Rosa Del Mar

Daily Brief

Issue 57 2026-02-26

Why Companies Stay Private Longer And When Ipos Still Matter

Issue 57 Edition 2026-02-26 9 min read
General
Sources: 1 • Confidence: Medium • Updated: 2026-03-02 13:01

Key takeaways

  • In David George’s public-market coverage universe, only three companies are growing revenue at over 30%.
  • Non-model-owning AI companies can remain defensible by compounding industry-specific context (workflows and data) and providing an accountable vendor relationship including support, integrations, and partnerships.
  • Founders generally dislike SPVs because they want transparency and control over who appears on the cap table.
  • AI infrastructure buildout may require on the order of $5 trillion over the next five to seven years.
  • Highly valued private tech companies represent about $5 trillion in aggregate market capitalization, roughly a quarter of the S&P 500.

Sections

Why Companies Stay Private Longer And When Ipos Still Matter

  • In David George’s public-market coverage universe, only three companies are growing revenue at over 30%.
  • One driver of companies staying private longer is that private capital markets have become deeper and more liquid, reducing the need to IPO until capital requirements become extremely large.
  • Private companies can partially replicate public-market liquidity and employee-compensation dynamics by running regular tender offers and managing stock-price volatility more tightly than public markets.
  • Public markets are increasingly tilted toward large-cap coverage and check sizes, making it hard for small-cap public companies to attract research attention and investor ownership.
  • Founder aversion to public-market stock volatility is heightened by recent tech drawdowns because large declines can reduce perceived employee compensation and create retention challenges.
  • A primary catalyst for going public is the need for much larger pools of capital and potentially cheaper cost of capital than is available in private markets, including for debt and other financing.

Ai Stack Structure And Implications For Incumbents (Budgets, Layering, Pricing)

  • Non-model-owning AI companies can remain defensible by compounding industry-specific context (workflows and data) and providing an accountable vendor relationship including support, integrations, and partnerships.
  • Incumbent software vendors are described as having high gross dollar retention, while net dollar retention has steadily declined since 2021.
  • Incumbent systems of record are expected not to be ripped out, but new AI vendors are expected to build action-taking layers on top of them, shifting value capture away from incumbents.
  • If software creation becomes much faster, vendors are expected to proliferate SKUs into adjacent domains, increasing competitive intensity for platform incumbents.
  • Model companies are expected to act as horizontal providers for general tasks, while many vertical functions are expected to be served by independent application vendors.
  • Public-market incumbent software vendors are expected to face slower growth because incremental buyer budget is shifting toward AI initiatives rather than expanding existing SaaS spend.

Secondary Liquidity, Cap-Table Governance, And Signaling

  • Founders generally dislike SPVs because they want transparency and control over who appears on the cap table.
  • Some SPV promoters are believed by founders to misrepresent how capital is assembled.
  • Founder or CEO willingness to sell (or refusal to sell) shares in a secondary transaction is treated as a confidence signal about a company’s future performance.
  • A common SPV pathway is presenting as a single fund to founders while aggregating many investors into a single-deal entity, obscuring end investors from the issuer.
  • SPVs are less diversified than diversified venture funds, and can be devastating for investors if the single underlying company performs poorly.
  • SpaceX is described as running employee tender offers about twice per year.

Ai Infrastructure: Capital Intensity, Throttling, And Utilization

  • AI infrastructure buildout may require on the order of $5 trillion over the next five to seven years.
  • AI infrastructure buildout at the scale described could push leading AI companies toward public markets as capital needs scale.
  • AI’s large capital needs could either push companies toward public markets for financing or further entrench private-market dominance via increased private-capital inflows to where growth is concentrated.
  • Large pools of Middle Eastern capital are characterized as a continuing funding source that may enable AI companies to stay private longer.
  • AI infrastructure investment decisions can be adjusted in roughly 12-month cycles by monitoring demand rather than requiring irreversible multi-year commitments.
  • The speaker claims there are effectively no “dark GPUs,” with new and older accelerators being immediately utilized and pricing holding up across generations.

Private-Tech Scale And Concentration

  • Highly valued private tech companies represent about $5 trillion in aggregate market capitalization, roughly a quarter of the S&P 500.
  • The 10 largest private companies account for about 40% of the aggregate private-tech market capitalization discussed.
  • Because many large tech companies remain private, a combined private-plus-public index would include more non-profitable firms and could make overall tech valuations look richer than public-market-only comparisons suggest.
  • Private-tech market capitalization has increased by roughly 10x over the last 10 years.
  • The number of public companies has halved over about 20 years.

Watchlist

  • AI infrastructure buildout may require on the order of $5 trillion over the next five to seven years.
  • AI infrastructure buildout at the scale described could push leading AI companies toward public markets as capital needs scale.
  • AI’s large capital needs could either push companies toward public markets for financing or further entrench private-market dominance via increased private-capital inflows to where growth is concentrated.
  • Large pools of Middle Eastern capital are characterized as a continuing funding source that may enable AI companies to stay private longer.

Unknowns

  • What is the precise methodology behind the $5T aggregate private-tech market capitalization estimate and its comparison to the S&P 500?
  • How frequently and at what pricing/discounts do large private companies actually run tender offers, and what fraction of employees/investors can participate?
  • How large and consistent are the incremental annual costs of being public across company size bands, and which cost components dominate?
  • To what extent are public-market small-cap coverage and ownership constraints quantitatively worse today than in prior periods?
  • How robust is the claimed shift in value creation timing (pre-IPO vs post-IPO) across sectors and IPO vintages, and what is the exact cohort definition?

Investor overlay

Read-throughs

  • Rising AI infrastructure capital intensity may pull leading AI firms toward IPOs mainly to fund scaled capex and enable stock-funded M&A, rather than for baseline liquidity.
  • Public markets may show fewer high growth revenue names as deep private capital and internal liquidity keep fast growers private longer, shifting where growth is accessible.
  • Non-model-owning AI application vendors can stay defensible by compounding workflow and data context plus accountable delivery, potentially shifting spend from systems of record toward action layers and outcome pricing.

What would confirm

  • Large AI companies publicly discuss IPO timing explicitly tied to capex scale, financing flexibility, or M&A currency needs rather than employee liquidity.
  • More frequent or larger tender offers at major private tech firms provide employee and investor liquidity without listing, alongside continued private funding inflows.
  • Enterprise AI buying shifts toward accountable vendors with integrations and support, with pricing conversations moving toward outcomes and increased SKU proliferation from faster software creation.

What would kill

  • AI infrastructure spend does not materialize at the implied scale, or utilization loosens meaningfully, reducing the need for outsized annual capex cycles.
  • Tender offers fail to provide broad, repeatable liquidity, increasing pressure for IPOs driven by employee and early investor liquidity needs.
  • Vertical AI vendors without model ownership fail to sustain differentiation despite workflow context and delivery, suggesting defensibility requires model ownership or incumbent systems remain dominant without budget reallocation.

Sources