I hold costs constant at $8B and get x = 4.4. $8B is probably a slight overestimate of current costs, I just took the losses from the article and discounted the last year's revenue to $3B. Users use inference which costs money so, in reality, costs will scale up with revenue, which is why I note this is a false assumption. But I also don't know how much of that went into training and whether they'll keep training at the current rate, so I can't get to a better guess.
If OpenAI starts making a lot of money on each subscription -- implied by your assumption that revenues will 4.4x while expenses stay constant -- the competition will aggressively undercut OpenAI in price. Everybody wants to take market share away from OpenAI, and that means OpenAI has to subsidize their users or sell at break even to prevent that from happening.
Furthermore, training also gets exponentially more expensive as models keep growing and this R&D is not optional. It's absolutely necessary to keep current OpenAI subscribers happy.
OpenAI will lose money, and lots of it, for years to come. They have no clear path to profitability. The money they just raised will last maybe 18 months, and then what? Are they going to raise another 20bn at a 500bn valuation in 2026? Is their strategy AGI or bust?
300m $/mo. * 12 mo. * x - costs = 7.85b
or
$3.6b * x = 7.85b + costs
I hold costs constant at $8B and get x = 4.4. $8B is probably a slight overestimate of current costs, I just took the losses from the article and discounted the last year's revenue to $3B. Users use inference which costs money so, in reality, costs will scale up with revenue, which is why I note this is a false assumption. But I also don't know how much of that went into training and whether they'll keep training at the current rate, so I can't get to a better guess.