Nvidia's GPU Loans Raise AI Doubts, Yosemite Bets $350M

A deep dive reveals Nvidia’s equity stakes and debt‑backed financing let CoreWeave and Nebius buy billions of GPUs, then use the hardware as collateral for more credit. The model fuels explosive growth but ties AI infrastructure demand to circular finance, questioning how much is genuine spend versus engineered leverage.

Reed Jobs says Yosemite’s second fund closed at $350 million, earmarked for building biotech startups that turn academic breakthroughs into cancer cures. The firm now has 17 staff, leverages AI for drug discovery and clinical trials, and splits capital, about a third into companies it creates itself. Blockbuster patent cliffs and AI progress are expanding the opportunity.

SAFEs and convertible notes aren't interchangeable; SAFEs are interest‑free, have no maturity date, and convert at the next equity round, while convertible notes accrue interest and carry a repayment deadline, impacting dilution and creditor rights. Picking the right instrument changes founder control and future financing terms.

Tom Tunguz shows that today’s LLM‑powered apps still require massive pipelines, tool‑orchestration, observability, and guardrails, essentially the same hidden infrastructure that plagued 2015 ML systems. The hidden technical debt now lives in data stores, retrieval‑augmented generation, and cost‑control layers, turning promised simplicity into an iceberg of engineering work.

The latest SaaS benchmarks show sub‑$1M ARR startups cut headcount from 13 to 7, while companies over $50M ARR saw ARR per employee tumble 34%. Public SaaS now needs 57 months to recoup CAC, flagging a looming cost‑efficiency crisis that could squeeze growth and investor returns.

LLM SEO (or GEO) lets brands appear inside ChatGPT, Claude, and Perplexity answers, turning a single AI response into a high‑value traffic source. Early adopters have lifted daily clicks from a few hundred to over a thousand, making this an underpriced, compounding distribution channel for founders.

Anthropic lured back Boris Cherny and Cat Wu, the Claude Code leads who left for Anysphere, just two weeks after their departure. The rapid rehiring underscores how fierce the AI talent war has become and forces companies to rethink retention strategies.

Doug argues that writing code remains a vital thinking tool, not a production task, because it forces engineers to engage directly with execution details that AI can miss. He warns that neglecting hands‑on coding erodes system fragility and team ownership, affecting long‑term reliability of AI‑assisted development.

Google paid $2.4 billion to acquire AI startup Windsurf primarily for its engineers, illustrating the premium ($100 M‑plus) top AI researchers now command. Meta is matching that spend, while Netflix treats talent like a pro‑sports roster, keeping stars, cutting under‑performers, to stay ahead in a race for artificial superintelligence.