Has OpenAI reframed the anchor price for AI?
Steven Forth is CEO of Ibbaka. Connect on LinkedIn
OpenAI is testing a new framework for pricing AI agents. One that resets the anchor prices for AI agents from tens or hundreds of dollars per month to thousands or tens of thousands of dollars per month. This is a big deal for business and agentic AI.
What is happening here? Will it work? What does this mean for other types of agents?
In the week of March 3, I noticed almost 30 mentions of OpenAI’s test of new pricing for its agents on my LinkedIn feed. It started for me with this article in The Information.
OpenAI Plots Charging $20,000 a Month For PhD-Level Agents which was published on March 5, 2025 (the article is behind a paywall, I’m afraid).
The key quote is
“OpenAI executives have told some investors it planned to sell low-end agents at a cost of $2,000 per month to “high-income knowledge workers”; mid-tier agents for software development costing possibly $10,000 a month; and high-end agents, acting as PhD-level research agents, which could cost $20,000 per month, according to a person who’s spoken with executives.
In the long run, OpenAI expects 20% to 25% of the company’s revenue to come from agent products, the person said.”
OpenAI framed pricing for co-pilots back in 2023
This and similar reports provoked a lot of comment. After all, OpenAI effectively set the pricing anchor for AI chatbots and search. They used a Van Westendorp study to set the price for a ChatGPT subscription at $20 and then priced the Pro Version at $200.
Initially there was a lot of negative comment, with some people saying this was way too expensive (comparing it to free, advertising funded, search) and others, many in the pricing community, saying that OpenAI had massively underpriced these AI services and that it would be difficult to recover.
Back in July 2023, Ibbaka conducted its own Van Westendorp research that largely confirmed OpenAI’s initial pricing decision. See Pricing AI assistants for productivity suites: Survey Results. This is an example of what a Van Westendorp survey looks like.
A Van Westendorp survey is a pricing research method used to determine the optimal price range for a product. It involves asking respondents four key questions:
At what price would you consider the product to be so expensive that you would not consider buying it?
At what price would you consider the product to be getting expensive, but you would still consider buying it?
At what price would you consider the product to be a good value for the money?
At what price would you consider the product to be so cheap that you would question its quality?
ChatGPT anchored pricing for at $20 to $200 per month for co-pilots, depending on functionality. Most other companies have followed along. And then DeepSeek challenged the status quo with a much lower price for a reasonably good model back in January 2025.
Foundation model companies needed to reframe the conversation on pricing
This is a real problem for AI companies as they move to reasoning or research models. These models are many orders of magnitude better than conventional models for some jobs. They also consume a lot of tokens (reasoning models don’t just have input and output tokens; they generate their own tokens as they churn through the chain of thought, OpenAI calls these reasoning tokens).
One of the most salient comments was by Egor Chirkunov in a LinkedIn post 𝗢𝗽𝗲𝗻𝗔𝗜'𝘀 $𝟮𝟬𝗞 𝗔𝗴𝗲𝗻𝘁𝘀: 𝗧𝗵𝗲 𝗠𝗮𝗿𝗸𝗲𝘁 𝗠𝗮𝗻𝗶𝗽𝘂𝗹𝗮𝘁𝗶𝗼𝗻 𝗡𝗼 𝗢𝗻𝗲'𝘀 𝗧𝗮𝗹𝗸𝗶𝗻𝗴 𝗔𝗯𝗼𝘂𝘁.
“The genius? By anchoring at $20K for Ph.D.-level work, the $10K developer agent and $2K knowledge worker tiers suddenly feel like bargains. Yesterday, walking into a corporation pitching a $5K/month AI service would get you laughed out of the room. Today? That's a “reasonable mid-tier” option.
OpenAI doesn't need anyone to actually buy these agents. The damage is already done. They've normalized the concept of AI-as-employee in corporate budgeting. Every competitor can now comfortably 5x their pricing while still appearing reasonable by comparison.”
A quick poll on “How much would you pay per month for a 'Ph.D.-level AI agent?”
On March 5 and 6, I put out a poll asking “How much would you pay per month for a 'Ph.D.' level AI agent?” N = 147 across my personal feed, the Professional Pricing Society, the Software as a Service - SaaS group, the Design Thinking Group and the Artificial Intelligence Exchange.
The results are much as one might expect. Regraphing this one sees a typical decline in volume with price.
Note that this is an informal poll, and a snapshot in time, done without any framing on the value of the agent. In the real world, one should only ask about price in the context of value. There are a lot of different types of Ph.D.’s doing a lot of different work so OpenAI’s positioning of the twenty thousand dollar a month agent is too vague to really be meaningful.
Before anyone looks at the above graph and concludes that the right price is $1,000 per month (already 5X lift over the $200 per month Pro subscription take a look at the below graph. The revenue optimizing price would be $5,000 per month with $10,000 and $20,000 far outperforming the $1,000 price point. This is for revenue optimization, pricing for profit optimization would be different again.
Of course if one was really pricing these offers (which don’t really exist yet, though they are coming in one form or another) one would do a lot more research.
Build and validate a value model
Map value drivers to functions and capabilities
Segment the market and design offers for each segment
Optimize pricing to align with goals (market growth, revenue growth, profit growth)
OpenAI has just delivered a master class in how to reframe pricing and change the anchor price. The anchor price provides a reference point that buyers use to decide if a price is reasonable (not acceptable, just reasonable). As you introduce new AI based functionality you may want to reframe the anchor price of your solution.
