20VC: Andrej Karpathy Joins Anthropic & Anthropic Raises $30BN at $900BN Price | SpaceX Files S1: How Does it Trade | Cerebras Smashes Day 1: What it Means for IPOs | Why Mass Layoffs Are More Worrying Than Anyone Sees

What It Covers

Harry Stebbings, Jason Lemkin, and Rory O'Driscoll analyze Anthropic's $900B valuation raise, Andrej Karpathy joining Anthropic, Cerebras' 68% IPO pop, SpaceX's planned $1.75T IPO, and the accelerating wave of AI-driven mass layoffs across Meta, LinkedIn, Cisco, and Intuit.

Key Questions Answered

• •Anthropic valuation math: At 18x ARR with 10x year-over-year growth and near-zero IPO risk, Anthropic's $900B round represents better value than most Series A/B deals priced at 20-50x ARR on companies five years from liquidity. Investors like Altimeter and Green Oak are rationally choosing the larger, de-risked asset at a lower multiple over early-stage bets with higher multiples and more uncertainty.
• •Token spend trajectory: Salesforce spends $300M annually on Anthropic tokens — roughly $15K per engineer per year, or 4% of total engineering payroll. For OpenAI and Anthropic to justify trillion-dollar TAM projections, token spend must reach approximately 20% of total engineering payroll across enterprise software companies. Current spend suggests most enterprises are only 25% of the way toward that level.
• •AI agent cost reality: Actual token costs to run autonomous AI agents are far lower than assumed. The Klaviyo CEO confirmed running full AI VP-level agents costs roughly $2.57 per month in direct token spend. This deflates both the bull case for token revenue growth and the fear around AI implementation costs — the real constraint is workflow design, not compute expense.
• •SaaS re-rating is permanent: Legacy SaaS companies will never return to 2021 valuations of 50x ARR. The new ceiling for high-performing public SaaS is 17-18x revenue (Datadog), mid-tier lands at 6-10x (Figma), and struggling businesses trade at 3x. Investors should evaluate these companies purely on revenue growth acceleration and cash flow, not on proximity to prior peak multiples.
• •IPO window is selective, not open: Cerebras' successful IPO at $1.85 with a 68% first-day pop does not signal a broad IPO window. The threshold for a successful IPO now requires a differentiated hardware or AI infrastructure position, a marquee customer like OpenAI, and a backlog exceeding $24B. Software companies below Figma's scale and growth profile face continued IPO market resistance.
• •Mass layoffs carry underestimated political risk: Meta cutting 8,000 jobs, Intuit cutting 1,600, LinkedIn cutting 875, and Cisco cutting 4,000 — all attributed to AI efficiency — creates a compounding political backlash. Unlike prior tech cycles where displaced workers found adjacent roles, AI-driven layoffs leave workers with limited rehiring prospects. Tech leaders face a choice between proactive workforce reinvestment or escalating regulatory and social consequences.