Anthropic Raises $45B but Falls Short on Compute & Thoma Bravo Hand Back Medallia Keys to Creditors

OpenAI, Anthropic, and the "Agent" Shift

• OpenAI's recent performance miss: OpenAI missed user growth and revenue targets, causing stock drops in CoreWeave (-5%) and Oracle (-7%).
  • Root cause: Failure to build superior models in the second half of last year led to a marked decline in market share to Anthropic.
  • Current status: Recent models (e.g., o5.5) are viewed as strong, potentially better than current Anthropic offerings, suggesting the market reaction is pricing in "old news."
• Anthropic's compute constraints: Despite model success, Anthropic is facing a critical compute shortage.
  • Hyper-scaler intervention: Hyperscalers (Google, Amazon, Microsoft) poured $45 billion into Anthropic to secure compute capacity.
  • Capital intensity: Building AI models now requires massive upfront CapEx; for every $1 of projected revenue growth, $4-$5 of CapEx is needed 24 months in advance.
  • Risk profile: Forecasting compute needs two years out with 10x growth targets creates a high-risk environment for over- or under-investment.
• The "Agent" Market Thesis: The conversation predicts a shift from human-led selection to AI-agent-led vendor selection.
  • Future dynamic: AI agents will autonomously choose LLMs and tools based on performance, potentially bypassing human brand preferences (e.g., humans preferring Claude for coding vs. agents preferring OpenAI).
  • Oligopoly structure: The market is converging on a three-way oligopoly (OpenAI, Anthropic, Google Gemini) as agents favor leaders with momentum and innovation.
  • Impact on legacy SaaS: Traditional "system of record" software (e.g., Jira, Canva, Project Management tools) faces existential risk as agents render these tools obsolete for autonomous workflows.
• Public Market Reaction:
  • Investor alignment: Public markets correctly identified that tools like Atlassian and Monday.com (human-centric) are less valuable than infrastructure enablers (Twilio, Cloudflare) in an agent-first world.
  • Agent Wars: Winning the "agent layer" (the interface where agents operate) is now considered the primary competitive battleground, potentially more critical than the foundation model itself.

Strategic Deal Activity and Regulatory Friction

• Meta's blocked acquisition: China blocked Meta's $2 billion acquisition of Manus (a Singaporean AI company).
  • Regulatory intent: The move is interpreted as leverage against Meta and a strategic counter to US sanctions on NVIDIA chip exports to China.
  • Outcome: Investors who exited the deal prior to the blocking are unlikely to return capital; the deal will likely be unwound entirely to prevent future US tech transfers.
  • Geopolitical signal: This marks an escalation in the "AI War," signaling that China will retaliate against US tech dominance by restricting access to Chinese research talent and markets.

Private Equity Distress and the "Medallia" Precedent

• Medallia liquidation: Tomo Bravo handed Medallia over to creditors, resulting in a $5.1 billion equity wipeout.
  • Nature of failure: The deal was not over-levered (mostly equity) but overpriced; the company could not service $2 billion+ in debt with low-growth, pre-AI revenue streams.
  • Sector implication: Enterprise software that requires significant transformation to AI (e.g., survey/customer feedback tools) faces "terminal value" questions and consolidation risks.
  • Vendor consolidation: AI-driven efficiency is driving a 30-50% shift in IT spend toward consolidation, targeting legacy products for elimination before major systems like ERP.
• Exit Market Contraction:
  • Loss of the "Plan B": The private equity exit route for B2B software is narrowing; PE firms are becoming highly selective, primarily targeting AI-enhanced winners at steep discounts.
  • IPO bar: Public market appetite for IPOs has shrunk; companies now need $400M+ ARR (preferably $1B+) and 40%+ growth to access liquidity.
  • Founder solutions: An emerging trend involves founders merging or selling mid-sized companies to peers/friends rather than exiting to the public market, as strategic acquirers (Big Tech) have limited appetite for non-AI-native B2B software.

Venture Capital Market Dynamics

• Revenue transparency (Gary Tan/Gary's memo): Y Combinator (YC) and investors are facing rampant ambiguity in Seed-stage revenue reporting.
  • Bullshit ARR: Startups are increasingly using multiple revenue definitions (committed, usage-based, core software) to inflate numbers, creating a fraud risk.
  • Market correction: YC's push for standardized, audited revenue definitions is a shrewd attempt to preserve trust and market value in the seed ecosystem.
• Investment Product Evolution:
  • Thrive Capital: Launching an "Eternal" fund focused on non-digital assets (e.g., sports teams like the San Francisco Giants) that are immune to AI displacement, contrasting with traditional venture's focus on tech.
  • Robinhood/AngelList Funds: Launching public-accessible funds for private mega-cap startups (OpenAI, SpaceX, Anthropic).
    • Fee structure: Charges ~3.6% annual fee, which is high for public markets but comparable to traditional VC fee drag (2% + 20% carry) given the high-risk profile.
    • Valuation risk: While offering exposure to top-tier assets, the valuations ($3B-$4B range for some) present significant downside risk if IPO performance lags public comps.

Portfolio Construction & Investment Strategy

• The "Fewer, Bigger" Thesis: Venture capital is shifting toward a Darwinian model where 10 massive exits must cover the losses of the remaining portfolio.
  • Early stage: Requires higher concentration in fewer, larger bets to ensure at least one "unicorn" outcome.
  • Late stage: Requires investing only in companies already at public scale, limiting the universe to a small number of viable candidates.
• Stock Recommendations:
  • NVIDIA: Recommended as the primary "pure play" on the AI infrastructure boom; the safest bet for upside despite market saturation risks.
  • Google: Recommended for risk-adjusted returns due to diversified cash flows and the ability to leverage surplus compute capacity for multiple AI strategies.
  • Sell Signal: Figma and Duolingo are identified as holdings to liquidate due to the threat of AI agents rendering their human-centric tools obsolete.
• Private Equity Outlook: The classic PE model for B2B software is breaking down; if the underlying software becomes non-durable due to AI, the debt-based leverage strategy becomes unsustainable, potentially leading to a systemic shift in the asset class.