Kurt Pfluger

AI Compute; Private, Secure, Compliant. Hardware and Infrastructure. DM me for Inquiries

[Kurt Pfluger @DeepLeafInfra]

https://t.co/52YYMFdeUz

[Kurt Pfluger @DeepLeafInfra]

https://t.co/0uJklMrPon

[Kurt Pfluger @DeepLeafInfra]

Frontier training compute has doubled about every 5 months since 2020, roughly 5 times a year. The installed base of AI compute doubles about every 7 months.

Data center power demand was 123 GW in 2025, 161 GW in 2026, and 211 GW in 2027. TrendForce puts 2030 demand at 491 GW against 223 GW the grid can spare. That is a 268 GW hole.

The chips scale by the season. The substations do not.

[Kurt Pfluger @DeepLeafInfra]

https://t.co/9yAAN4cirx

[Kurt Pfluger @DeepLeafInfra]

@JadeYi25 @Shoulders9i I may be able to help. DM your requirements

[Kurt Pfluger @DeepLeafInfra]

@mrm8488 I may be able to help. DM your requirements

[Kurt Pfluger @DeepLeafInfra]

https://t.co/2CrFwKAo9v

[Kurt Pfluger @DeepLeafInfra]

https://t.co/VMgd5INEmN

[Kurt Pfluger @DeepLeafInfra]

We are actively working with teams that need private GPU capacity or hardware this quarter.

Isolated H100 / H200 / B200 (and AMD) capacity
US data residency options. Verified hardware with more realistic lead times than the open market. DM me if this applies to you.

[Kurt Pfluger @DeepLeafInfra]

https://t.co/h6HH0swcsB

[Kurt Pfluger @DeepLeafInfra]

https://t.co/sh8qE38vkz

[Kurt Pfluger @DeepLeafInfra]

https://t.co/XNoy6Xicyx

[Kurt Pfluger @DeepLeafInfra]

AI GPU pricing still shows wide spreads in September 2026.
On-demand rates for the same generation of hardware can differ significantly depending on the provider. Neo cloud and specialist platforms often sit well below hyperscaler list prices, while committed or reserved capacity can further reduce the effective cost.
For teams running sustained training or inference workloads, the difference between the highest and lowest available rates can meaningfully change project economics.
We help organizations evaluate private dedicated capacity and hardware sourcing as alternatives to pure on demand public cloud.
If you are reviewing current GPU pricing or capacity options, feel free to message me.

[Kurt Pfluger @DeepLeafInfra]

Quick week in review on AI compute - Sept 2026:
– Lead times still brutal for most buyers
– Pricing spreads between neo clouds and hyperscalers remain huge
– Power and transformers continue to constrain new builds
– Private / dedicated capacity is solving the problem for a growing number of teams

We run private GPU infrastructure single or multi tenant and help source hardware.

If you’re dealing with any of the above, reply “private” or DM me. I’ll respond personally.

[Kurt Pfluger @DeepLeafInfra]

36–52 week lead times for new B200s.
1% vacancy in primary data center markets.
Power gear still the long pole.

If your data needs isolation or you need capacity this quarter, the public cloud isn’t the only option.

Private dedicated GPUs available. Hardware sourcing available.
DM if useful.

[Kurt Pfluger @DeepLeafInfra]

@carmenli @computeexchange Looks great !

[Kurt Pfluger @DeepLeafInfra]

@MelvinInvests Power / Compute the new oil

[Kurt Pfluger @DeepLeafInfra]

@sonnet_xu @fluidstack Agreed

[Kurt Pfluger @DeepLeafInfra]

@liquidcompute @LHVGarcia @WSJ Congratulations !