Dear Reader,
Good morning,
This is Dylan Jovine with Behind the Markets.
Happy Wednesday.
Today's Wednesday, February 18th.
What a great day to be alive.
Today I want to talk about a new report from Bloomberg that really puts something we've been discussing for years into stark perspective — the coming power needs of AI.
You've heard me say dozens of times that AI is, at its core, an energy story.
But these new numbers?
They're almost hard to wrap your head around.
Global data center power demand is expected to rise to nearly 1,600 terawatt hours by 2035.
Let's translate that into English.
A watt hour is simply one watt of power used for one hour.
If you turn on a 100-watt light bulb and leave it on for an hour, that's 100 watt hours.
A terawatt hour?
That's one trillion watt hours.
And we're talking about 1,596 terawatt hours.
That's 1,596 trillion watt hours of electricity.
It's almost absurd.
In the U.S. alone, demand is expected to reach roughly 430 terawatt hours.
China is projected to add around 397 terawatt hours.
Europe's demand is expected to surge more than 300%.
Between now and 2030 alone, we'll need over 600 additional terawatt hours — enough to power roughly 60 million homes.
That's not incremental growth.
That's an energy shockwave.
And it's going to change everything.
That's why TOMORROW, at 1:00pm est., I'm going live to tell you everything you need to know about this AI power crisis, and which companies are positioned best to profit from it.
Make sure to click here to reserve your spot asap.
Here's the key point:
AI growth is contingent on generating more power.
Without energy, AI doesn't scale. It doesn't jump off the screen into the physical world. It doesn't automate factories. It doesn't train next-generation models. It doesn't power robotics.
It stalls.
If you want to see how real this bottleneck is, go Google the 48-megawatt data center in Santa Clara, California.
It's a beautiful facility.
State-of-the-art.
And it's been sitting largely empty for years.
Why?
No power.
That's in NVIDIA's backyard.
Google has even warned it could take up to 12 years to get some new data centers connected to the grid.
Twelve years.
Think about that.
We're in an AI arms race with China — and we're telling companies they can plug in… sometime in the 2030s.
China doesn't have that problem.
For all the long-term flaws of centralized systems — and there are many — when their government says "build," it gets built.
Here, we negotiate with local governments.
We debate rate increases.
Residents push back.
Electric bills become political.
And they should — nobody wants their monthly bill to double because a hyperscaler needs more juice.
So what's happening?
A major shift.
The administration is now allowing — and in some cases encouraging — companies to bypass the traditional grid and build their own dedicated power infrastructure.
Private power plants.
Private generation.
Private energy ecosystems.
That changes everything.
It changes how AI scales.
It changes who wins the energy race.
And it creates a whole new set of opportunities — and risks.
We've been talking about AI and energy since 2022, when it was still mostly theoretical.
Now it's operational.
The story has moved.
And for the first time, the government is effectively saying:
"If the grid can't move fast enough, build around it."
That's a profound shift.
We're going to be breaking this down in detail — the bottlenecks, the policy changes, the national security angle, and the investment implications — on tomorrow's webinar at 1:00 PM.
So if you haven't reserved your spot yet, click here to sign up now.
If you care about where AI goes next, you need to understand where the power is coming from.
Because no power = no AI.
Anyway, that's what's on my mind today.
Have a wonderful day.
I'll see you tomorrow.
Tidak ada komentar:
Posting Komentar