A study by Epoch AI illustrates the monetization issue faced by language model developers, where continual needs to invest in the next iteration of a model essentially wipes out profits, raising questions about whether any given standalone model can sustain itself solely by direct revenue sources.
Evaluating OpenAI, the researchers suggest that ChatGPT 5 has, so far, failed to generate sustainable profits, considering all the iterations (GPT-5 and GPT-5.1, GPT-4o, ChatGPT).
But model lifecycle losses aren’t necessarily permanent. AI models arguably do not need to be profitable today, as long as companies can convince investors that they will be in the future; can create clear value and therefore develop new revenue mechanisms and cost-saving mechanisms.
As sometimes is the case, avoided costs become the value driver, in part, as much as higher revenue is important.
And, over time, costs to create the next iterations and run inference operations should also decline. At the level of market structure, an oligopoly (“winner take most”) should develop, as has been the case in other computing market segments.
That alone should boost revenues and profit margins for the eventual leaders.
In some ways, the challenge is similar to that faced by upstart internet firms including Google search and Facebook social media, neither of which had a revenue model at all, at first.
But indirect monetization strategies include the “freemium” model and a mix of subscription, licensing, advertising or other services offering a path to sustainability.
The key is leveraging value in the broader ecosystems, bundling AI capabilities into existing offerings to boost user engagement, retention, conversion, or operational efficiency, leading to revenue growth elsewhere.
It's particularly effective when AI costs are absorbed as a "value-add" rather than a line-item expense, allowing companies to recoup investments through increased customer lifetime value or ancillary streams.
But that also suggests the the firms best able to monetize their language model investments are those with robust existing revenue models that AI can materially enhance, such as:
Google's integration of Gemini into its search engine (advertising)
Amazon personalizing product recommendations and ads (commerce)
Meta personalizing social media ads
Robotics (hardware sales and support)
Autonomous vehicle services or hardware such as Waymo enabled by AI sold as a vehicle purchase, upgrade or subscription
Enterprise software licenses
“Insight” data such as credit card transaction trends and behavior
AI infrastructure (chips, data center services)
Freemium or marketplace models (premium versions, fees, commissions)
The issue for firms without preexisting revenue models or platforms is that some of these options do not exist, or would be hard to create.
But we also have to remember that was the case for Facebook and Google; Airbnb and Uber; Amazon and Netflix. It might also have been equally true for Microsoft Windows or Android.
Sustainable models were discovered over time. Language models might be a developing case of that.
No comments:
Post a Comment