The AI build-out is estimated to cost $10.3tn.
That’s roughly 3.6% of US GDP, every year to 2032.
Everyone is funding it. That doesn’t mean everyone gets the compute.
More concrete doesn’t mean more access. Making existing supply accessible does.
That’s what io.net was built for.
Lock-in isn’t on your invoice.
But it’s built into most GPU cloud contracts.
It’s the 40 TB egress bill, the multi-year terms with penalty clauses, the single provider with no fallback.
At 50 H100 hours a day: $44,820/yr locked in at typical neocloud rates vs. $26,820 on-demand.
The lock-in tax is real. And so is the exit.
io.net changes the architecture, so you don’t pay it.
23,000 GPUs.
One flagship data centre.
And Microsoft just pulled out because the grid can’t deliver the power until the 2030s.
Money can buy chips and customers, but it can’t speed up a grid connection.
That’s why distributed compute is so important. That’s what @ionet was built for.
Unroasted coffee waits in a warehouse for the next buyer.
An unused GPU hour doesn’t.
That’s why 90% vs 85% utilisation matters more than most pricing debates.
The @hyperbolic_labs paper runs analysis around 60,000 price paths. The assumption I would spend more time on is that 90% of capacity finds a buyer.
This is where coffee feels like a useful comparison, I have been on the receiving end of coffee bean inflation ☕️. A benchmark does not make every coffee bean origin and grade equally saleable. In compute, networking, location and reliability still determine which workloads a cluster can serve.
The year three numbers in their analysis show why that matters. At the modelled selling price of roughly $4.90 per billed hour, 90% utilisation earns about $4.41 for every available hour. The full contracted capacity cost is $4.30, including the prepaid deposit. At 85% utilisation, revenue becomes about $4.17 per available hour. That falls below the full capacity cost before paying for the hedge.
The selling price has not changed. Just five percentage points of utilisation.
My addition would be to let utilisation and customer mix move alongside price. Include a scenario where rates fall and fewer hours sell. Feed those assumptions with actual bookings, repeat usage and reasons customers leave. This can also explain the recent @CoreWeave debt news flow.
Building inference products at @ionet has made me pay attention to how demand changes around model launches and new AI tech trends. The workload customers wanted last month may need different capacity or a different service today.
That is product work with a direct effect on the economics: which workloads to support, how to package, extract peak performance from the capacity and how much demand actually returns.
Coffee traders care about the specific micro lots and its buyers. Compute needs that same attention, with one extra constraint.
Unroasted coffee beans can wait in a warehouse for another buyer. An unused GPU hour cannot.
Thank you @hyperbolic_labs team on putting this paper out there.
The GPU shortage isn’t really a shortage.
It’s a queue.
And the biggest buyers already skipped it.
Everyone else gets the leftovers, a markup and a long lease.
Good thing there’s a shorter line. Compute that’s already built, already running, and priced for the people who actually build.
io.net. AI for the many, not the few.
io.net/blog/the-ai-oligopoly…
A few companies control the most powerful AI.
Now they're debating whether it has feelings.
But this is just a way to shift accountability onto the models.
“It’s not us, it’s the AI.”
That’s what concentrated control looks like. A few firms build it, run it, and decide who answers for it.
Open access means more people can build, inspect, and hold it accountable.
That’s why we built io.net.
Gartner expects over 80% of neoclouds to cease independent operations by 2030.
The AI data centre build-out needs $8.5T in revenue to cover costs and returns.
Gartner predicts $3.6T at most.
When the math doesn’t work, the builders relying on that infrastructure are the ones left on shaky ground.
io.net is different.
We orchestrate supply that already exists. We don’t bet on demand that hasn’t arrived.
The big AI question isn’t which model you can use.
It’s whether you can get the compute to build at all.
Trillions of dollars of data centers are in the pipeline. But they’re built mostly for buyers who sign 3-year commitments.
If you’re a researcher who needs 8 GPUs for three weeks, a startup validating a product, or a company without massive budgets, you’re locked out.
More capacity under the same access model just means the same gap, only bigger.
io.net is here to fix the access model, not just the supply.
AI is pushing prices up.
Everyone’s paying for the build-out, from power to electronics.
Meanwhile, access to the compute itself stays locked behind long, expensive contracts.
A few tech giants collect the upside while the rest of us get the invoice.
You don’t have to use AI to pay for it. But you need a big budget to build with it.
That’s what io.net is working to change.
Europe doesn’t just have a chip problem.
It also has a power problem.
@ionet's David Watts is in Vienna today at Data Center World Europe making the case for right-sized AI infrastructure.
Send the work to where the energy is and let software do the routing.
Catch him at 2:50pm in Schubert 6.
Microsoft’s president says "AI is less regulated than milk."
He’s right.
But rules written by the companies building the problem will never fix it. They just protect the people at the top.
Safe AI comes from more access, more participation, and more voices at the table.
That’s why @ionet is building AI for the many, not the few.
