The Invisible Trillion

The Other Ledger

What All This Infrastructure Is Actually For

Earlier chapters laid out what chips and data centers cost — the energy, the water, the mining, the disposal. I tried to be honest about each of those, including the parts that are genuinely bad.

This chapter is going to do the same thing in the opposite direction. What does all this infrastructure actually do? What gets made, possible, better, cheaper, safer, or more widely available because the chips exist? Because that list is also real, and leaving it out would be as dishonest as pretending the costs don’t exist.

I want to warn you about the tonal trap in this chapter, because it’s the one that every tech-friendly book falls into and which I am going to try to avoid. The trap is that once you start listing good things, you start selling. The writing gets warm, enthusiastic, a little breathless. Cool inventions! World-changing breakthroughs! Better future! By the end of the chapter, you are not reading a ledger; you are reading a brochure.

I will try not to do that. This chapter is a ledger. The items on it are real, some of them are extraordinary, and none of them erase what was in the preceding chapters. You should finish this chapter feeling that the costs and the benefits are both large and both real, and that the interesting questions — which the rest of the book is about — are how to get more of the second while paying less of the first.

Fair?

Medicine

Let me start with the thing that is, in raw human terms, the biggest item on the ledger.

A CT scanner is a chip-dependent machine. So is an MRI, an ultrasound, a PET scanner, a modern blood analyzer, a pulse oximeter, a continuous glucose monitor, a pacemaker, an insulin pump, a cochlear implant, a hospital telemetry system, and an electronic health record. If you walked into a 1970 hospital and a 2026 hospital, the biggest visible difference would not be the architecture; it would be that the 2026 hospital has computers and screens attached to essentially everything, and the 1970 hospital does not.

The consequences of this, averaged across the population, are gigantic.

Breast cancer mortality in the United States has dropped by roughly 40% since 1990. A major reason is mammography screening and, more recently, AI-assisted mammogram reading that catches tumors earlier and with fewer false positives than human radiologists working alone. Earlier detection means earlier treatment, which means more survival.

Heart attack mortality has dropped by more than half since 1980. Chip-enabled improvements are not the only cause — drugs matter, and surgery matters, and anti-smoking campaigns matter — but continuous ECG monitoring, automated defibrillators in public spaces (all of them chip-driven), and better imaging that lets cardiologists see exactly where an artery is blocked are all part of the story.

Childhood cancer survival has gone from around 10% in 1960 to around 85% today. Almost no one over 50 has not lost a relative to a cancer that today’s children would survive. The reasons are many, but precision imaging, targeted radiation (which requires chip-controlled delivery), and treatment protocols generated from enormous patient databases are major pieces of the pie.

In 2020, a team at DeepMind — the AI research lab owned by Google — released a system called AlphaFold that predicts the three-dimensional shape of proteins from their genetic sequences. This doesn’t sound dramatic unless you know the backstory, which is that predicting protein shape had been one of the great unsolved problems of biology for fifty years. Tens of thousands of researchers had spent entire careers on individual proteins. AlphaFold solved the general problem, published its predicted structures for essentially every known protein — more than 200 million of them — and released the database for free. This has accelerated drug discovery, enzyme design, and basic biology research in ways that are still playing out, but already visible. AlphaFold runs on GPUs in a data center. Without the chips, the solution does not exist.

None of this means medicine is perfect or fair. Care is still distributed unevenly, access is still gated by money, and plenty of medical AI is overhyped and oversold. But on average, across the population, the chip-dependent part of medicine is saving an enormous number of lives, and the ones it is saving include, almost certainly, people you love.

All 19 chapters of The Invisible Trillion · Library · All courses