What became visible

Reading McKinsey Global Institute’s The AI Economy: Interconnected Forces, Feedback Loops, and Speeds of Change, I was struck less by its account of accelerating AI capability than by the sequence of constraints that acceleration keeps revealing. Frontier chips encounter advanced-packaging constraints. Expanded compute encounters electricity and grid constraints. Physical infrastructure encounters construction constraints. As those constraints are relieved, the pressure moves again—toward applications, skills, organizational workflows, and the institutions expected to absorb what the technology makes possible.

There is a useful inversion hiding inside this pattern. We tend to speak about technological progress as though each breakthrough removes a bottleneck. It does, locally. But at the level of the larger system, the bottleneck often does not disappear; it migrates. Solving one constraint transfers pressure to another part of the system, where the capacity to respond may have been developing on a very different clock.

That makes the location of scarcity more interesting than the rate of capability improvement.

Abundance moves scarcity outward

The same pattern may extend much further than physical infrastructure. When computation becomes abundant, power becomes scarce. When intelligence becomes abundant, context becomes scarce. When generation becomes abundant, evidence and provenance become scarce. When autonomous execution becomes abundant, legitimate authority becomes scarce. When possible actions multiply, judgment about which actions should be taken becomes more consequential.

This suggests that AI does not lead us toward a world without scarcity. It continually changes what is scarce.

That distinction matters because enterprises are still concentrating enormous attention on the thing becoming abundant: models, tokens, agents, generated software, automated workflows, and machine intelligence. Yet strategic advantage may increasingly accumulate around the complementary resources that abundance makes scarce. Proprietary context, trusted evidence, enforceable policy, legitimate authority, institutional memory, human judgment, and the capacity to coordinate them may become more important precisely because intelligence itself becomes easier to obtain.

The architecture of the enterprise therefore has to follow the moving bottleneck. An organization designed around yesterday’s scarcity will continue optimizing what is becoming plentiful while neglecting what has quietly become limiting.

Different clocks

McKinsey’s systems framing makes another feature of this movement visible. The components do not evolve at the same speed. Model capability can change over months. Data centers take longer. Electrical infrastructure can take years. Organizations redesign workflows more slowly still. Regulation, institutional norms, professional practices, and social expectations operate on their own clocks.

The difficulty is therefore not simply that the bottleneck moves. It can move from a fast system into a slow one.

That may explain some of the strange character of the present AI transition. Extraordinary capability can coexist with disappointing organizational outcomes without either observation being contradictory. The technology may have crossed one threshold while the surrounding system remains constrained somewhere else. What appears to be an “AI problem” may actually be a coordination problem between systems moving at different velocities.

For leaders, this changes the question. Instead of asking only how quickly AI is improving, we should ask where the constraint created by that improvement is moving next—and whether the surrounding institution can move when it arrives.

The question I want to keep

Perhaps every significant increase in abundance creates a corresponding scarcity somewhere else.

If so, the enduring strategic question of the AI economy may not be:

What can intelligence do next?

It may be:

What becomes scarce when intelligence can do it?

That feels like a question worth carrying into the woods.