Analysis: assuming API pricing, the $200/month Claude Max and ChatGPT Pro plans offer up to ~$8,000/month and ~$14,000/month worth of tokens, respectively
Recently, we purchased one of each Anthropic/OpenAI subscription plan and randomly ran long horizon coding tasks until we exhausted the weekly limit. It's widely believed that a $200/month plan maxes out at ~$2000/month worth of tokens (assuming API pricing). However, we found [image]
@semianalysis_
Context & Ripple Effects
OpenAI’s consumer AI subscription ladder has expanded from the original $20/month Plus plan to $100 and $200 Pro tiers with materially higher Codex usage limits. The related coverage also shows OpenAI pursuing corporate adoption alongside these individual plans.
This analysis puts a concrete API-priced ceiling on what the highest-end plans can deliver under sustained coding use. It matters because the plans’ stated monthly fees are far below the measured value of the tokens users may be able to consume, though that comparison is not the same as OpenAI’s underlying cost.
First-order effects
For heavy coding users who can reach the relevant limits, ChatGPT Pro and Claude Max can provide substantially more token access than buying equivalent API usage at listed prices, according to the analysis.
OpenAI and Anthropic are effectively extending high-value usage through fixed-price subscriptions, making the limits and eligibility rules—not the headline monthly fee—the immediate constraint for power users.
Second-order effects
API-based developers and subscription users face sharper product segmentation: users with irregular or very high interactive workloads have more incentive to use premium chat plans, while programmatic workloads still require API access.
The finding raises pressure on both providers to tune rate limits, coding-agent access, and plan differentiation so that premium subscriptions remain attractive without becoming an uncontrolled substitute for metered API consumption.
Third-order effects
If high token-to-fee ratios persist, frontier-model vendors may increasingly use subscriptions to lock in individual power users and monetize workflow integration, while reserving more predictable, usage-based economics for APIs and enterprise customers.
The durable competitive issue becomes how providers package scarce inference capacity across consumer, developer, and corporate channels; the reported API-price equivalent alone does not establish whether the plans are profitable.
The trend: Premium AI subscriptions are evolving from simple feature bundles into capacity-allocation products that price sustained access to advanced models and coding tools far below nominal API-equivalent usage.
Subscription plans are massively subsidized. And by massively, I mean absurdly: Claude Max 20x: $200/month, with usage reportedly worth around $8,000 ChatGPT Pro 20x: $200/month, with usage reportedly worth around $14,000 [image]
@scaling01 nah people have done this analysis in the past . i don't think public discussion of it matters much to the labs, the main thing is that they retain freedom to adjust limits as they desire plus that average usage in fact remains low (see all the non-interactive claude c…
IMHO, nothing about OpenAI is sustainable. They are surviving on hype. Even with a big subsidy they are still losing market share, and they have completely lost whatever technical advantage they once had. They are spending too much money on public opinion (lobbyists,
been on both anthropic and openai's subscriptions for a while and this aligns very well with my real experience - you get a lot more value from openai's plans right now and this analysis hasn't even taken into account that gpt 5.5 get the same thing done with much less tokens
SemiAnalysis quantified token-maxxing. Claude Max: $200/mo gets you $8,000 in tokens. ChatGPT Pro: $200/mo gets you $14,000 in tokens. Doubt that will last much longer. [image]
Hot take: They're not subsidized their margins are insane. They are just absolutely raping api customers. Anyone who has used deepseek or hosted anything and done the math on hardware/power costs knows this
The margin on a subscription plan is a function of the average utilization. If we assume both companies have 75% API gross margins, this results in the following subscription margins. (3/4) [image]
API is overpriced vs subscriptions by 40-70x Do you know what this means? They have DeepSeek-level compute efficiency if not better. Which makes perfect sense (tbh still too good for DS), given frontier budgets for overtraining, autoresearch, larger and older staff etc.
Still laughing at this. Even if you believe that OpenAI/Anthropic have 75% margins on tokens, all it takes for the margins to go negative on any of their subscriptions is for a user to use 25% of their rate limit. In any other context we'd be calling these people losers!
Read these two facts together. Because there's something important to learn from it. Anthropic just told investors it's on track for its first profitable quarter, with revenue more than doubling to ~$10.9B. OpenAI is projected to burn well into the double-digit billions this [ima…
obviously, maximally utilized subscriptions are negative. but I think the negative margin isn't that dramatic anymore (because it's too easy to max out with agents now). I'd say current subscriptions must be profitable up to ≈35% utilization. https://x.com/...
@kimmonismus You could say that, but at same time API prices are likely significantly inflated and many users of sub plans probably not using it all too. In the end the math is working out for these companies. And doubt it will change much in future, at least regarding models of …
If this price war escalates, this probably will be the reality check AI needs. The spending needs a ROI & fast. Expect a blowback on chip stocks. OpenAI Considers Drastic Price Cuts, Anticipating War for Users With Anthropic https://www.wsj.com/...
This is in line with my understanding. Subsidization is a lot more intense than most people think. AI is extremely expensive. Will be a big shocker when Anthropic and OpenAI stop subsidizing. One engineer can easily spend $100k of tokens/year
Obviously this is way worse than API overall. However, explicitly nerfing subscriptions leads to huge public backlash, and the rapidly falling cost of intelligence means you'll be able to profitably serve Opus 4.8 level models for $20/month in the near future. We therefore think
Both OpenAI and Anthropic are considering price cuts. Neither of these companies but ESPECIALLY OpenAI can under any circumstances afford this pay cut. This is fundamentally insane. A sign of desperation that will erode revenue while costs stay linear. https://www.wsj.com/... [im…