20VC: Mag7 Earnings: Google & Amazon Win - Meta and Microsoft Falter | Anthropic's $50BN Raise & What it Means for a Potential IPO | Atlassian, Twilio and Five9 Beat: The SaaS Apocalypse Over? | Sierra's $15B Valuation: Peak or Potential

What It Covers

Harry Stebbings, Rory O'Driscoll, and Jason Lemkin analyze Mag7 earnings week — Alphabet and Amazon outperform while Microsoft stagnates without AI revenue and Meta faces CapEx skepticism. They cover Anthropic's $50B raise at a $900B valuation, Palantir's 134% RPO growth, SaaS rebounds from Atlassian and Twilio, Sierra's $15B valuation, and the Musk vs. Altman trial opening.

Key Questions Answered

• •Microsoft's AI dependency: Strip out Copilot and Azure AI revenue from Microsoft's results and the entire corporation runs flat. With $37B in AI ARR and $190B in planned CapEx, every dollar of Microsoft's growth narrative depends on AI bets paying off. Investors and operators should pressure-test any large-cap tech thesis by isolating AI contribution — if the underlying business is stagnant, the valuation risk is structural, not cyclical.
• •Coding as the token demand benchmark: To size the LLM market accurately, track token spend as a percentage of engineering salary at AI-native organizations. If that ratio stabilizes at 20–30%, Anthropic and OpenAI can grow into $500B+ revenue categories. If it settles at 5%, the math breaks. Current survey data shows most engineering teams spending 2–15% of salary on tokens — far below the threshold needed to justify frontier model valuations.
• •Palantir's enterprise AI positioning: Palantir wins because it can credibly deploy $10–100M AI transformation initiatives enterprise-wide — a scale no two-year-old AI startup can match. CEOs under board pressure to show AI progress need a single large initiative they can report on by June 30. Palantir fills that gap the way IBM and EDS did for prior technology waves, compressing multi-year sales cycles into weeks because every stakeholder now shows up to the first meeting.
• •Two-pronged SaaS survival test: SaaS companies need two simultaneous wins to escape the apocalypse: monetizing AI with existing customers AND attracting net-new customers. Atlassian passes the first prong via Rovo.ai upsell but shows slowing net-new customer counts. Twilio passes both — net-new customer growth may be up 40% year-over-year driven by AI-native companies like Eleven Labs. Companies achieving only one prong are deferring decline, not reversing it.
• •Autonomous AI agents cost less than expected: Running two fully autonomous AI agents — one handling VP of Marketing functions, one handling VP of Customer Success — costs approximately $254 per month in total token spend. This deflationary reality challenges assumptions about token consumption growth. Operators building AI agent workflows should audit actual monthly token costs before projecting LLM market size, as real-world consumption may be an order of magnitude below theoretical estimates.
• •Anthropic's CapEx-to-revenue leverage ratio: For every $1 of Anthropic revenue, roughly $3–4 in CapEx must be committed — much of it a year in advance, when revenue is 10x lower than forecast. At $10B run-rate revenue, the implied forward CapEx commitment reaches $30B. This structural dynamic means Anthropic's $50B raise at a $900B valuation is not excess — it is the minimum viable capital buffer to avoid being caught short on compute during a 10x growth year.
• •The non-manager mandate: Coinbase cutting 14% of staff signals a structural shift: executives who cannot personally execute — ship code, run campaigns, deploy agents — are being eliminated regardless of seniority. A CMO who cannot run their own campaigns, or a CCO who cannot autonomously reach 150 customers overnight using AI tools, is a liability. Founders should promote individual contributors with AI fluency over traditional managers, as the productivity gap between the two groups is now measurable and widening.