ChatGPT (AI)

[charltjr]

It did take them 17 years to be profitable, most of these companies are less than 5 years old.

That’s fair, but the amounts involved here are crazy. Musk largely funded Tesla himself in the early days, and they IPO’d after seven years. Total investment into Tesla from private equity and raising shares across it’s entire history is less than Anthropic burnt in VC cash in 2025 alone, and Anthropic has raised a further $45bn in VC funding in 2026.

The problem for me is that they have no obvious path to profit. Compute isn’t getting cheaper, their models are getting more complex and consuming more compute, they exist only because huge amounts of cash are constantly pumped into them. Sooner or later, investors need a return.

The big question is will enough businesses pay to use their products when they start having to pay a sensible market rate that allows for some profit, and that’s looking doubtful.

 

[Archtronics]

That’s fair, but the amounts involved here are crazy. Musk largely funded Tesla himself in the early days, and they IPO’d after seven years. Total investment into Tesla from private equity and raising shares across it’s entire history is less than Anthropic burnt in VC cash in 2025 alone, and Anthropic has raised a further $45bn in VC funding in 2026.

The problem for me is that they have no obvious path to profit. Compute isn’t getting cheaper, their models are getting more complex and consuming more compute, they exist only because huge amounts of cash are constantly pumped into them. Sooner or later, investors need a return.

The big question is will enough businesses pay to use their products when they start having to pay a sensible market rate that allows for some profit, and that’s looking doubtful.

I mean sure but the upside is massive for the 2 or 3 companies that come out on top.

Once we get to a place where you have to be using ai to compete as business they’ll be able to turn the screws and jack up the price.

Look at VMware they jacked up the price 800%+ and a lot of businesses have had no choice but the pay it.

 

[Ray Gin]

Musk vs Altman atm, init?

 

[charltjr]

I mean sure but the upside is massive for the 2 or 3 companies that come out on top.

Once we get to a place where you have to be using ai to compete as business they’ll be able to turn the screws and jack up the price.

Look at VMware they jacked up the price 800%+ and a lot of businesses have had no choice but the pay it.

That’s what they’ve been betting on, absolutely. That was my take too, that when the bubble pops and most of them fail then just like in the dotcom boom you’ll still have an Amazon and an eBay who come out on top and become the money making machines.

The problem is that it’s starting to look like the issue is that they can‘t get a functional enough product at a low enough price. When they try to raise prices, usage drops significantly. That’s a killer blow because they have absolutely enormous fixed costs and it makes their burn rate even worse.

They are massively constrained on compute resource for future growth too, and it’s not exactly the work of a minute to throw up a data centre and get it populated.

Investing in the AI boom is starting to stretch some of the biggest funds in the world. Softbank are massively exposed for example. It’s a question of can they sustain the burn rate for long enough and it’s starting to look ugly.

 

[massivewangers]

I understand the investments, as it could be a massive money saver for businesses, and therefore a massive earner for the companies behind it. The impact it could have on the world is on a similar scale to the steam engine, probably more. That said, it could also implode, so it's a big gamble. I suspect @charltjr is right in that there could be some major casualties, but a few will come out of it as absolute behemoths. And of course, just like Amazon, they'll find new ways to f**k people in the name of profit/convenience.

 

[Steve]

It did take them 17 years to be profitable, most of these companies are less than 5 years old.

Musk’s got SpaceX (xAI Grok & starlink) as well as Tesla (optimus) $$$$$$$$$$$

 

[ChrisR]

Spacex also potentially buying Cursor for $60bil….

SpaceX secures option to buy AI startup Cursor for $60bn or partner for $10bn

Cursor is a Silicon Valley startup using AI to automate coding as Elon Musk’s firm seeks foothold in the AI market

www.theguardian.com

 

[charltjr]

Spacex also potentially buying Cursor for $60bil….

SpaceX secures option to buy AI startup Cursor for $60bn or partner for $10bn

Cursor is a Silicon Valley startup using AI to automate coding as Elon Musk’s firm seeks foothold in the AI market

www.theguardian.com

This is where it gets all the murkier for me - Cursor runs on top of Anthropic’s models. Anthropic can’t make money, and Cursor somehow have to add enough value that people won’t stop using it when Anthropic hike their prices.

So many AI startups are using Anthropic or OpenAI models and building their own frameworks around them. That means the whole sector relies on the pricing model of a couple of big players who can’t make any money at the moment.

The business model at the moment seems to be to get AI workflows so embedded into businesses that when the inevitable price rises happen it can’t easily be unpicked. It’ll be interesting to see if the sector can actually pull off the bait and switch before running out of money.

