Every few years someone announces that the last technology wave is over. It never is. The web did not end when the cloud arrived, and the cloud did not end when phones took over. The waves do not replace each other. They stack, like tectonic plates, each one resting on the ones below and pushing everything above it a little higher.


Look at the stack from far enough away and one pattern shows up every time: each plate makes one expensive thing nearly free, and the value moves to whatever is still scarce.


Each plate rests on the ones below it. None replaces the others.

1995: the dot-com web made distribution free


Netscape went public in August 1995, and for the first time a small company could put its catalogue in front of anyone with a modem. Printing, shipping and shelf space stopped being the price of reaching a customer.


The bubble that followed burst in 2000, and the lesson people took was "the internet was hype." The real lesson was quieter. Distribution did become free. What stayed scarce was trust: the brands people already knew how to find and believe. Search made that even clearer, because being found became its own discipline.


2006: the cloud made infrastructure free


Amazon opened S3 and EC2 to the public in 2006. Before that, starting a software company meant buying servers, renting space for them and hiring people to keep them running. After it, a credit card and an afternoon were enough.


When infrastructure became a utility, the scarce thing became speed: who could ship, learn and ship again the fastest. That is why so much of the next decade was built by small teams.


2007: mobile made being reachable free


Apple introduced the iPhone in January 2007 and opened the App Store in July 2008. The computer moved into the pocket, and with it the customer's attention, location and camera.


Being reachable everywhere stopped costing anything, which made attention the scarce resource. Businesses learned to meet customers in the feed, the map and the notification rather than at the counter. Reviews, photos and hours on a map listing quietly became part of the storefront.


The 2010s: SaaS made running software someone else's problem


Salesforce had been selling software as a subscription since 1999, but it took the cloud and the phone to make it the default. Through the 2010s, every business function got its own app: one for the website, one for the books, one for email, one for scheduling, one for the ads.


Software became cheap to start and easy to rent. The scarce thing became coherence. A typical small business now runs ten tools that each hold a slice of what it knows, and none of them talk to each other. The seat price is low; the cost of the gaps between the seats is not.


2022: AI made the first draft of thinking cheap


ChatGPT opened to the public on November 30, 2022, built on a decade of work that began with deep learning in image recognition and the transformer architecture in 2017. For the first time, the first draft of a lot of knowledge work became almost free: the email, the summary, the ad, the product description, the reply to the customer.


This plate is different because it touches judgment-shaped work, not just tools. But the pattern holds. When drafts are free, the scarce things are the ones a model cannot invent on its own: your data, your rules, your taste, and someone accountable who says yes or no before anything goes out. A business that gives AI its knowledge and keeps the approvals gets leverage. A business that gives AI a blank page gets noise.


Now: robotics starts on physical work


The newest plate is moving under factories, warehouses and kitchens. Collaborative robots that work beside people, vision systems that inspect a part in a fraction of a second, and machines rented by the hour instead of bought for a fortune are already here. What changed is the same thing that changed for software: the brains got cheaper, so the arms became useful in more places.


Robotics will not make physical work disappear any more than the web made stores disappear. It will make repetitive physical work cheaper, first in the places where it is boring, dangerous or impossible to staff. The scarce thing will be what it always becomes: knowing which job to hand over, designing the work around people, and having someone who can fix the line when it stops.


What the stack means for a business today


Read the plates together and three rules fall out.


  • Own the layer that stays scarce. Distribution, servers, apps and drafts all became cheap. Your relationships, your know-how and your data did not. Keep them in one place you own, not scattered across tools that can change their prices or close.
  • Keep a human where the decision is. Every wave tempted people to automate the judgment along with the labour. The businesses that lasted automated the labour and kept the judgment, with clear approvals for anything public, priced or paid.
  • Rent the new plate before you buy it. Nobody needed to build a data centre to use the cloud. Nobody needs to train a model to use AI, and most businesses will not need to buy a robot to use robotics. Start with one job, measure it, and let the result decide the next step.

The plates will keep moving. The question is never which wave is next. It is which of your scarce things the next wave will make more valuable, and whether you are ready to put them to work.