- OpenTools' Newsletter
- Posts
- 💸 ChatGPT Breaks Ad Promise
💸 ChatGPT Breaks Ad Promise
Plus Wikipedia finally charges AI companies and Israel launches supercomputer
Reading time: 5 minutes
🗞️In this edition
OpenAI Breaks Promise, Adds Ads to ChatGPT
Wikipedia Finally Charges AI Companies After Years of Free Access
Israel Launches AI Supercomputer for Startups and Researchers
In other AI news –
DeepMind CEO: China AI Only Months Behind US
Microsoft Predicts Personal AI Companion for Everyone in 5 Years
Tesla AI Leader Says Engineers Still Need Coding Skills
4 must-try AI tools
Sometimes the most telling moments are when someone does exactly what they said they wouldn't.
One company is adding the thing its CEO called "uniquely unsettling" just two years ago. Another is finally charging for access after 25 years of being free. And a third launched infrastructure smaller than competitors but critical for survival.
Each decision reveals something about desperation masked as strategy. About running out of options while claiming evolution. About promises that expire when the math stops working.
The gap between what leaders say and what they do keeps widening. The real story is always in that gap.
This edition looks at the moves companies make when they're out of better choices.
What's happening:
OpenAI announced Friday it's bringing ads to ChatGPT. Free users and $8 ChatGPT Go subscribers will see sponsored products at the bottom of answers starting soon.
The ads show up when relevant to your conversation. They're labeled "sponsored" and OpenAI says they won't influence what ChatGPT tells you.
Plus, Pro, Business, and Enterprise subscriptions stay ad-free. Only free and Go tier users in the US will see ads during testing.
OpenAI says it won't sell your conversations to advertisers. Ads won't show for health, mental health, or politics topics. No ads for users under 18.
Sam Altman previously said he "hates" ads and called the idea "uniquely unsettling." But with $1.4 trillion in infrastructure commitments, OpenAI needs more money.
Why this is important:
This shows subscriptions alone won't fund OpenAI. Only 5 percent of 800 million weekly users pay. That math doesn't work.
The timing matters. OpenAI is preparing for an IPO and needs to prove it can monetize free users. Ads are the path Google and Meta already proved.
If OpenAI makes ads work in conversations, Google will do the same with Gemini. The entire AI industry will follow.
ChatGPT ads sit at the bottom of answers. Users just got helpful information. They're less skeptical. That makes them more likely to click than search ads.
Comments from the editor:
Altman said he'd only add ads as a "last resort." This is him admitting they're out of options. Subscriptions aren't scaling fast enough.
The "uniquely unsettling" quote is now awkward. Altman called AI plus ads disturbing in 2024. Now he's launching it in 2026.
This makes ChatGPT more like Google Search and less like a neutral tool. Once ads fund the business, advertisers gain influence even if OpenAI denies it.
For users, this is the tradeoff. Free access continues, but you're now the product. Your conversations fuel ad targeting even if the data isn't "sold."
What's happening:
Israel launched its national AI supercomputer this week, giving startups and researchers access to computing power equivalent to 1,000 Nvidia B200 accelerators. The resources will be available at below-market rates.
Seventy percent of the capacity goes to AI startups working on model training. Thirty percent goes to academic researchers conducting fundamental studies.
The Innovation Authority is managing the allocation through a high-availability supercomputer platform. Dror Bin, the Authority's CEO, said companies and researchers now have direct access to advanced computing resources through discounted accelerators.
The launch comes as Israel faces limited domestic access to high-performance AI infrastructure. Most Israeli companies currently rely on foreign cloud providers for large-scale training.
Why this is important:
This shows Israel betting on AI infrastructure as a competitive advantage. Without domestic computing resources, startups can't experiment or train models affordably.
The 70-30 split favors commercial applications over pure research. That's deliberate. Israel wants startups building products, not just publishing papers.
For companies, below-market pricing changes the economics. Training large models costs millions on AWS or Google Cloud. Subsidized access makes that feasible for smaller teams.
The 1,000 B200 equivalent is modest compared to global competitors. The US approved 500,000 Nvidia processors for the UAE alone. But Israel is starting somewhere rather than waiting.
The real test is whether allocation expands with demand. If startups actually use this capacity, the government will need to fund more processors. If not, the infrastructure sits idle.
DeepMind CEO: China AI Only Months Behind US – Google's AI chief says Chinese models are closer than anyone thought, but there's one thing they still can't do that matters most.
Microsoft Predicts Personal AI Companion for Everyone in 5 Years – Within five years, you'll have an AI that sees what you see, hears what you hear, and knows you better than your closest friends do.
Tesla AI Leader Says Engineers Still Need Coding Skills – Former Tesla AI director pushes back on Nvidia CEO's claim that coding is dead, saying AI tools aren't reliable enough yet.
Rizemail - An AI-powered email summarization tool that helps users get to the core of their unread newsletters and long email threads
Kastro Chat - An AI-powered chatbot platform that allows businesses to create their own chatbots without any coding knowledge
Verbalate - A video translation and lip sync software designed to help businesses reach a global audience
Taranify - A platform that uses AI technology to provide mood-based recommendations for music, Netflix shows, and books
The pivots are getting harder to spin.
From "never ads" to launching ads. From free forever to paid access. From waiting for scale to starting small and hoping.
These aren't confident moves. They're necessary ones. The kind you make when subscription revenue isn't enough, traffic is dropping, or you need infrastructure now instead of later.
What matters is whether users notice the gap between what was promised and what's being delivered. Or whether the explanations work well enough to keep them around.
If you have any feedback for us, please reply and let us know how we did. We're always looking to improve.
How did we like this version? |
Interested in featuring your services with us? Email us at [email protected] |