computer and editing programs back in the early 2000s. Adjusted for inflation, that is close to $28,000 today.

Now you can buy a strong Mac for around $2,000–$3,000 and subscribe to professional editing software instead of buying everything outright like we had to do before.

1 hour ago, GammaGlobulin said:

The point is:

b. What will people do if what was once Free now costs Money?

Bit of a big ‘if’ that though. Free tiers might get tighter, but AI going fully paywall-only across the board isn’t likely. Feels more like another speculative doom-thread than anything concrete tbh 😃

55 minutes ago, Hummin said:

That is how most technology markets work. The real price will eventually find its level, so we do not need to speculate too much yet.

New technology will come, competition will grow, and the cost may also come down over time. Honestly, we do not know the future or what comes next.

For now, AI is available, many can use it for free, and some choose to pay for more options. If the value stays strong enough, people will continue. If it becomes too expensive, people will stop paying or move to alternatives.

In comparison:

I paid more than $15,000 for a computer and editing programs back in the early 2000s. Adjusted for inflation, that is close to $28,000 today.

Now you can buy a strong Mac for around $2,000–$3,000 and subscribe to professional editing software instead of buying everything outright like we had to do before.

You should NEVER buy anything other than a Linux computer, these days.

Use OpenSUSE Leap 16.

Then, buy the components you need.

And then, assemble the entire thing.

Very easy to do.

And, there are so many more reasons why Linux is better than Windows or Apple these days.

Unfortunately, the DRAM prices, and memory in general, is skyrocketing.

Building a a computer this year is a far different prospect than last year, July, for example.

18 minutes ago, Nemises said:

Bit of a big ‘if’ that though. Free tiers might get tighter, but AI going fully paywall-only across the board isn’t likely. Feels more like another speculative doom-thread than anything concrete tbh 😃

You will find out, soon enough, that you will need to pay for what you use.

The only question is: Will you pay?

Only you can decide this question.

20 minutes ago, GammaGlobulin said:

You will find out, soon enough, that you will need to pay for what you use.

The only question is: Will you pay?

Only you can decide this question.

Maybe 👍 but that’s not really the point. Most digital services have free tiers, paid tiers, and competition keeps things in balance. ‘Everyone will have to pay for everything soon’ has been predicted for years… hasn’t quite played out that way.

If it changes, people will adapt like they always do. Until then, it’s just speculation dressed up as certainty 😄

At the moment AI still needs training. It's just not reliable enough so you can trust it fully. You have still to guide it along the way to get the result. Like a child that has all the knowledge but I you have to lead them to the result.

Example: I wanted to do a repair on my car. So I asked AI to show me a step by step guide, if possible with pics or even a video. I got it, but from a totally different random car, useless. So I had to correct it, ask more questions and in the end I got at least a step by step that was useful. So I still have to teach AI, at least in parts. Btw. I use a paid subscription.

That was just one example. There are many more. They may ask more money and cover the cost from the day they provide trustworthy results every time. That is not the case now.

1 hour ago, GammaGlobulin said:

You will find out, soon enough, that you will need to pay for what you use.

The only question is: Will you pay?

Only you can decide this question.

I don't use it and will NEVER pay for it.

Money in AI will not come from individual people but from business.

I am already being bombarded with companies promising to streamline my business processes, speed up mundane tasks and do it all with less people. It will be sold as bespoke and certainly at premium prices. We have already seen this with Anthropic withholding its latest and most advanced AI model, internally dubbed Claude Mythos, from general public release (but talking to some big corporations), citing it as 'too powerful/dangerous" when it showed a capability to find and exploit software vulnerabilities at a level superior to humans. They have put a lid on it so far as it demonstrated worrying capabilities when it 'escaped' its sandbox, tried to cover its tracks and attempted to hide its actions from researchers. This is yet another reason why this should all be so much better regulated and put in the hands of people with an idea of what the hell it is capable of, but without a finacial incentive to just let it all loose.

These companies will eventually have to monetise their products due to the vast investments they have already made but much like Facebook when it first hit the scene, how exactly that will come about will depend on where the product itself leads them and who in turn is willing to pay for such a competative advantage. I suspect business to be at the forefront.

Let's just hope the don't just let the genie completely out of the bottle, to the detrament of all mankind.

14 minutes ago, johnnybangkok said:

Money in AI will not come from individual people but from business.

Companies will NOT continue to lose huge money providing free products to individuals.

Did you watch the vids I attached to this Topic?

What seems like the most likely outcome is that OpenAI will fail financially. ChatGPT could get taken over or absorbed by a competitor. Anthropic could face a similar fate, although it seems more likely to survive than OpenAI.

So the AI space could end up being reduced mainly to Google and possibly Grok. I don’t think either of those will fail for various reasons. Apple could also end up buying OpenAI or Anthropic and rebrand one as their own AI product and eventually part ways with the upcoming Gemini integration into Siri that is scheduled for later this year.

If companies like Microsoft can’t make their AI services profitable, they may just walk away. Big tech companies are known for dropping projects that don’t work and moving on. Meta just did something similar with the metaverse.

So the situation is fluid and evolving. Some companies will fail, some will survive, and others will merge. I don’t think there will be a total collapse of the industry because the corporate need for AI as a product is real and still growing. The most likely failure, in my view, is OpenAI. The rest will probably adapt, find ways to monetize, or shift direction.

The “AI bubble” might not end in a dramatic crash. It could instead play out as a series of mergers, adjustments, and repositioning.

5 hours ago, Hummin said:

So what do it get wrong ? What is the main problem ?

AI is a great tool to widen your horizon and digg deeper, and if you are sceptical, just continue searching with available sources. But if you believe everything is wrong with this world, of course you will see wrong everywhere.

I'm with you and don't understand where some of these posters are coming from. My guess is that they have just brushed the surface of AI's capabilities and found it flawed.

It is wrong often and what you can learn in the process of working with AI on a problem is indescribable and borderline miraculous. I have custom AI modules for most things that are important in my life. Each module has instructions and I provided with the appropriate information about me and the context AI needs. I adjust AI in each module to meet my needs. Some I give specific instruction to find my blind spots and other I ask to be my personal doctor being they have decades of data on me that I provided.

I could go on forever - it is incredible and I would pay 20k-60k baht anually if I had to for AI. I don't use the free plans so maybe that is part of my disconnect. Maybe they are less functional.

1 hour ago, GammaGlobulin said:

Companies will NOT continue to lose huge money providing free products to individuals.

Not sure what you are getting at here when I clearly stated 'These companies will eventually have to monetise their products due to the vast investments they have already made...'

Take it from the start, Would internet have had the success it has, if Google or other services were paid ? Definately not. So if they want to make users pay for Ai, fine,. But they will have far less users so up to them.