brain/
← all entities
entitypersonstock-market

Anjney Midha

Founder & CEO of AMP PBC (compute grid / independent system operator) · First angel check into Anthropic; former General Partner at a16z · Stanford visiting scientist (CS153 Frontier Systems)

aka Anj

Quotes

Today the average data center in the industry, in the independent ecosystem, is running at less than 70% utilization. The Colossus 2, which is running in Memphis, Elon's 500,000 GB300s, was running at less than 60% node utilization and less than 11% MFU model flop utilization... At Google the utilization is roughly 99%.

2026-06-13-podcast-odd-lots-anjney-midha-s-plan-to-radically-lower-the-price· 2026-06-13#compute-utilization-overhang-as-latent-supply

We take all of that unutilized compute, no matter what format it is... and we turn it into one fungible resource... today we're able to do that in ways that improve utilization, sometimes from 50, 60% at labs we have incubated to close to 95, 96%.

2026-06-13-podcast-odd-lots-anjney-midha-s-plan-to-radically-lower-the-price· 2026-06-13#compute-utilization-overhang-as-latent-supply

The effective price per hour that you're paying is closer to 25 to $28, whereas the marketed rate that you think you're paying is $2.50. So that spread due to wastage is just insane. That's the deadweight loss.

2026-06-13-podcast-odd-lots-anjney-midha-s-plan-to-radically-lower-the-price· 2026-06-13#compute-utilization-overhang-as-latent-supply

About 80 cents of every dollar a lab spends today on their R and D flows to a chip provider like Nvidia... so as a result your margins are just super rough. So from a unit economic perspective you want more control over your margins... and you want supply chain independence. That's how it works at the foundry level today: TSMC gets to decide which compute provider's business grows or not because they only have so much production capacity.

2026-06-13-podcast-odd-lots-anjney-midha-s-plan-to-radically-lower-the-price· 2026-06-13#cuda-moat-erosion-at-inference#compute-utilization-overhang-as-latent-supply

If you look at the compute prices of long term rentals over the last six months, from between January and now, they're trading up 2x. We started securing our capacity in January at these long term rates. We could resell that at a 2x markup if we wanted to. [Demand slope is] perpendicular.

We're actually procuring capacity in advance through call options on compute clusters... I'm not that thrilled about the financialization of these products. When you add speculation to production goods, it creates scarcity of a different kind, because then you have financial traders trying to trade the speculative value of the asset.

I just don't think there's only three frontier models that are going to win. It's going to be an ecosystem... corporate leaders don't care about the underlying model. They just want the work done in the cheapest, most efficient way. Increasingly which particular model is helping you out will just be abstracted... under the hood they might be using 200 different models to orchestrate your task.

Notes

Anjney Midha

One-line summary: AI-infrastructure operator/investor building a software 'grid' to make heterogeneous GPU compute fungible; tracked for compute-utilization economics and the latent-supply contradiction to GPU scarcity.

What they're known for

Brief factual context — fill in.

Why they matter to stock-market

Why this person's claims are tracked here — fill in.

Said

Speaker-attributed claims extracted from diarized sources. Each bullet mirrors one entry in quotes: frontmatter — keep them in sync.

Sources

Related

Cross-links — fill in.

Referenced by