20VC: The State of Compute: What Everyone Needs to Know | Why Neo-Clouds are Great Businesses | What Everyone Is Missing About GPU Prices? | How AI Infrastructure Investors Will Win with Kush Bavaria, Co-Founder @ Ornn
20VC: The State of Compute: What Everyone Needs to Know | Why Neo-Clouds are Great Businesses | What Everyone Is Missing About GPU Prices? | How AI Infrastructure Investors Will Win with Kush Bavaria, Co-Founder @ Ornn
Podcast1 hr 12 min
Listen to Episode
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights
  • Consider CoreWeave (CRWV) and Nebius (NBIS) for high-risk exposure to constrained AI-compute capacity, but size positions cautiously: the thesis depends on sustained demand, reliable tenants, and GPU values holding up.
  • Look beyond GPU providers to data-center power and equipment suppliers, including transformers, as a potential AI-infrastructure theme; no specific public tickers or targets were provided.
  • Treat these as long-term, high-uncertainty themes rather than short-term trades: the discussion gave no price targets or entry points, and weaker AI demand could pressure compute prices and investment returns.
Detailed Analysis

CoreWeave (CRWV) and Nebius (NBIS)

  • Kush said he would go long CoreWeave and Nebius if he were able to invest publicly, arguing that AI-compute demand and GPU prices could keep rising.
  • He likened these “neo-cloud” providers to landlords: they own compute infrastructure and earn revenue by renting it to AI companies.
  • He said GPU supply is constrained: CoreWeave orders may be quoted for Q4 2027 or Q1 2028, while building and installing a smaller system could take roughly half a year.
  • He cited rising GPU-hour prices: around $7–$8 for a five-year arrangement today, and speculated that future Rubin GPUs could cost $20–$30 per GPU-hour. That is his forecast, not a stated market target.
  • He also argued that older GPUs may retain value longer than investors once expected: he said A100s and H100s were still being rented, including for inference workloads.

Risks and Takeaways

  • The bullish case depends on AI demand continuing and tenants being able to pay for compute. If enterprises do not find meaningful business value in AI, providers could face weaker demand and falling rental prices.
  • Kush highlighted leverage, tenant credit quality, and the uncertain resale value of GPUs as risks. If customers default, lenders may be left with chips worth less than expected.
  • He said neo-cloud providers may lack a strong moat, although owning and operating infrastructure can still support a rental business.
  • The discussion supports a high-risk, high-upside AI-infrastructure thesis, not a guarantee that either stock will perform well. Kush’s stated preference was to go long, but he also described meaningful demand, credit, and asset-value risks.

AI Compute and Data-Center Infrastructure

  • Kush argued that compute could become a major commodity market, with pricing shaped by supply and demand.
  • He sees investment opportunities in the physical infrastructure needed to expand data centers, including transformers and other specialized equipment.
  • He praised Crusoe’s approach of building and integrating electricity infrastructure, transformers, and related data-center components.
  • He said rising compute demand is drawing capital from large technology companies and AI firms, while smaller startups may struggle to secure capacity.

Takeaways

  • The discussion suggests looking beyond GPU makers and cloud providers to the electricity, equipment, and data-center supply chain that enables AI growth.
  • Kush’s bullish case rests on sustained demand and constrained supply. He also noted that large-scale AI adoption by enterprises remains uncertain, so infrastructure investment may carry significant demand and financing risk.
  • No specific public stocks, price targets, or investment timeline for transformer or data-center suppliers were recommended.

NVIDIA (NVDA) and GPU Pricing

  • NVIDIA GPUs were discussed as the standardized compute product rented by many neo-clouds.
  • Kush said prices for H100s and newer GPUs had risen as demand increased, and that some buyers had previously underestimated how long GPU prices might remain high.
  • He argued that older GPUs can remain useful for less demanding inference workloads even after newer chips arrive.

Takeaways

  • The transcript presents a positive view of GPU demand and pricing, but it does not offer a direct recommendation to buy NVIDIA stock.
  • GPU prices and chip availability are important variables for the wider AI-infrastructure investment case. A downturn in AI demand could reverse the current pricing strength.

Meta (META) and Hyperscaler Capital Spending

  • In response to a question about Meta’s bonds, Kush said they had become more expensive amid market concerns about the amount of spending going into AI infrastructure.
  • He noted that hyperscalers are directing substantial cash from their established businesses toward data-center capacity, and that investors are concerned about the shift from positive to negative free cash flow at companies such as Google and Facebook.

Takeaways

  • The discussion points to AI capital spending and its effect on cash flow as an important issue for large technology companies.
  • It did not include a clear buy or sell recommendation on Meta stock or its bonds. The stated concern was investor uncertainty about the scale of spending and the eventual return from AI demand.

Shopify (SHOP)

  • Shopify was cited as an example of market volatility: its shares reportedly fell 15% after what Kush described as a small earnings miss.
  • He used the example to illustrate how difficult it is for investors to price companies when their five-year outlooks are uncertain.