 

[ChrisR]

The business model at the moment seems to be to get AI workflows so embedded into businesses that when the inevitable price rises happen it can’t easily be unpicked.

Which is the way in general now with SaaS/consumption based products, when selling them the aim is to get things as embedded and part of the general business processes as possible so that it makes it hard to just switch that 'thing' off.

 

[R3k1355]

The problem for me is that they have no obvious path to profit. Compute isn’t getting cheaper, their models are getting more complex and consuming more compute, they exist only because huge amounts of cash are constantly pumped into them. Sooner or later, investors need a return.

Investors have already signalled they aren't going to wait.

Meta, Microsoft, Amazon all saw their share price dip on the release of quarterly financial results, with AI spending being the big issue.
$650bn just this year alone, it's simply not sustainable, and the longer it goes on the deeper the hole they need to claw their way back out of.

 

[Matt Cup]

Hucking Fuge Data Centre approved in Utah. It will use more than double the power than the entire state uses 😳

 

[Ray Gin]

I've got 4 agents working for me atm on SuperGrok, it's so strange seeing them sharing ideas with each other and coming to a final verdict etc. Proper weird 😂

 

[Jamess182]

Hucking Fuge Data Centre approved in Utah. It will use more than double the power than the entire state uses 😳

And 90% of that power is going to be used by people generating videos of a horse driving a car. The world is fucked

 

[mace¬]

I've just discovered if you are buying something, ask GPT to give you a discount code and it will find you one if out there. Just paid for the monthly subscription.

 

[Ray Gin]

I wonder when Hollywood will start to put the pressure on for regulations in terms of content generation? Why would Disney spend hundreds of millions making a film when it can just be generated for a few thousand dollars?

 

[Aaron..]

I wonder when Hollywood will start to put the pressure on for regulations in terms of content generation? Why would Disney spend hundreds of millions making a film when it can just be generated for a few thousand dollars?

Wasn't AI use part of the actors strikes a few years back?
Companies wanted them to sign away their voice and image rights so they could AI them in to future films?

 

[R3k1355]

Yea it was, and there is going to be laws brought in to ensure that anything produced with an actors image or voice triggers automatic licencing payments.

but AI companies don't seem to give a s**t and will let their platforms churn this crap out regardless.

 

[R3k1355]

Surely if those are real, and that good. They'll be on the market in the next 12 months?

 

[R3k1355]

 

[Louis]

Mines exactly like genspark

 

[Ray Gin]

I've not seen a pic of mine in ages so can't remember what it looks like.

The genspark one looks like the Subaru logo in reverse

 

[Ray Gin]

@Daz... is a f**king textbook Gemini

 

[Matt Cup]

😂

 

[Archtronics]

Started using Anthropic's Fable 5 model(rumoured to be Mythos) and it’s noticeably better.

Ate through all my credits within an hour tho.

 

[Ray Gin]

Started using Anthropic's Fable 5 model(rumoured to be Mythos) and it’s noticeably better.

Ate through all my credits within an hour tho.

They need to add in some sort of auto mode imo. Changing between agents depending on the complexity of the task, just to save tokens, is really annoying.

I could go up to the highest tier tbf but I don't really need that much.

 

[James]

Had to turn the memory setting off on chatgpt as it was freaking me out and made it seem even more like I was speaking to a real person rather than AI

 

[Touring_Rob]

Jumps to conclusions sometimes too. If you ask it something a bit general it assumes you are still talking about something else from a few days ago.

 

[R3k1355]

Careful with how hard you push that AI on your workforce

 

[Andy.]

Careful with how hard you push that AI on your workforce



View attachment 1791204

SLT - don’t restrict our users. They need to learn and not to worry about costs.
Me - we need guardrails to stop things getting out of control.

$100,000 a short while later….

SLT - We need guardrails to stop things getting out of control.

😆

 

[Donny_Dog]

I wonder how many I.T folk have been asked in the last month alone...
"can you write us an A.I policy..."

 

[dann2707]

Jumps to conclusions sometimes too. If you ask it something a bit general it assumes you are still talking about something else from a few days ago.

Yeah it's started to do my head in with that

 

[R3k1355]

View attachment 1791256
SLT - don’t restrict our users. They need to learn and not to worry about costs.
Me - we need guardrails to stop things getting out of control.

$100,000 a short while later….

SLT - We need guardrails to stop things getting out of control.

😆

Yea I'm sure I read a similar story to that video in the news, companies aggressively pushing AI till they burn through so many tokens they push themselves to the edge of bankruptcy.

We're still in the honeymoon period with AI as well, when prices start ramping up companies will really need to have tight control of token usage or they'll rapidly run up bills they cannot cover.