Matt Atkinson

Building the land + energy + AI compute platform for the next decade. CEO @wearekizer. American Southwest.

[Matt Atkinson @mattkinson7]

Today, Kizer AI (@wearekizer) is launching a 500,000-acre turnkey land platform for AI data centers.

We’re working with one of the largest private landowners in the US and have assembled sites across Texas, New Mexico and Oklahoma that can support multiple 1.0 GW+ AI campuses, with access to the power, gas, water, fiber and transmission infrastructure needed to build them.

The goal is simple: give developers access to sites that are ready for AI infrastructure at scale.

If you’re a data center developer, energy company, infrastructure provider or capital partner looking for your next site, reach out to us.

Read the full announcement: https://t.co/s02Yd386UZ

[Matt Atkinson @mattkinson7]

Remember this graph every time you think the datacenter space is in trouble

[Warren Pies @WarrenPies]

Instructive to see a post like this get so much traction.

B200 rental rates up 30% YTD.

Hyperscaler cloud revenues tracking along.

Remember: GPU rental rates and token prices were supposed to fall over time. The demand metrics this year have beaten all informed expectations.

[Warren Pies @WarrenPies]

Instructive to see a post like this get so much traction.

B200 rental rates up 30% YTD.

Hyperscaler cloud revenues tracking along.

Remember: GPU rental rates and token prices were supposed to fall over time. The demand metrics this year have beaten all informed expectations.

[Matt Atkinson @mattkinson7]

@negligible_cap SpaceX and XAI alone are targeting 10 GW by the end of 2027, $BABA's 20 GW over 8 years pales in comparison. You always have to put the narrative in perspective

[Matt Atkinson @mattkinson7]

@edtarnowski The grid reality caught up

[Matt Atkinson @mattkinson7]

@ShanuMathew93 The total pipeline is even more impressive - @SemiAnalysis_ puts total lifetime AI capex at about $11T by the late 2020s with outstanding debt passing $7T. This is the gold rush, ladies and gentlemen.

[Matt Atkinson @mattkinson7]

You always need to watch the underlying ROI on these deals - too many bad deals went through during the hype cycle/gold rush and now when the ROI isn't holding up, the banks are pulling out from similar deals. Demand for data centers still exists though - funding supply from the banks exists as well. The only difference is that now major projects will get access to it.

[kabeer ahmed @KKabeer2004]

The 23rd largest private landowner in the US just opened up 500,000 acres of land for AI data centers.

Today, Kizer AI (@wearekizer) announced a platform for developing multiple 1GW+ AI data center campuses across the US.

I’ve been a part of this project since the very beginning and it's amazing to finally see it go public.

Full announcement:
https://t.co/RvE0vECa1D

[Matt Atkinson @mattkinson7]

@LambdaAPI @nvidia 23% efficiency gains is pretty sweet! Hardware and software optimisations need to go hand in hand to make sure we can keep up with the demand that is coming - AI system power consumption is on track to rise 6x over the next five years!

[Matt Atkinson @mattkinson7]

@SemiAnalysis_ Nvidia's backtops are super interesting because they allow them to do more than just sell GPUs - they're helping create the creditworthiness needed to finance the infrastructure that buys those GPUs

[Matt Atkinson @mattkinson7]

@MikeZaccardi Sure they can turn the spigot back on - the real question is can they afford to?

If AI capacity remains constrained, cutting capex doesn't just improve free cash flow. It means giving up compute capacity that competitors are still building.

[Matt Atkinson @mattkinson7]

AI is just as much a construction boom as it is a compute boom

More GPUs eventually means more campuses, substations, cooling, transmission and generation. At this scale, the limiting factor isn't just how many chips Nvidia can produce. It's how quickly America can build the infrastructure around them.

[Matt Atkinson @mattkinson7]

@SiegelScribe Regulatory requirements are necessary for a society to function safely but when regulation starts to trip up progress, we should carefully reconsider and re-balance

[Matt Atkinson @mattkinson7]

@curious_founder Wonder who that tenant is - and the conviction required for them to do all that work without even having a tenant is mind-boggling. They took a huge bet and its might just pay off very well.

[Matt Atkinson @mattkinson7]

@IEA Negative electricity prices aren't just a reason to build more batteries.

They're a signal that we have electricity at the wrong times and demand that isn't flexible enough to use it.

[Matt Atkinson @mattkinson7]

I think the contract duration is more interesting than the headline pricing.

$20M+ per MW on a 3-year deal means customers are paying a massive premium for near-term capacity without locking themselves in for a decade.

That looks less like normal data center leasing and more like a market trying to price infrastructure that can't be built fast enough.

[Shay Boloor @StockSavvyShay]

IREN went from $9.7M per MW with Microsoft and $11.3M with Nvidia on 5 year terms to $20M+ on recent 3 year deals showing how aggressively shorter duration AI capacity is repricing.

Annualized contract value across the AI infra stack
• $IREN ~$2.6B across $MSFT + $NVDA
• $NBIS ~$5.9B across $MSFT + $META
• $CRWV ~$7.8B across $META + OpenAI
• $AMZN >$10B with Anthropic
• $SPCX ~$15B with Anthropic

That gap between pricing and scale is what makes IREN so interesting since its already getting some of the best pricing per MW while carrying one of the smallest contract books leaving a lot more room to pair that pricing with scale even if corporate governance issues remain a real caveat.

[Matt Atkinson @mattkinson7]

This is being framed as a warning sign, but "returns take years to realize" is a strange criticism of infrastructure.

Power plants, transmission lines and data centers are supposed to generate returns over decades. The bigger question is whether the underlying demand will still be there when the assets come online.

So far, the problem doesn't appear to be finding demand. It's building enough infrastructure to serve it.

[Matt Atkinson @mattkinson7]

@MichaelDell @nvidia @CoreWeave @DellTech Big milestone!

Now the power infra needs to catch up to give these things the juice they need 🚀

[Matt Atkinson @mattkinson7]

@knowledge_vital I'd start with a different question: where are they going to get 10 GW?

At this scale, the limiting factor isn't whether Nvidia can put GPUs on a balance sheet. It's whether SpaceX can turn 10 GW of ambition into energized infrastructure by the end of 2027.