$11,000,000,000,000
$11,000,000,000,000.
That's the number to be kept front of mind.
That's the number that makes all the other numbers make sense.
But it's a hard number. Because it's so impossibly big.
We are not used to thinking in trillions. Tens of trillions? Fuggetaboutit.
Yet this is a very real number. A very important number.
Possibly the defining number of this moment.
Let's put this very big number into context.
Eleven trillion dollars is:
- More than one third of all the gold mined in all of history
- More than one third of US GDP
- Six times the annual GDP of Australia
All of Australia, working as hard as we can, for more than half a decade, only just touches $11,000,000,000,000.
Why is this number important?
It's the amount that's being spent on data centre infrastructure between 2024 and the end of 2029.
Around two trillion of that eleven trillion has already been spent.
Nine trillion will be spent over the next three and a half years.
That's the plan, at least.
These kinds of numbers, when they reflect spending, produce all sorts of unexpected consequences.
In general, it's simply not possible to spend that much money so quickly toward a single goal.
Even when you can throw effectively unlimited money at that goal.
The world is complex and irregular and does not always yield to money.
That $11,000,000,000,000 means - and what we're already seeing - a massive twist in economic space-time. Eleven trillion dollars gathered together toward a single goal form a sort of economic singularity: a black hole of capital, that bends everything into its orbit.
This event horizon has become the particular object of fascination at Noops Research.
No one has ever seen capital deployed so quickly at such scale. The prior examples: railroads, electrification, and automobiles all took decades, while this is advancing in months.
Some systems can move that quickly. Many can not. Some will break under the strain.
All of that is already becoming visible:
- RAMageddon making cheap smartphones and computers impossible
- Data centre electricity use doubling from 485 terawatt hours to ~1000 twh
- Global interest rates spiking due in part to increased demand for capital
Finding $11,000,000,000,000 in capital isn't easy. SemiAnalysis projects $7,000,000,000,000 in debt will need to be raised - on roughly the same cadence that the US finances its current $2T-a-year deficit.
Lest anyone call this out as 'circular financing', it should be stated clearly and for the record: Data centres print money from the moment they're operational, and do so long past they've paid themselves off, 24 months later.
Demand for compute is both insatiable and increasing exponentially. This $11,000,000,000,000 is being invested to satisfy demand, not - as was the case with WorldCom's dark fibre - in advance of that demand.
So this economic singularity isn't going anywhere, even if the frontier labs all go dark. The demand for compute to deliver AI to new customers is already enough to justify the spend.
But that spend, as noted earlier, is going to be uneven. That unevenness creates incredible opportunities - and specific cautions.
Which is the point of AI Investor, Noops Research's half-day event designed to provide clarity and tools for thinking about the biggest capital deployment in history.
So with that $11,000,000,000,000 planted firmly in mind, we're going to 'walk the stack' of all of the areas impacted by this spend, from 'sand to software'.
This is a far bigger story than data centres.
Over the next six posts, we'll explore them all.