Chapter 1: What does a chip actually do?

In the preface, we said "our lives ride on top of chips." But the moment someone asks, "So what is a chip?", plenty of people find themselves stuck for an answer. That's perfectly fine. In this chapter, we unravel the very foundation — what a chip is, as a part, and what it does — from zero, carefully. Once this settles in your gut, everything that follows will come into clear view.

1.1 What is a "chip," in the end?

Let's start with the word. A semiconductor chip (often just called a "chip," an "IC," or an "integrated circuit") is, put very roughly, "an extremely small board that computes using the force of electricity." The material is mainly silicon — yes, the same silicon found in sand and quartz. A relative of the sand that's everywhere on Earth has become the brain that supports human civilization. That alone is a slightly wondrous thought.

So what is that "small board" doing inside? In fact, the root of what it does is astonishingly simple. It is "just flipping switches — an enormous number of them, at an enormous speed, on and off."

1.2 A chip's true nature is a "swarm of switches"

Inside a chip, parts called transistors are packed in, densely. No need to overthink it. A transistor is, in essence, a very small switch that flips between letting electricity "through / not through." Same role as the light switch in your room. What differs is size, number, and speed.

"Tiny switches that just flip between letting electricity through or not" — moving in unbelievable numbers, at unbelievable speed. This is the true nature of a chip. So how does that plain on-off become "computation"? Here is the most interesting part.

1.3 The modest magic by which on-off becomes "computation"

The key is to treat a switch's on-off as a "number." We decide that a switch turned on is "1," and one turned off is "0." With just this one agreement, the swarm of switches becomes able to represent "numbers."

   The single agreement that turns switches (transistors) into "numbers"
 ───────────────────────────────────────────────────────────────
     electricity is flowing      →   "1"
     electricity is not flowing  →   "0"
 ───────────────────────────────────────────────────────────────
   Line up many switches, and a run of 1s and 0s can represent any number.
     e.g.)  0 1 0 1 = one number      1 1 1 1 = another number
   → Then build, as circuitry, "how to rearrange 1s and 0s into addition or
      comparison," and the swarm of switches begins to 'compute.'

This way of representing numbers with only 1s and 0s is technically called "binary." No need to memorize the name. What matters is this one point: a "run of on-offs" can represent numbers, and by devising how the switches are connected, you get addition, comparison, and memory. A chip is a machine that keeps doing this "rearranging of 1s and 0s," endlessly, at a speed faster than the eye can follow.

To put it another way. Writing every number, letter, and image out of just two characters, 1 and 0 — this is like building any enormous creation out of just a few kinds of LEGO brick. Or like writing every book in the world with the 26 letters of the alphabet. The kinds of part are surprisingly few, but the ways of combining them are infinite, so infinite things become possible. A chip's underlying power comes from here.

1.4 But this is the "digital" story ― there are plenty of semiconductors that don't compute

Everything so far — "computing with 1s and 0s" — is in fact only one region of semiconductors (the digital region). The world is also full of semiconductors that don't compute; if anything, these are the ones supporting daily life. The three representatives are:

The lead of this corner is digital (the computing chip), but semiconductors are not only that — keep this map in a corner of your mind.

1.5 On a board the size of a fingernail, billions of them

Back to the matter of numbers. A chip these days packs billions to tens of billions of transistors (those switches) into an area about the size of a fingernail. Hard to picture, isn't it? The population of Tokyo is about 14 million. All of Japan, about 120 million. Many times that number of "switches" sits on a single fragment of a fingernail — put that way, does the strangeness of the density come across a little?

※ The exact numbers differ completely by generation and product. The figures "billions" and "tens of billions" are only rough guides for grasping the scale. The order of magnitude changes with the kind and generation of chip, and it increases year by year with advances in technology. In this corner, we won't compete over exact counts. It's enough if you feel: "at any rate, it's a quantity too large for a human to intuit."

This competition — over how small, and how densely, you can pack them — is the history of semiconductors itself. And the common sense of that competition is now wavering, which connects to the "2 nm" and "going 3D" story we'll take up later (Chapter 6). For now, it's fine to just remember "extremely small, extremely densely packed."

1.6 The "black part" and the "small board inside" are different things

Let me untangle one easily-confused point. Look at the board of an electronic device and you'll see black, square parts with lots of metal legs sticking out. Many people think that thing itself is "the chip." But strictly, that's a little off.

That black part is the "package" — a kind of lunchbox. The real lead is the very small silicon board, carefully stored inside that box. This is called the "die." The computing body we call "the chip" is this die. The black package is merely the box and terminals that protect the delicate die from shock and dust, and exchange electricity with the outside world.

Why does this distinction matter? Because the "shift in how chips are made" (Chapter 7) and the "going 3D" (Chapter 6) that appear later in this corner are mainly stories about this die. Not the lunchbox, but the way the contents are made, is what's about to change — hold that, and what's ahead will slide right in.

1.7 This chapter's summary, and the bridge to the next

Take just three things home from Chapter 1.

• A chip = a small board that computes with electricity. The material is silicon, a relative of sand.
• Its true nature is a "swarm of switches (transistors)." By treating on-off as "1 and 0" and rearranging them in tremendous number and speed, computation is born.
• The "black part" is the lunchbox (package); the computing body is the small board inside (the die).

Now, up to here we've talked about "what a chip is" — the shared ground floor. But — say "chip" and, in fact, there are several kinds of "brain," each good at different things. Some are all-rounders; some are specialists, uncannily fast at one particular job. And right now, the kinds are increasing explosively. In the next chapter, Chapter 2, let's go and see the heart of the "Cambrian explosion" — the diversification of the "species" of brain: CPU, GPU, TPU, and further, quantum and bio.