Agents Pick AI Models: OpenAI Resurgence and PE Playbook Breakdown

The Gist

• Agents now choose AI models and vendors, sidelining human decision-making
• OpenAI regains edge as AI workflows prioritize model effectiveness over brand loyalty
• Thoma Bravo's $5.1B Medallia wipeout signals PE-for-B2B playbook collapse
• Foundation model wars shift focus to agent-driven workflows

• AI agents, not humans, will determine the winners in the foundation model wars and the future of B2B companies.
• The PE-for-B2B playbook is breaking due to AI rendering classic B2B software non-durable.
• OpenAI missed Q4 numbers due to demand softening, while Anthropic ran out of compute despite better models.
• The capital intensity of AI requires $4 of capex for every $1 of run-rate revenue, creating a high-stakes bet.
• Medallia's equity wipeout highlights the risks of debt-heavy PE deals in low-growth, pre-AI companies.
• Founders may increasingly hand over their companies to larger peers rather than pursuing traditional exits.

• Evaluate whether AI agents will use your software to determine its long-term viability.
• Avoid debt-heavy PE deals for low-growth, pre-AI companies.
• Focus on profitability and scale for IPO-stage companies, as SaaS pixie dust credit has expired.
• Consider handing over your company to a larger peer if traditional exit paths are unviable.

• $45 billion (Anthropic raised from Google and Amazon to meet compute demand.)
• $5 trillion (Nvidia's market valuation.)
• $4 trillion (Google's market valuation.)
• $5.1 billion (Equity wiped out in Thoma Bravo's Medallia deal.)
• 21% (Medallia's reported net retention drop.)
• $300 billion (Estimated capex required for 5x growth on a $10B base over 24 months.)