Tuesday, July 1, 2025

Will AI Disrupt Internet Business Models?

One reason for studying the business implications of artificial intelligence is that we may be seeing a significant shift in the role played by near zero marginal cost in shaping feasible business and revenue models. 


We might all agree that near-zero marginal cost has been fundamental for content, social media and other app pricing and revenue models. Network effects and scale have been fundamental precisely because marginal cost has been so low. 


Low marginal cost arguably is responsible for “winner take all” market structures; the ability to support software products using advertising; the global character of markets and the importance of “time to scale.”


Marginal cost is the cost of producing one additional unit of a good or service. For non-tangible products such as software, streaming, or cloud services, this cost is often close to zero once the initial product is developed and infrastructure is in place. 


Low marginal cost means that digital businesses can grow a user base quickly without proportional increases in expenses. That is vital for businesses built on network effects, where each additional user increases the value for others (social networks, marketplaces).


Also, low marginal cost also means attackers can undercut incumbent pricing levels, often making higher margins on lower retail prices (“free” use and “freemium” models). 


Business Element

Traditional Business

Internet Business (Low Marginal Cost)

Cost per additional user

Increases with scale

Remains near zero with scale

Profit margin

Shrinks with volume

Grows with volume (after fixed costs)

Growth constraints

Physical/logistical

Virtually unlimited (digital)

Network effects

Limited

Strong, self-reinforcing


All that can enable a business model where adding more users means lower average cost per user as scale grows. That is less true for a traditional “physical” model, where cost tends to scale in a more-linear fashion. 


The implication is that AI possibly disrupts many foundational internet app, service and content models, where zero-to-low marginal cost is the economic foundation. 


The essential difference for AI-based models is that very-low marginal cost might not be so substantial.


Large language models incur non-trivial costs for training, inference, and maintenance that arguably are more linear cost drivers than we have gotten used to for many internet apps. 


Unlike traditional cloud-based internet delivered  software, where serving additional users involves negligible database or bandwidth costs, AI inference costs are directly proportional to user activity. 


On the hardware side, AI processing tasks arguably also involve data infrastructure requirements that also scale in a more-linear way. 


Traditional internet platforms might have marginal costs per user interaction estimated at $0.0001–$0.001, for example. 


AI services might have marginal costs per interaction closer to $0.01–$0.50, depending on model complexity and usage patterns.


The business model implications, if this gap does not close, is that AI model marginal costs could be higher by 100 times or more. 


And that could key implications for the value of scale and likely revenue models. Is the internet business model about to be disrupted?


No comments:

Stablecoin Use Cases Seem Tailor-Made for Cross-Border Transactions

With the caveat that new technologies sometimes enable use cases we have not thought about, and despite the interest a few hyperscale consum...