Episode summary

Why Neo-Clouds are Great Businesses & The State of Compute with Kush Bavaria, Co-Founder @ Ornn

20VC · · Hosts: Harry Stebbings · Guests: Kush Bavaria

Why Neo-Clouds are Great Businesses & The State of Compute with Kush Bavaria, Co-Founder @ Ornn – 20VC

Summary

Kush Bavaria argues that AI infrastructure is becoming a rent-collecting business, with data centers and GPU owners acting like landlords while model builders and inference providers are the tenants. He says compute is still undersupplied, lead times remain long, and chip prices have continued to rise because demand for tokens and training has stayed strong. In his view, the public market is still too skeptical about AI usefulness, while the physical layer of the stack is increasingly valuable.

The conversation centers on why Ornn exists: to build a market for compute through price indices, futures, and eventually broader financial products. Kush says compute needs standardization to trade like a commodity, similar to Brent oil, and that his company had to go deeper into chips and data centers than expected because the market data is not readily shared. He sees liquidity, credit analysis, and ratings as key missing pieces, and expects compute futures and data products to become meaningful businesses.

Kush is also bullish on specialization. He expects inference to split into many verticals, with horizontal providers alongside dozens of focused players, and he thinks older GPUs will keep finding use in cheaper or slower inference rather than becoming obsolete. On the company-building side, he prefers New York over Silicon Valley, values working with close friends, and frames Ornn as a long-term family business rather than a fast flip.

Key takeaways

  • Kush Bavaria says AI infrastructure investors can still win because data centers and GPU owners collect rent as long as demand persists.
  • He thinks GPU supply is very tight, with CoreWeave quotes stretching into Q4 2027 and early 2028 and OEM lead times around 20 to 22 weeks.
  • Kush believes chip prices are rising, not peaking, and says buyers now have to think in weeks and months ahead to secure capacity.
  • He argues compute needs a standardized index and futures market, like oil benchmarks, before it can trade efficiently.
  • He expects inference to become highly specialized, with a horizontal base provider plus many vertical niche providers.
  • He thinks older GPUs will remain useful for lower-cost and specialized inference, rather than becoming obsolete immediately.

Notable quotes

“I think infrastructure investors actually like end up winning because they will always keep collecting rent”

— Kush Bavaria at 1:06

“I think compute is a larger market than oil will be or any of the other commodities have been for the past like 100 years, 150 years”

— Kush Bavaria at 1:06

“the price per GPU hour is still increasing”

— Kush Bavaria at 17:37

“I think there will be like a horizontal provider that is like very good at like providing like base level inference”

— Kush Bavaria at 27:57

Topics

  • ai compute marketplace
  • gpu supply and pricing
  • hedging compute risk
  • training vs inference demand
  • ai investing and market overheating

Companies and tickers mentioned

  • Ornn
  • Andreessen Horowitz
  • Thrive Capital
  • BoxGroup
  • CoreWeave

Watch the episode

Watch “Why Neo-Clouds are Great Businesses & The State of Compute with Kush Bavaria, Co-Founder @ Ornn” on YouTube

Summary, takeaways and quote selection by Octagon. Quotes are short excerpts; the full conversation is in the episode.

More from 20VC

All 20VC episodes →

20VC, this episode and its artwork belong to Harry Stebbings. Octagon is not affiliated with the show; this page is an independent summary.