Takeaways

  • The mention was about market volatility, not a Shopify-specific investment thesis.
  • The transcript gave no price target or recommendation for Shopify.

OpenAI and Anthropic

  • Asked which company he would prefer to invest in, Kush chose OpenAI, saying it had access to more compute.
  • He argued that access to compute could support faster model development and product growth, while noting that chip manufacturing capacity cannot be expanded instantly.
  • He also acknowledged that the relative quality of the companies’ models can change quickly.

Takeaways

  • The discussion emphasizes compute availability as a competitive factor for AI companies.
  • Both companies are private, and no actionable public-stock recommendation or price target was given. The view is a comparative opinion, not a complete assessment of either company’s valuation or prospects.

Factory

  • Kush said he would put all of his personal investment money into Factory, the coding-tools company, because he prefers its product and believes users might switch to it from competing tools.
  • He said the product works well for his team, including because it can use different AI models and offer a better experience.
  • He also qualified his view by saying he did not know whether the underlying business was strong.

Takeaways

  • This was an enthusiastic, product-led opinion about a private company, not a researched valuation or a balanced assessment of its business.
  • The “all my money” comment describes a highly concentrated hypothetical bet. The speaker’s own uncertainty about the fundamentals is an important caveat.

Ornn (Private Company) and Compute Markets

  • Ornn is described as building price indices and futures for GPU compute, with the goal of creating reference prices and financial products for the market.
  • Kush said the company is also developing a physical market to obtain transaction data, since providers may be unwilling to share commercially sensitive pricing information.
  • He compared the opportunity to established market and data businesses such as ICE, CME, and S&P, arguing that compute could become a much larger commodity market.
  • He also identified data-center debt ratings as a potential future business line.

Takeaways

  • The potential investment theme is the market infrastructure around compute: pricing data, indices, trading, and credit ratings.
  • Kush’s case depends on compute becoming standardized and widely traded. He noted that different GPU configurations and customer requirements make standardization challenging.
  • The discussion describes Ornn’s business opportunity but does not provide a public investment recommendation or valuation.

TSMC (TSM) and Chip-Supply Constraints

  • Kush referred to TSMC’s fabrication capacity when explaining that demand for chips cannot simply make manufacturing supply expand faster.
  • He used this as part of the argument that compute availability could constrain AI-company growth.

Takeaways

  • Chip-manufacturing capacity is a key dependency in the AI-compute thesis.
  • The transcript did not provide a direct recommendation or valuation view on TSMC.

Broader AI Investment Themes

  • Kush was skeptical of some fast-moving AI fundraising, describing it as driven by FOMO and fear of missing the next major AI company.
  • He said established AI categories such as legal and customer-support tools may be difficult for new entrants to challenge.
  • He expects some inference businesses to specialize by workload—for example, speed, cost, image or video generation, or biological applications—rather than every provider serving every use case.
  • He warned that some AI startups and neo-cloud tenants will fail, and that providers could be exposed if they lend compute without adequately assessing customers’ ability to pay.

Takeaways

  • The discussion favors careful attention to compute access, real customer demand, and customer credit quality over investing solely on AI enthusiasm.
  • Potential opportunity areas mentioned include specialized inference, data-center infrastructure, and compute-market services. The transcript does not give specific public-company recommendations for these themes.
Ask about this postAnswers are grounded in this post's content.
Episode Description
Kush Bavaria is the Co-founder and CEO of Ornn, building the financial market for AI compute. Ornn helps companies price, finance and hedge GPU capacity—bringing the tools of commodity markets to AI infrastructure. Kush is just 23 years old and has raised funding from some of the best in the world including Andreessen Horowitz, Thrive Capital and BoxGroup. AGENDA:  05:00 Building a Viral Sports Betting Platform from a Dormroom at MIT 09:00 Can AI Win While Infrastructure Investors Lose? 13:00 What Is Everyone Missing About GPU Prices? 14:00 Why Is Everyone Betting Against CoreWeave and Nebius? 19:00 Are Older GPUs Really Becoming Obsolete? 24:00 Where Are AI's Biggest Untapped Opportunities? 29:00 What Would It Take for Compute to Trade Like Oil? 43:00 What Wins the AI Race: Better Talent or More Compute? 48:00 Why Raise Money When You Don't Need It? 59:00 How Do You Build a Team That Hates Losing? 01:05:00 Would You Buy OpenAI or Anthropic?
About The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

By Harry Stebbings

The Twenty Minute VC (20VC) interviews the world's greatest venture capitalists with prior guests including Sequoia's Doug Leone and Benchmark's Bill Gurley. Once per week, 20VC Host, Harry Stebbings is also joined by one of the great founders of our time with prior founder episodes from Spotify's Daniel Ek, Linkedin's Reid Hoffman, and Snowflake's Frank Slootman. If you would like to see more of The Twenty Minute VC (20VC), head to www.20vc.com for more information on the podcast, show notes, resources and more.