Is AI a Platform Shift?

You can think of a platform shift as a change in the dominant layer that applications are built on. It might look like Software > Hardware. Or Hardware > Software. A new layer becomes the dominant layer (e.g., most applications are being built on top of the new layer rather than the previous layer). As a result, the new layer might be able to capture the majority of the value from the previous generation (e.g., it might accrue the majority of the profits). New companies become dominant. New or different architecture on top of the layer. A change in the business models for the layers. Some potential platform shifts over the last few decades:

• Personal Computers

• The Internet

• Mobile

• Cloud

From those examples, some necessary but not sufficient requirements for a platform shift:

• Changes where distribution aggregates. Mobile shift distribution from desktop. Consumers spent more time interacting with software via their phones than they did on their desktop computers. This is why Google needs to pay Apple billions of dollars every year to remain the default search option on mobile.

• Changes the business model (possibly in a disruptive way). Cloud shifted the business model from being deployed on-prem to software-as-a-service. The delivery model dictated a new business model, from licensing to subscription (or usage-based). The Internet changed the dominant software business model — instead of selling individual copies of software (e.g., productivity software), the most profitable option was to capture the zero marginal cost of distribution. Cloud offered a different financing model for businesses (elastic pay-as-you-go operating expenses versus the traditional capital expenditures of buying or leasing data centers).

• Changes what’s possible. A platform shift usually enables completely new workflows in addition to augmenting older ones. Mobile offered new contextual APIs (e.g., Location and Camera) that enabled new applications that weren’t possible on desktop (e.g., Uber and Instagram). Digital marketplaces couldn’t exist before the globally networked computers (e.g., eBay). I guess you could sell something over LAN.

So, through this lens, is AI a platform shift?

• Changes where distribution accrues? Potentially. AI turns the marginal cost of content to zero. Whenever the marginal cost of something in the value chain is set to zero, this usually has a downstream effect on where distribution aggregates (e.g., the Internet turned the marginal cost of software distribution to zero). This could manifest itself in a few ways — new aggregators replace the old ones (e.g., Google, Reddit, StackOverflow, Quora). In Google’s case, it makes the aggregation of quality content much more difficult (there will be much more content), and in the case of the sites that both aggregate and own content, they suffer a quality problem and a moat problem (it’s easy for competitors to bootstrap with AI data).

• Changes the business model? Probably not. It could shift software to follow more usage-based pricing (vs. subscription), but it would still likely be delivered as SaaS. Maybe smaller models will go on device, and that could shift us from a Software > Hardware cycle (commoditize the software, sell the hardware).

• Changes what’s possible. Clearly true. There are workflows that will be drastically changed by AI.

So, will AI be a platform shift? Probably. Yes, there will be incumbents who accrue significant value — but that’s true of any platform shift. Microsoft navigated most of these (and often came out stronger). Amazon was borne of the Internet era and individually kicked off the cloud era. Meta was started in the Internet era but successfully pivoted to mobile (and might turn the corner on the potential AI shift). But there will also be new platforms that replace the old ones. Many Internet-era companies were outcompeted by their mobile counterparts. Cloud software obsoleted a great majority of licensed software (save Adobe, who pulled off a miracle).