Moore's Law, the rule of thumb that has shaped strategic planning for sixty years, says the number of transistors on a chip doubles roughly every twenty-four months. [1]
METR, an AI evaluation research organization, measured how long a task an AI system can complete autonomously and found that capability is doubling every seven months. Post-2023, the doubling time has accelerated to about 4.3 months. [2]
Moore's Law: every twenty-four months. AI capability: every four to seven months. That is somewhere between three and six times faster than the cadence executives have been using to plan around technology shifts for their entire careers.
The problem is not that Moore's Law is wrong. The problem is that most senior leaders are still using it as the mental model for AI, and AI is not running on the same clock.
What Moore's Law Actually Said
In 1965, Gordon Moore observed that the number of components on an integrated circuit had been doubling every year and projected that pace would continue. In 1975, he revised the doubling time to every two years.
That observation became the rhythm of the technology industry. Twenty-four months between meaningful upgrades. Twenty-four months between the chip you bought and the one your competitor bought. Twenty-four months was enough time to plan a product launch, train a team, and amortize a vendor contract.
Most strategic planning still implicitly assumes that twenty-four month rhythm. Five-year plans. Three-year IT roadmaps. Eighteen-month vendor contracts. The cadence assumption is baked into the math.
It is also a hardware curve. Moore's Law describes how many transistors you can pack onto a piece of silicon. The improvement happens when manufacturing gets better.
What Is Actually Happening With AI
AI capability is not a hardware curve. It is a software curve being driven by hardware, training data, model architecture, and increasingly, AI is itself helping to design the next generation of AI.
METR's measurement is concrete. They tested whether AI systems could autonomously complete tasks that take humans a known amount of time. Five years ago, the frontier was tasks that took a few seconds. Today it is tasks that take a few hours. That capability has been doubling roughly every seven months on average since 2019, and every 4.3 months since 2023.
Project that forward. If the trend holds, autonomous AI capability will reach week-long tasks in two to four years. Then month-long tasks. Then quarter-long projects.
That is not a smooth Moore's Law curve. That is a much steeper one.
Why the Curve Is Steeper
Three reasons compound.
First, AI improvements stack. A better model, run on better hardware, trained on better data, with better fine-tuning and better orchestration, multiplies. Each layer can improve independently and at its own pace. We are dealing with exponential growth now.
Second, the AI models are starting to help build the next generation. Researchers are using AI to write the training code, to design the model architectures, to evaluate outputs. The technology is recursively improving the technology.
Third, the diffusion is faster. Moore's Law improvements showed up in shipped products months or years after the chip improved. AI improvements show up the same week the model is released. There is no manufacturing lag.
This is why a tool you bought eighteen months ago can already feel obsolete. It is not your imagination. The substrate moved.
What This Does to Your Planning Windows
If you assumed Moore's Law was the right cadence for AI strategy, your planning windows are between three and six times too long.
Three-year IT roadmaps that assume the AI tools available today will still be relevant in 2029 are operating on a fiction. The vendor you sign a thirty-six-month contract with is locking you into a product you would not buy at the eighteen-month mark.
The five-year strategic plan that has "AI" listed as a 2027 priority is a five-year strategic plan that has already missed its window for that priority. The competitive landscape will look nothing like today's by then.
This is the reason fluent executives are now budgeting in quarters for AI tooling, requiring exit clauses on AI vendor contracts, and rerunning their AI strategy as a recurring agenda item rather than an annual review.
The Question Worth Sitting With
If your three-year IT roadmap is built around the assumption that the AI tools available today will still be the AI tools that matter in 2029, what is your roadmap actually planning for?
The cadence under your strategy is not the cadence the technology is running on.
Sources
[1] "Moore's law," Wikipedia.
[2] METR, "Measuring AI Ability to Complete Long Tasks," March 2025 and "Time Horizon 1.1," January 2026.
For those of you warming up in the comments, yes, I obviously used AI to write this. That's my whole point: the ideas are mine, drawn from a five-page free-flowing brain dump and from real conversations I've had with people at all ends of the AI knowledge spectrum. AI helped me organize, tighten, and get the words on the page faster than I could on my own. I have been telling you throughout this series that AI is not here to replace you, it is here to make you more efficient. This article is the proof.