Compute: The Cost of Thinking-by-Machine

Computation has gotten relentlessly cheaper for decades — and when a useful thing gets nearly free, it stops being a discount and starts unlocking new worlds.

The Future of Technology · Lesson 7 · 7 min read

A task that would have cost a fortune in computer time decades ago — recognizing faces across millions of photos, or simulating tomorrow’s weather — now happens on the device in your pocket for a fraction of a penny. Nobody “decided” these should be cheap; a steady force made them so. What is that force, and why does a falling price change what is possible, not just what is affordable? Hold that question; by the end you should see why a boring-sounding cost decline is one of the deepest forces of all.

Compute is operations on information — and it keeps getting cheaper

“Compute” just means doing operations on information: arithmetic and logic steps on data. The deep force is not that computers exist — it is that the cost of each operation has fallen dramatically and steadily for decades. More gets packed into each chip, and the price per calculation keeps dropping.

It is not one magic trend but a stack of S-curves (Lesson 3): each chip-making technology eventually saturates, and a new one picks up the climb. The result, viewed from far away, is one of the most reliable cost declines in the history of technology.

Falling cost changes what is possible, not just what is affordable

Here is the part people miss. When a useful capability gets, say, a thousand times cheaper, you do not just do the old things for less — you cross thresholds. Things that were absurdly expensive become routine, and you start using the capability lavishly, in ways that were never worth it before.

Cheap compute is why we can afford to store everything, brute-force hard problems, and put a computer in every object and pocket. The phone, the internet, and modern AI are not separate miracles — they are what happens when thinking-by-machine becomes nearly free.

Weather prediction: decades ago, simulating tomorrow’s weather needed a room-sized supercomputer and a national budget — only governments could. As compute cost collapsed, the same simulation became cheap, so forecasts are now on everyone’s phone, updated constantly, and far more accurate (you can afford to compute much more). The forecast did not just get cheaper — it became ubiquitous and better, and wholly new things, like constant global models, became possible.

Why compute is the keystone force

Run it through Lesson 6’s three tests. General? Every domain uses it. Improving? The decades-long cost curve. Enabling? The internet, mobile, and AI all ride on cheap compute.

But compute has a special role: it amplifies the other forces. Cheap computation makes energy grids smarter, turns biology into something you can analyze and design, and is the literal substrate AI runs on. It is the force that multiplies forces — which is why it sits near the root of almost every story about the future.

Suppose a capability — say, automatically reading text in any photo — costs $1 per image at first. At that price almost no one uses it. Now let compute cost fall:
1. At a few cents, a company digitizes its archives — a cost saving on an old task.
2. At a fraction of a cent, your phone reads signs in real time as you travel — a new everyday use.
3. At effectively free, it gets embedded everywhere by default — in cameras, apps, and tools you never think about.
Notice the capability never changed; only its price did. Each time the cost crossed a threshold, a new class of uses switched on. That is why cheap compute keeps surprising us.