Table of Contents – Full Book = 2 hour read
Chapter 1: Nvidia and their GPUs (now live)
Chapter 2: Data Centers: Construction and Function (now live)
Chapter 3: Hyperscalers and their Trillion Dollar Bets (now live)
Chapter 4: Wall Street is a Casino With A Church Attached (coming soon)
Chapter 5: Big, Cash Burning Players: Anthropic and OpenAI (mid-December)
Chapter 6: Real Demand for AI compute? (mid-December)
Chapter 7: A vision of superintellience (mid-December)
Chapter 8: A vision of global depression (mid December)
Introduction (45 second read):
The title and cover of this book are both misleading.
I love compute. If you don’t, I assume you’re an enlightened shepherd in the mountains. I believe that A.I. is a revolutionary technology that will change the world for the better, and I acknowledge the necessity of data centers to make this new, more abundant, more prosperous world happen.
If you are looking for a book to hate on A.I., how it is eroding our ability to reason, crowding out human artists, exacerbating propaganda, increasing cyberattacks, spreading scams, widening inequality, creating fear of impending unemployment, or encouraging the mentally stable towards heinous acts, this is not that book. There are plenty of those already.
If you are looking for reasons to bomb a data center: the noise they create, the enormous water and electricity they consume, the displacement of residents, the political backlash, this is also not that book, although data center’s impacts will be touched upon throughout the narrative.
This book is about capitalism’s excesses and pressures, a deep analysis of the largest infrastructure build out in human history, what this means for our near-future (3-5 years), and how we can prepare.
At the time of this writing, $700-$900 billion dollars a year are being spent on building data centers. To put this spend in perspective, during the 1800s, when Americans were building out railroads, $100-$200 billion/year (inflation adjusted) was being spent on railroads. During the Internet infrastructure build out, we were spending $150-$250 billion a year on fiber optic cables. Even if we account for the world’s increased GDP since the late 2000s (x3.6), we still surpass Internet infrastructure spend by $50-$150 billion.
To understand the root of this buildout, we must understand the soil and materials, the players and incentives, the assumptions and beliefs.
The blossoming tree of AI technology, which optimists are declaring is our generation’s electricity, and pessimists are claiming is a nuclear bomb that could kill us all, began with a seed.
We will start with a deep analysis of this seed, the essential ingredient of AI compute: GPUs.