newsfilter
← Interviews
20VC with Harry Stebbings1h 13m

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

Kush Bavaria, Harry Stebbings

Interview, Fireside Chat, Podcast
  • Orn's Core Mission: Kush, co-founder and CEO of Orn, is building a compute market that creates price indices and futures for GPU compute to standardize pricing and risk management.
  • Market Valuation Thesis: The speaker argues compute will become a larger global commodity market than oil, potentially reaching a valuation in the trillions over the next decade.
  • Infrastructure Investment View: Infrastructure investors (landlords/data center owners) are expected to "always win" by collecting rent, similar to real estate, as long as there is demand for compute.
  • Chip Pricing Trends: GPU prices have been increasing daily for the last 60 days, with forward contracts for Q4 2027 and Q1 2028 already being quoted due to severe supply shortages.
  • Supply Chain Lead Times: Acquiring and installing new GPUs takes approximately 20–22 weeks just for the chips, plus 2–3 weeks for installation, leading to a total lead time of roughly 6 months to build on-premise infrastructure.
  • Neo-Cloud Moat Assessment: Neo-clouds like CoreWeave and Nebius are viewed as having no inherent moat in infrastructure alone; they must expand product lines to gain competitive advantage.
  • Investment Position: Despite acknowledging the lack of a moat in neo-clouds, the speaker expresses a strong preference to go long on CoreWeave and Nebius due to their ability to monetize infrastructure assets and secure tenant payments.
  • Bear Case Scenario: A downturn in enterprise demand or failure of AI to prove ROI to large companies (e.g., McDonald's, Coca-Cola) could lead to tenant defaults, leaving creditors with chips worth only as metal scraps.
  • Depreciation Dynamics: Older chips (A100s, H100s) retain value longer than predicted; they remain viable for three-plus years for inference tasks rather than becoming obsolete immediately.
  • Inference Market Fragmentation: The inference market will likely split into specialized verticals (e.g., fast inference, slow inference, biological inference) rather than being dominated by a single horizontal provider.
  • Investment Behavior: Late-cycle VC momentum is characterized as "FOMO investing," driven by fear of missing the next OpenAI or Anthropic, leading to rushed due diligence and 2-hour funding rounds.
  • Overheated Categories: Horizontal consumer and professional tools like customer support (e.g., "Rogo") and legal AI (e.g., "Harvey") are considered saturated and difficult to compete against.
  • Undersupplied Sectors: The speaker identifies the physical AI stack, specifically data center infrastructure components like transformers and electricity, as significantly underinvested.
  • NYC vs. Silicon Valley Culture: The speaker rejects the "coin-operated" transactional nature of Silicon Valley, preferring New York's culture of loyalty and long-term relationships over equity-driven recruitment.
  • Orn's Founding Context: The company was founded to solve the problem of finding and securing compute, stemming from the founder's realization that software-only approaches failed because data centers wouldn't share sensitive pricing data.
  • Index Standardization Strategy: To build a reliable index, Orn runs a physical exchange and controls the underlying assets to gather transaction data, bypassing the need for partners to share proprietary data.
  • Future Debt Ratings: A planned revenue line for the future is providing debt ratings for data center projects, acting as a "Moody's for data center debt" to help lenders assess credit risk.
  • Talent vs. Compute: The speaker posits that "more compute" is the primary determinant of winning in the AI race, as better models eventually make up for talent gaps, though a combination of both is ideal.
  • Longevity Projections: The speaker predicts human lifespans will extend to 150–200 years within a few years, with high quality of life, driven by advanced biotech and AI.
  • Human-AI Relationship: A future belief is that humans will act as "biological vessels" for AI, teaching AI biological feelings and emotions rather than controlling them.
  • Founding Team Composition: Orn hires primarily ex-college athletes to ensure high communication skills, competitiveness, and the ability to handle uncertainty without immediate gratification.
  • Founder Culture: The team lives in a six-floor walk-up apartment in NYC to maintain "grit" and a "hunger" for success, avoiding the complacency associated with luxury housing.
  • Naming Origin: The name "Orn" was selected because it was a cheap four-letter domain derived from a character in League of Legends, adhering to a strategy that shorter names have higher success rates.
  • Investor Selection: The company chose boutique investors (Thrive Capital) over multi-stage giants initially because the partners cared more about the founders as people, fostering a long-term "family business" dynamic.
  • Personal Philosophy: The speaker views dating as a distraction during the pre-IPO phase, joking that he will remain single until returning investor funds, while emphasizing the mental toll and emotional volatility of founding.
  • AI Tool Evolution: The speaker has shifted from skepticism to being "pro-harness," using tools like Cursor and Factory to switch models rapidly and find the best AI for specific coding tasks.
  • Personal Investment: The speaker would invest all personal capital into "Factory" (the coding harness company) due to its superior product and utility compared to competitors like Cognition.
  • Model Comparison: When forced to choose between OpenAI and Anthropic, the speaker would back OpenAI for its current compute availability, noting Anthropic's cash is useless if it cannot secure TSMC fabrication capacity.
  • 2030 Skepticism: The speaker admits the belief in humans becoming "biological vessels" for AI by 2030 is likely incorrect, viewing it as a long-term possibility rather than a near-term reality.
  • Ultimate Goal: The speaker views his company's role as enabling humanity to flourish by democratizing access to compute, which supports biotech and longevity research.