Editing Openai/69557275-0da8-8004-a03a-8eb02e10a61f (section)

===== III. Compute Time as Class Marker =====

There will be free AI.

There will be “normal-person AI.”

And then there will be compute aristocracy.

Right now you can already see the ladder forming:
* $20/month gets you Plus-tier access. OpenAI<ref>{{cite web|title=OpenAI|url=https://openai.com/index/chatgpt-plus/|publisher=openai.com|access-date=2026-01-02}}</ref>
* $200/month gets you Pro-tier access (explicitly positioned as “more compute”). OpenAI<ref>{{cite web|title=OpenAI|url=https://openai.com/index/introducing-chatgpt-pro/|publisher=openai.com|access-date=2026-01-02}}</ref>

And the easiest way to feel compute is video.

Example: Runway’s Standard plan is $12/month and is roughly ~62 seconds of Gen-3 Alpha video per month; Pro is $28/month and is ~225 seconds. Runway<ref>{{cite web|title=Runway|url=https://runwayml.com/pricing|publisher=runwayml.com|access-date=2026-01-02}}</ref>
Luma’s Dream Machine shows the same reality: credit systems that make “ten seconds” a real budget decision. Luma AI<ref>{{cite web|title=Luma AI|url=https://lumalabs.ai/learning-hub/dream-machine-support-pricing-information|publisher=Luma AI|access-date=2026-01-02}}</ref>

In other words: today, “make me a movie” is still expensive enough that people do math.

But the direction is obvious: as prices fall and tiers expand, compute time becomes the new leisure.

Not “I watched a movie last night.”
“I generated one that was for me.”

And once rich people can rent always-on frontier compute the way companies do—e.g., an H100-class GPU at $3.79 per GPU-hour—a single always-on GPU is roughly $33k/year in raw rental. Lambda<ref>{{cite web|title=Lambda|url=https://lambda.ai/pricing|publisher=lambda.ai|access-date=2026-01-02}}</ref>

That’s how you get a world where “What plan are you on?” becomes a class tell.