“I just told a Fortune 500 CEO his data centres are worthless.”

That was how Guy Massey of CommScope opened his LinkedIn post, and I bet if you were there, you could almost hear a collective gasp across the server room. The CEO laughed at first, as any confident exec sitting on a billion dollars of data center assets might. Then Massey showed him NVIDIA’s power specs. The laughter stopped.

It’s the kind of moment that cuts through the industry noise with brutal clarity. GB200 NVL72 racks pulling 120kW each. Most existing data centers were designed for 10kW per rack. Massey summed it up perfectly:

“That’s like trying to power a jet engine with a car battery.”

And he’s not exaggerating. Welcome to the AI era, where physics, thermodynamics and raw grid capacity have become the new Moore’s Law. The data centers we built five years ago? Obsolete, Fini, Toast, Garbage. The ones being designed today? Already behind.

The Brutal Reality

Let’s start with the math.

In just five years, we’ve gone from 10kW per rack to over 100kW. Air cooling? That’s now like using a desk fan on a blast furnace. One AI cluster can draw more power than 50,000 homes. And the humble substation down the road — the one feeding your “state-of-the-art” Tier 3 facility? How quaint. That’s now the bottleneck.

Big Tech knows this.

• Meta is mandating liquid cooling, pushing 700W+ per GPU
• Google has its servers literally swimming in dielectric fluid.
• Microsoft has stopped retrofitting entirely — it’s now “liquid-first.”
• And NVIDIA is shipping 120kW racks right now.

So yes, the data centers we’ve been so proud of are today’s digital relics. They’re the IT equivalent of a Blockbuster card — nostalgic, maybe, but useless when the world has moved to streaming.

The Bigger Picture

If every hyperscaler is suddenly demanding ten times the power, what happens to the rest of us?

Where are all these AI-ready data centers going to live? The answer may surprise you: Everywhere and nowhere.

Some are going under the sea, where seawater offers free cooling and land is cheap — if you don’t mind corrosion and marine life. Others are popping up above the Arctic Circle, or being planned for space-based data centers orbiting Earth. I wouldn’t rule out Antarctica next. At least there, you don’t need to worry about air conditioning — just penguins wandering through your fiber lines.

All of this sounds wild until you realize that global AI compute demand is outpacing the world’s ability to build physical infrastructure. Desperate times, desperate measures, and all that.

So the multi-trillion-dollar question isn’t how we build enough new data centers — it’s whether we can.

Because if you’re counting on the power grid to bail you out, good luck. Grid expansion takes years. AI’s timeline runs in months.

Meanwhile, there’s another factor nobody wants to talk about: water. Even liquid-cooled systems need pristine, filtered water loops. Algae in your system? That’s a $10 million outage. Clean water may soon be as critical as clean code. Even dielectric fluid needs water.

Shimmy’s Perspective: Innovation Will Find a Way

If you’ve read my stuff before — or caught last week’s Shimmy Says — you know where this is going. We can’t spend our way out of this mess. We can’t build our way out. We have to out-innovate our way out.

The race isn’t about who builds the most massive, liquid-bathed, megawatt-hungry temple to AI. It’s about who figures out how to do more with less.

If Mohamed can’t come to the mountain, the mountain must come to Mohamed.

If NVIDIA’s chips need a Rolls-Royce data center, maybe the real innovation lies in building a Toyota that can do 80% of the job at 20% of the cost and footprint.

That’s where I’d be placing my bets — on the bridge builders. The companies are finding ways to make AI feasible for the rest of us. They’re the ones designing new cooling materials, smarter power management, or software layers that dynamically optimize energy draw.

Because if we don’t find that bridge, we’re stuck. And even $3 trillion in new data center spending might not be enough to cross the chasm.

The Real Achilles’ Heel

That’s why I think NVIDIA’s $5 trillion valuation might be built on shaky ground. Not because its products aren’t brilliant — they are. But because the infrastructure needed to run them at scale may be unsustainable.

When the limiting factor isn’t GPU supply but power density and thermal management, the playing field changes. Suddenly, the next big disruption isn’t in silicon — it’s in physics, fluid dynamics and materials science.

In every tech cycle, there’s a moment when scaling hits a wall. The companies that survive aren’t always the ones with the most powerful products; they’re the ones that find clever ways around the wall.

The bridge builders — the ones who deliver enough capability, in smarter, lighter and more efficient ways — may end up inheriting the earth.

The Bottom Line

Guy Massey wasn’t exaggerating when he said most data centers are worthless. They’re not dead yet — but they’re definitely terminal. The jet engine is coming, and we’re still trying to hook it up to a car battery.

So yes, $3 trillion in new data center spending doesn’t sound so crazy when you realize the old infrastructure can’t power the new world. But before we pour concrete and wire up another gigawatt farm, let’s ask a harder question:

Do we really want to build a future that only the top five hyperscalers can afford to sustain?

Because if that’s the plan, innovation won’t survive the heat — literally.

We need efficiency breakthroughs, new architectures, maybe even new physics. We need to science the sh^t out of this problem.

And when we do, when we find a way to make cutting-edge AI work in Toyota-class data centers — that’s when the next generation of innovators will take the lead.

The bridge builders. They’re out there. And as I said, they may just inherit the earth.

Shimmy’s Take

As Guy Massey said, “You can’t air-cool a volcano.”

You can’t run AI on yesterday’s data centers.

But you also can’t keep building jet engines when the world’s running out of fuel.

Somewhere between brute force and brilliance lies the real innovation.

That’s the road we need to find — before the grid goes dark.