"We know AI is important. We do not yet know where the durable value will accrue."
AI is rapidly becoming one of the most important technology shifts in decades.
However, while adoption is accelerating, the industry's economic structure remains unclear.
The key uncertainty is no longer:
✗ Will AI matter?
The key uncertainty is:
→ Who captures the profits?
The technology is no longer the question.
The business model is.
Many future winners may not even exist today.
Hyperscalers are spending hundreds of billions of dollars on AI infrastructure.
The spending continues despite uncertain ROI.
Simple.
Overspending can be fixed.
Missing the next platform shift cannot.
Every major technology cycle produced a similar investment surge.
Expect:
Do not automatically expect:
Most discussions focus on:
Evans argues this may be the wrong place to look.
The model becomes:
Electricity.
The value shifts to:
Appliances.
The largest AI businesses may not be model companies.
They may be workflow companies.
Why has software development moved first?
Because success is measurable.
Compare that with:
where outcomes are harder to measure.
AI adoption spreads fastest where ROI can be quantified.
Perhaps the most useful mental model in the deck.
Think of AI as an unlimited supply of:
Instead of asking:
"How many employees do we need?"
Ask:
"How many AI-assisted workers can one expert supervise?"
Future organisations may be built around:
Expert + AI Team
rather than
Manager + Large Human Team
A critical distinction.
A job is a collection of:
Tasks are:
AI attacks tasks.
Not necessarily jobs.
Most professions are more likely to evolve than disappear.
This may be the most under-discussed AI challenge.
Historically:
Junior Analyst → Senior Analyst → Portfolio Manager
Junior Lawyer → Associate → Partner
Junior Engineer → Architect
If AI performs junior work:
Who trains future experts?
AI may improve short-term productivity while creating long-term capability risks.
Many organisations have not yet thought through this problem.
The deck ultimately revolves around one framework.
Chips. Cloud. Models. Data Centres.
Characteristics:
Workflow software. Industry solutions. Operational tools.
Characteristics:
Potentially the most attractive layer.
Proprietary operational data.
Characteristics:
Customer access. Trust. Brand. Ecosystems.
Characteristics:
The market currently values AI companies based on intelligence.
History suggests the largest businesses are usually built on:
rather than technology alone.
If that pattern repeats, the biggest AI winners of the 2030s may look less like AI labs and more like industry-specific infrastructure and workflow companies.
The most important question is not:
"Which model wins?"
The more important question is:
"Where in the AI value chain will durable profits ultimately reside?"
BENEDICT EVANS – SPRING 2026 AI DECK