The Temporal Trilemma

The energy trilemma has no time axis. For AI infrastructure, that is the axis that matters most.

The energy trilemma is usually drawn as a triangle. Three corners: security, affordability, and sustainability. The lesson it teaches is that you cannot fully satisfy all three at once, so you choose a balance and accept the trade-offs. The World Energy Council has measured national performance against those three dimensions since 2010, and the framework has become the default way the industry talks about energy choices.

‍A triangle is a useful picture. It is also a still picture. The thing it describes does not hold still.

Why Power Constraints are Limiting AI Data Center Infrastructure

I argued in 2016 that power, not compute, would ultimately be the binding constraint on data center infrastructure, and onsite power would be a necessity. The consensus at the time was that chips and capital set the pace. They did not. The pace is set by how quickly you can put a megawatt where the load is. That shift, from compute as the scarce resource to power as the scarce resource, is what exposes the missing piece in the trilemma.

‍The trilemma tells you what to balance. It says nothing about when, or about how fast the balance can move, or about which direction it moves in. It has no time axis. For most of its working life that omission did not matter, because energy systems changed slowly. Policy moved in decades. Capacity moved in years. The balance point drifted, and an annual index was a perfectly good way to watch it drift.

How AI Growth Interrupted Traditional Grid Planning

‍AI infrastructure broke that assumption.

‍The International Energy Agency (IEA) reports that electricity demand from data centers rose 17 percent in 2025, against 3 percent growth in global electricity demand overall, and projects data center consumption to double by 2030 with AI-specific demand tripling. In the United Kingdom, Engineering & Technology reports that the proposed pipeline of data center projects could require 50 gigawatts, around 5 gigawatts more than the country's entire current peak demand.

This is not slow drift. It is a fast transient landing on a frame built for slow drift.

The balance moves backward, not just forward

‍Watch what the transient does to the three corners. Faced with a multi-year wait for a grid connection, developers are choosing speed and paying for it on the sustainability corner. Axios reports that roughly 30 percent of planned data center power capacity is now expected to be generated on-site, up from almost nothing a year earlier, that Chevron is building a natural gas plant dedicated to a single data center, and that operators are islanding sites, running them off-grid for years rather than waiting for the grid to catch up. The industry has a name for the priority driving all of this: speed to power, the time between project approval and energization. An industry framework launched by NEMA, ASHRAE, and the Pacific Northwest National Laboratory is built around shortening exactly that gap.

‍Read against the trilemma, this is a system giving ground on sustainability to buy time on security and delivery. Decarbonization progress that took a decade to accumulate can be set aside in a single procurement cycle. The balance does not only move. It moves backward.

‍That is the part the static triangle cannot show, and it is the part that matters most right now.

Naming the missing dimension: the Temporal Trilemma

‍So let me name it. I call this the “temporal trilemma”. The three corners do not change. Security, affordability, and sustainability are still the things in tension. What the temporal trilemma adds is the axis they travel along: time. This is not a fourth corner. Time is not another thing to balance against the other three. It is the dimension in which the balance moves, and it now moves fast enough, and reversibly enough, that ignoring it produces bad decisions.

Isn't it obvious the trilemma changes over time?

‍This is worth saying plainly, because the obvious objection is that everyone already knows the trilemma changes over time. They do. The World Energy Council has been evolving the framework for years, broadening what each corner contains and exploring concepts such as dynamic resilience. But those revisions enrich the corners. They make security and sustainability mean more. They answer the question of what to balance. They still do not answer the question of at what speed, and in which direction, the balance is moving, and whether you can respond before it moves again? That is a different question, and it is the operative one for anyone building infrastructure on an AI timescale.

Managing Energy Grid Delays with a Structured Transition Model

‍If the problem is temporal, the answer is sequencing.

‍This is the work the Structured Transition Model (STM) does. Rather than picking a single balance and defending it, you decide what to deploy now, what to run as a bridge, and what to converge on, so that meeting today's speed requirement does not foreclose the sustainability requirement you will still have to meet later. The trilemma asks you to pick a balance. The temporal trilemma tells you the balance will move. STM is how you stay ahead of the movement instead of reacting to it.

‍None of this makes the triangle wrong. Security, affordability, and sustainability are the right three corners. The point is narrower, and more useful: the trilemma is incomplete in exactly one dimension, and that dimension, time, is now the one doing the most work. Draw it as a still picture and you will keep being surprised by a system that refuses to stand still.‍