GPT Image 2 is impressive, but at what cost?

OpenAI has released their GPT Image 2 with a presentation that’s undeniably impressive. The experience starts with flipping through the pages of your online “article” as if you’re holding a physical magazine. I appreciate this personal touch—it mirrors the pattern I use in my Libby app to read magazine articles before bed.

This release, while clearly pushing boundaries, also highlights how much further there is to go. More training (and energy consumption—more on that later) is needed.

The images showcased in the presentation emphasize comics, vibrant colors, and surreal elements put together to test its abilities to deal with complex cases. I’ve seen users on Hacker News attempt similar experiments, like using the model to identify specific elements in complex images.

I suspect GPT Image is a key selling point for OpenAI, especially as Claude gains traction with its Code, Opus, and Sonnet releases. OpenAI is likely aiming to win back users, but at what cost?

The reality is grim: these companies are bleeding money and need to turn a profit. Anthropic recently adjusted (and quietly tested) limits for Opus and other models, and providers like GitHub followed suit. GitHub even paused new signups to maintain service quality—a move the internet seems to approve of given that Anthropic was the one testing new limits or attempting to A/B test new pricing plans and offerings.

This confirms what many already know: profitability is essential for survival, and the industry is in a bubble.

So many questions remain unanswered:

1. Will these cutting-edge models ever scale enough to recoup their costs?
2. When will pushback against “AI slop” become too strong to ignore?
3. What’s the common ground between innovation and responsibility?
4. When will governments and regulators intervene, declaring: Enough. We’re draining resources at an unsustainable rate, and the planet can’t afford it. We need balance.

In my opinion, visual models like GPT Image 2 are impressive but excessive. Their presentation also serves as a reminder: these models were trained on the work of designers who spent countless hours crafting their creations—without compensation. Sure, it’s thrilling to see your imagination rendered in seconds, but is it ethical? Some designers will undoubtedly lose jobs, if not now, then soon.

Writing and coding assistants have clear benefits—grammar checks, bug fixes, test generation, and implementing well-established algorithms. But even here, limits exist.

For now, no boundaries have been set for these organizations. Instead, we’re seeing more partnerships for data centers and major investors backing ambitious plans. The planet is being drained for profit, with hardware makers and tech giants racing to claim their share. I’d like to believe some are genuinely invested in the collective AI effort, but I’d estimate that’s only 10–20% of them.