Comment OpenAI CEO Sam Altman complained on the platform that was formerly known as Twitter Saturday, that the service, even at $200 per month for ChatGPT Pro struggles to make a profit.
“Insane thing: We are currently losing money on OpenAI Pro subscriptions!” He wrote in a blog post. The problem? According to @Sama “people use it much more than we expected.”
Altman’s admissions came about a month after OpenAI announced plans to charge customers $200 per month — ten-times the cost of Plus subscriptions — for its Pro tier of ChatGPT. Users get unlimited access to the advanced voice, o1 model for the extra money. “which uses more compute for the best answers to the hardest questions.”
Altman was taken by surprise when the new pricey tier became popular. Altman wrote in a separate blog post that he had “personally chose the price and thought we would make some money.”
We think the real problem is:
Altman did not elaborate on the issue, but it could be related to how OpenAI’s O1 models work.
With LLM inferencing, a greater number of simultaneous users can be advantageous for maximising utilization – if you own or rent hardware instead of paying by token. o1 generates responses using a chain-of-thought (CoT).
CoT breaks down problems into the constituent steps needed to solve them. Chain of thought models often require more tokens to encode words, punctuation and other jargon than a traditional model such as GPT-4.
We understand that this process is obscured by OpenAI’s O1 models, which pauses while it is “thinking.” At the end, the user only receives the final answer.
The more tokens, the longer the generation time, which means that GPUs are occupied for a longer period of time, leading to higher operating costs.
It remains to be seen what this means for ChatGPT and whether OpenAI will increase subscription fees or decrease them to boost revenues or amortize costs over a larger number of casual users. We wouldn’t surprise if OpenAI decided to remove “unlimited” the access to expensive models.
We’ve contacted OpenAI to get their comment. We’ll let you all know what we find out. OpenAI’s for profit path is blocked
OpenAI has been struggling to generate a profit with ChatGPT Pro just weeks after revealing its latest step on its long-term transformation from non-profit to for profit.
OpenAI’s latest corporate restructuring will see its for-profit arm transition into a public benefit company (PBC), allowing it to make larger investments. According to the plan, its PBC arm will have the final say on operations and business while the non-profit will maintain “significant interest in the existing for-profit,” for things like hiring, charitable initiatives, and so on.
It is a departure for OpenAI, which had a non-profit overseeing a for-profit.
Altman and his team are still figuring out how to run an effective business. The firm has raised $6.6 billion of new funding since early October.
- Microsoft’s AI PC plans are spared by the jury, but Arm is considering a retrial.
The hype man will continue to hype
As the New Year approaches, it appears Altman still has a flair for the dramatic. He shared bold predictions about where AI will go next. Altman, in a long blog post, said that he is now confident that the company knows how to “build AGI [artificial general intelligence] as we have traditionally understood it,” by 2025, the first AI agents will be “join the workforce.”
As if AGI weren’t sufficient, Altman claims OpenAI has begun to focus on “superintelligence,” that, if genAI hype is to believed, will bring “massively accelerate scientific discovery and innovation well beyond what we are capable of doing on our own, and in turn massively increase abundance and prosperity.”
Success for whom, of course, is still an open question. (r)
