data center revenue · X 热门 · 2026-09-16 17:12 UTC

# data center revenue — X 热门讨论 (2026-09-16 17:12 UTC)

## @ZitoSalena (ZitoSalena) · 09-16 15:24 · ♥41 ↻14 💬3 The building of AI data centers is not landing in a neutral country. It is landing in one where the debate over artificial intelligence has turned political, cultural, and increasingly fearful. That opposition is not theoretical.

In Archbald, in Conshohocken, in Wilkes-Barre, in Hampden Township outside Harrisburg, Pennsylvanians have packed municipal meeting rooms to fight data-center proposals over noise, water, air quality, property values, and the fear that their power bills will subsidize somebody else’s chatbot.

And yet Shippingport, a borough of 154 people that has watched its tax base evaporate twice in one lifetime, is not Archbald. The mayor is on the podium.

The building trades are counting jobs. The county is counting revenue. When your town has already lost the nuclear plant and the coal plant, a two-gigawatt anything looks less like an imposition and more like a lifeline.

Both of those reactions are real. And people — actual complicated people and not the caricatures that the loudest voices on either side reduce them to — are perfectly capable of holding both views at once.

https://t.co/KtfPeqv0nU https://x.com/ZitoSalena/status/2100244556281151911

## @LoveOkon01 (LUV💚) · 09-16 13:07 · ♥33 ↻12 💬7 Everyone talks about GPU price. Very few talk about the cost of waiting.

That was the question I took to AethirClaw today:

What if “time to first GPU” became a real AI infrastructure metric?

When an AI company needs compute, the obvious questions are usually:

How much does the GPU cost? How much capacity is available? What are the contract terms?

But there is another question hiding underneath all of that:

When can I actually start using it?

Because a GPU that arrives months late isn't really available capacity today.

The delay can affect:

→ Model training timelines → Product launches → Engineering productivity → Revenue timing → Ability to respond to market opportunities

And this is where the previous AI infrastructure cycle becomes interesting.

GPU shortages showed that access to compute can become a bottleneck in itself.

But the problem doesn't necessarily end when GPUs are available.

You still need the infrastructure to house them.

Power. Grid connection. Cooling. Electrical systems. Data center capacity.

So the real race isn't simply:

Who can get GPUs?

It is increasingly:

Who can get usable compute online fastest?

I asked CARA to research the historical evidence, find examples from the previous GPU supply cycle, examine the hidden costs of waiting, and then challenge the thesis.

One of the strongest counterarguments was important too:

If GPU prices fall significantly over time, waiting could sometimes save money.

So “faster” isn't automatically “better.”

The economics depend on what the customer loses while waiting.

That distinction matters.

And this is where Aethir's ACCELERATE strategy becomes an interesting part of the conversation.

If powered shells and existing grid-connected capacity can move infrastructure deployment from a multi-year greenfield process toward months, then the value isn't just the infrastructure itself.

It's the time saved getting that infrastructure into production.

Maybe AI infrastructure procurement needs a new line on the spreadsheet:

Time to first GPU.

Because the cheapest GPU isn't necessarily the cheapest compute if you spend months waiting before you can use it.

I explored the thesis, counterarguments and supporting research with AethirClaw.

The next phase of AI infrastructure may be less about simply having compute...

and more about how quickly that compute becomes usable.

#AethirClawTribe https://x.com/LoveOkon01/status/2100209970243358870

## @nextbigfuture (nextbigfuture) · 09-16 14:02 · ♥35 ↻8 💬2 OpenAI has $40 billion of annualized revenue. This is a 20% increase from august. This is $3.5 billion per month. SpaceX has more annualized revenue. Last quarter spacex was at $2.5 billion per month average but had more in June. Now spacex has added rhe Google contract at $0.9 billion per month. In August spacex finished rhe acquisition of cursor adding $0.4 billion per month. Two more AI deals added 2.2 billion per month . New Ai deals are at $6 billion per month. The Ai revenue for Spacex is more than OpenAi. By December spacex should be at $120 billion ARR. $10 billion per month. $2-3 billion Per month will be the Starlink and launch. The rest will AI for Spacex. Spacex is and will be ahead of OpenAI on AI revenue. Spacex growing AI leasing with four times as much data center will be ahead of OpenAI and Anthropic combined in 2027. 8GW. $50 billion per year per gigawatt will be $400 billion per year. > 引用 @wallstengine: OpenAI’s annualized revenue surpassed $40B last month. That run rate jumped about 20% after the launch of GPT-5.6. - FT https://t.co/1Tsf3i3Am3 https://x.com/nextbigfuture/status/2100223782497788312