At OpenAI’s annual developer day on Tuesday, the new tool called dots drew most of the attention, but recent changes to subscription pricing may reveal more about the company and the AI sector. The $200-per-month plan now includes half as many tokens as before, and OpenAI introduced a new $500-per-month tier. Those moves highlight a growing compute bottleneck and the high costs of running ambitious, always-on AI services.
What changed in pricing
OpenAI reduced the token allotment for its $200 monthly plan and added a conspicuously expensive $500-per-month option. In practice, that means users receive fewer tokens for the same $200 price while an upper-tier, much more costly alternative is available. The shift suggests the company is trying to manage limited compute and token resources while monetizing heavy usage.
Compute limits and the industry context
The price adjustments are another sign that compute capacity is a serious constraint across the AI industry. Running always-on, agentic assistants is expensive; OpenAI and Meta — which recently launched an agent called Muse — will need substantial resources to offer those services continuously. It remains unclear whether such agents will perform reliably and usefully at scale, because enabling them at top capability requires many tokens and high compute, which is costly.
Early user experiences and limitations
Although dots received the spotlight, reports indicate that Meta’s Muse users encounter paywalls and CAPTCHA-like hurdles they must work around. By contrast, the author notes that Codex is operating smoothly and providing useful functionality, and it’s uncertain whether dots will replace some projects already built on other tools. New AI features often impress, yet still fall short of full readiness.
A concrete example of high costs
OpenAI has also spent large sums on ambitious technical problems: the company reportedly spent roughly $10 million over about a weekend to solve part of the Navier–Stokes equations. That example underscores how expensive certain research and compute-intensive tasks can be under current token-pricing and infrastructure conditions.
Fundraising and IPO status
Bloomberg reported that OpenAI is seeking $30 billion in a new funding round at a $1.4 trillion valuation, while delaying its initial public offering (IPO).
Why this matters
The changes in pricing and the examples of high compute costs indicate that scaling powerful, always-available AI assistants is not only a technical challenge but also a financial one. Companies must balance ambitious product plans with the reality of limited compute resources and the need for sustainable funding and pricing models.
Conclusion
The demos at developer day — especially dots — were notable, but the pricing revisions and the size of the bills for compute-heavy work reveal fundamental constraints that will shape how and when these AI capabilities become broadly reliable and affordable.
