ChatGPT’s 'compute-intensive' next act better be dazzling, because the competition is outshining its recent efforts

Sam Altman says something big is coming to ChatGPT. Something “compute-intensive.” Something powerful enough to justify a higher price tag, or at least a ChatGPT Pro subscription. But it will have to really shine, as Google and other OpenAI rivals are already lighting up the sky.

Altman's X post managed to sound both ambitious and apologetic. He promised that features are rolling out soon, and that they’ll push the limits of current AI. He then admitted that they may also push your wallet's limits. So some tools will be available only to Pro subscribers, and others might demand an even steeper fee. Because, as he put it, these upcoming experiments are going to demand a lot of computing power.

Altman wasn’t vague about the costs. The company wants to explore what AI can do when you crank the dials to eleven, and this tiered pricing is the way to make it happen.

As someone who has experimented with ChatGPT for years now, I would expect the news to match Altman's hype; but that doesn't mean that what he thinks is exciting will actually come off that way.

This wouldn't be such a problem for OpenAI were it not for the fact that the competition has been putting on a real spectacle lately. OpenAI’s latest offerings, for all their utility, are starting to look like they’ve lost the showmanship.

Google's Nano Banana, silly name aside, has proved to be an impressive, absurdly fast contender among AI image generators, able to make 3D figurines of people realistic enough to look like they escaped from a collectibles shelf.

Meanwhile, Veo 3 from Google has kicked AI video generation for the masses into high gear. Though the clips are short, the cinematic lighting, consistent characters and believable motion are all stand-out features. The same goes for Meta’s recent open-source models, and those coming from smaller players are arriving with stunning demos every other week.

ChatGPT's native AI image creation is still good, and Sora's videos can look quite amazing with the right prompt, but most of what people think of when they think of ChatGPT is utility around text; and helpful isn’t the same as thrilling. And while GPT-4o’s voice upgrades make the AI assistant even more useful, they're not eye-catching like some of the other multimodal features popping up from OpenAI's rivals.

AI power play

Which brings us back to the “compute-intensive” mystery box. OpenAI is clearly investing in something ambitious. It wants to see what happens when it throws massive computing power at new ideas. They want to figure out how far the model can go when you don’t limit it by budget constraints. They’re pushing the envelope, but asking paying users to help fund the postage.

Fine. But then show us something we haven’t seen before. If you’re going to call something compute-intensive, I want to feel the power being deployed. I'm picturing something like uploading a video of someone assembling IKEA furniture, and a photo of my room and of me, and seeing a new video showing me assembling the piece and putting it in different spots in my room with real-time annotations of what to do and the pitfalls to avoid.

I'm not saying I would pay a lot for that, but it's the kind of "wow" moment that would at least tempt me to add another monthly bill to the pile. On the other hand, part of ChatGPT's appeal is how available it is for the average person; but that accessibility shouldn’t come at the expense of innovation.

The company may be building the most robust, aligned, reliable AI model behind the scenes, but from the outside it sometimes feels like we’re just waiting for the next big thing, without any clear sense of what it will be. But still, if you’re going to break the bank – or ask you users to – you're going to have to show us something worth the outlay.

Maybe ChatGPT is about to stun us all with some gleaming, blazing, multi-sensory upgrade that reminds us why we fell in love with this tech in the first place. Because OpenAI can’t keep assuming it will have the lead among AI assistants forever. If the next act is going to cost more, it had better feel like something that’s worth every cent.

You might also like

• OpenAI's high-minded approach to AI-human relationships ignores reality
• OpenAI has a new scale for measuring how smart their AI models are becoming – which is not as comforting as it should be
• GPT‑5 Pro is brilliant, but it’s still nowhere near real AGI, says one of the professors who coined the term