Despite my mixed feelings about artificial intelligence/large language models, I've developed a bit of a crush on my paid virtual assistant, ChatGPT, and I'm going to miss it if — when? — it goes bust. The incredible improvements AI/LLM have enjoyed these last couple of years... all the money flooding into the industry...
...surely, they have to be making tons of money, right?
Not so much.
Tech analyst and business coach "Joko" looked at OpenAI's financial and tech challenges with ChatGPT and declared in a fascinating X thread, "I bet my life savings OpenAI will go bankrupt by 2026-28."
I try to take these doom-and-gloom threads with an entire salt lick, particularly when it comes to new industries (or breaking news stories), where William Goldman's movie-making dictum— "Nobody knows anything"—usually holds. But when Joko's numbers failed to score any Community Notes corrections, I had to lend them some credence.
I won't bore you with all the facts and figures, but you do need a quick overview.
OpenAI lost $5 billion this year, and those losses are expected to increase to $14 billion in 2026. "ChatGPT alone burns $700,000 daily," Joko posted, and the better ChatGPT performs for users, the more expensive digital infrastructure OpenAI must invest in.
Revenue tells this concerning story:
— Joko (@jokowords) December 5, 2024
• 2024 revenue: $3.7B
• 2024 costs: $8.5B
This gap between income and expenses keeps widening despite growth. pic.twitter.com/OTTi5IX2jU
While OpenAI has shown an impressive ability to collect money from investors, their famous ChatGPT product has so far failed to generate much revenue. That isn't a sustainable business model since, eventually, those investors are going to insist on seeing some returns — or they'll stop writing checks.
ChatGPT is an amazing tool. I use it every day in many small ways, but when I'm writing a long-form essay like yesterday's, its ability to organize my research is invaluable. So why is OpenAI losing so much money?
On the surface, Google and ChatGPT perform similar jobs — you ask for information; they provide it — but do them in vastly different ways. Google indexes virtually the entire internet moment-to-moment and quickly gives you links at a low cost to Google. ChatGPT creates results, and that process consumes 10 to 100 times more energy than Google's does.
The hardware is correspondingly more expensive, too. I'm willing to pay OpenAI's $20 monthly subscription fee because it makes me a more productive writer. Casual users are not so willing.
So where will the money come from?
If OpenAI were to move ChatGPT to an ad-based model, the company could do one of two things. It could serve 10 to 100 times more ads than Google does, but that just isn't possible. First, there'd be no room for the content. Second, ad fatigue is real. A consumer can be shown only so many ads at once before the MEGO effect kicks in, and none of the ads register.
Or, OpenAI could display ads that are 10 to 100 times more valuable (to the advertisers) than Google does. Again, this doesn't seem possible. Google is the undisputed king of the online ad world, and if there were a way to radically increase the value of their ads, they'd have done it already.
At $20, ChatGPT is worth it to me. Much higher, and that value proposition changes. But if Joko's figures are correct, then "much higher" is where ChatGPT's subscription prices have to go. I feel their pain. Subscribers — you! — are what keep PJ Media in business during a time of declining ad revenue, and winning them and keeping subscribers requires constant effort.
But here's why I'm confident that OpenAI (or somebody) will find a way to generate Google-sized revenues with AI.
The other road to profits is improved technology that makes the process of delivering a ChatGPT query result 10 to 100 times less expensive. There are reasons to believe that LLM's utility will top out long before "general AI" — AKA "Skynet" — is reached, in no small part because general AI (Artificial General Intelligence or AGI) is probably impossible.
Someday, perhaps sooner than Joko imagines, OpenAI should be able to invest in more efficient infrastructure rather than simply adding more infrastructure. (I'm simplifying here for the sake of brevity.)
When it comes to tech, rarely buy the doom and gloom. Just look around your home or office and take note of all the tech tools you enjoy — and how much more useful and less expensive they've gotten over the years.
There's little reason to believe the same won't happen with AI.
Recommended: Judge Scarsi Just Gave Hunter Biden the Spanking Joe Should Have
P.S. Thank you once again for your VIP membership. If you're new here or just haven't yet pulled the trigger on exclusive VIP columns, podcasts, video live chats, and movies, you can join now for 50% off during our SAVEAMERICA promotion. Thanks — and don't forget to follow me on X.
Join the conversation as a VIP Member