With OpenAI being at the center of the AI hype, I would’ve thought they’d be raking in the dough instead of losing $5 billion. So it’s really just Nvidia making money on this bullshit, huh? It’ll hurt when the hype dies down and Nvidia drops from the second top spot on the S&P 500. We’re all going to feel that one.
You know, to make money in a gold rush, don’t dig, sell shovels.
And Nvidia has really fancy shovels.
What is AI gold then?
Autonomous vehicles, robotics, LLMs
nah its whitelabelling AI credits. You build an API connector, charge 3x for a credit and sell it to a business.
Obsolescence of human workers/employees.
Eh all these companies operate as loss leaders until they capitalise the market.
- CNBC has confirmed that OpenAI expects about $5 billion in losses on $3.7 billion in revenue this year — figures first reported by The New York Times.
- Revenue is expected to jump to $11.6 billion next year, a source with knowledge of the matter confirmed.
So yeah some small loses here and there to make back far more in the future.
Which is fine in theory, but “expected” based on what?
They haven’t demonstrated any ability to meaningfully improve their models (“meaningfully” meaning "sufficient to actually address the very serious concerns about their practical usability), they haven’t shown any ability to meaningfully capture enterprise sales for their API, and their conversion rate on free users to paid users is abysmal. Their only stated plan to increase revenues is doubling their prices, which given their already terrible user retention doesn’t actually seem like a reliable way to bring revenue up. Jacking up prices only works when your users find you indespensible, and everything OpenAI offers can be found elsewhere for less.
The assumption is that they’ll develop some kind of moat, but there are plenty of other AI models on offer or in development. It would also be useless capturing a market when the companies that would be their customers realize they’re not making money on the AI themselves.
You have to get people hooked on your product, though.
If they and every other AI company just evaporated no one would really be bothered.
You can’t capitalize a market that doesn’t really exist.
This is exactly the problem. There are plenty of people who will crawl out of the woodwork to tell you how they’ve found a way to make AI “useful”, but very, very few could put their hand on their heart and say that it was “essential” to their workflow or their own happiness and wellbeing in any meaningful way.
I think this is just OpenAI marketing.
“Insane thing: We are currently losing money on OpenAI Pro subscriptions!” he wrote in a post.
The problem? Well according to @Sama, “people use it much more than we expected.”
Oh no, ChatGPT is too useful to customers! Altman isn’t going to be telling any real problems that OpenAI has to the whole world over Twitter.
You’re right, that’s definitely what Sam is trying to do here. Unfortunately for him, he’s still an idiot, and he’s inadvertently telling on himself here by openly confirming what’s been well understood for a while; ChatGPT simply is not profitable to run because the models are so stupidly inefficient. That’s a real problem, and one that they’ve shown no meaningful plan for solving.
Fuckin LLM bubble needs to burst already. I want some crazy compute cards to play with.
Also, who knew the only people who would pay 2400/yr for access would bebe the ones who plan to make hundreds of queries per day. What do like, people try to think about value for money before they buy stuff?! What are you all… like not filthy rich?! Ew.
Strange coincidence right as they want to convert to a for-profit company structure. “Bro we are not even making profits, nothing to see here bro”
That’s because the whole thing is stupid. Is made by stupids, marketed to stupids, paid by stupids, and for the most part used by stupids. Because they’re stupid.
There’s a pattern in there if you look closely.
I feel like you’re putting it down, but for the life of me, I am not picking it up. Can you please try explaining again, but slower? Use simple words. Like I’m stupid or something.
I think he’s trying to say it’s marketing’s fault.
That’s stupid
The amount of money stupid gives to stupid, though. Makes my stomach churn. So much for so little.
There is value in current ai but everyone who’s in charge thinks it’s economic value.
If AI cost peanuts to run, this would be a very reasonable point. But it doesn’t. It’s staggeringly expensive to operate something like ChatGPT.
So any use of genAI has to consider the question “Do the benefits provided actually justify the cost?”
Obviously, in a capitalist society this turns into “How can we monetize this?”, but even in a fully socialist society it would still be necessary to ask if this technology is actually providing sufficient societal benefit to actually justify the material resource cost of running it.
Or you’re stupid because you can’t use LLMs effectively, don’t understand their value, and now you’re angry because of that.
What is the use case for a $200 a month AI subscription? It’s a lot of money to spend on a novelty, clearly people are finding it useful.
That’s nothing to a business.
What business though?
I run tech for a midsize business, and consult for several small businesses. Aside from one 4-person company, all of the businesses I oversee found it less expensive to host their own LLM in Azure than to pay for OpenAI’s subscriptions. I’m talking 10% of the cost of subscriptions for the same functionality.
The midsize business in particular has only seen measurable benefit from more specialized/global applications of “AI” tools, such as integrating machine learning into data analytics. There are a ton of people who use the LLM chat, but I think the mishaps caused by the LLM may have undone any efficiency gains. Either way, I’m sure glad they’re not paying hundreds of thousands of dollars annually for it.
Lol, well obviously it would be cheaper to host an LLM that is smaller. Imagine the cost of hosting o1
Fair. We’re using 4o and o1-mini right now, because access to the full o1 is waitlisted in Azure. However based on some brief review of their pricing for o1, I’d say we’re still going to save a metric fuckton of money compared to per-user subscriptions.
Businesses that think shoving AI tools down everyone’s throat will create superior productivity so they can lay off workers
Any of the businesses that have hopped on the AI train. $200/month is basically the price of a single Indian call center employee. A company can pay for the AI subscription and fire 90% of the call center, using humans only for escalation.
Lazy idiots
Bruh. Meanwhile I’m still with my free and libre Mistral-7B I refined using my own WhatsApp messages and I almost never use it…
You don’t use it because the AI already took over your life and murdered your biological counterpart
Well she ain’t too smart then cuz now she has to pay rent so I’d say we’re even
