💰 Supabase Says No to $Millions

Hey folks,

Alphabet is approaching $4 trillion in market cap, becoming the fourth company ever to reach this valuation milestone, with its stock up 70% this year far outpacing Microsoft and Amazon. The surge is driven by Gemini adoption and broader AI infrastructure momentum, as the Magnificent Seven now command a combined $21.5 trillion in market value. This concentration of capital in AI compute providers signals where enterprise budgets are flowing, and technical managers need to understand which platforms are winning institutional backing to inform multi-year vendor strategy and hiring priorities.

Let's dive in..

💰 Supabase Hit $5B By Saying No to $Million Checks

🐝 The Buzz: Supabase went from $2 billion to $5 billion valuation in just months by repeatedly rejecting seven-figure contracts from deep-pocketed customers demanding extensive customizations. CEO Paul Copplestone calls these decisions "very painful," but his bet is simple: stick to the product vision, optimize for developer happiness over enterprise sales teams, and the market will follow. The strategy appears to be working as Supabase positions itself as the backend of choice for the vibe-coding world.

• The trade-off that hurts: When a company waves a million-dollar contract in your face but wants features that don't benefit your broader developer community, most startups fold. The reasoning is calculated, enterprise customizations create technical debt, fragment your roadmap, and turn you into a services company disguised as a product business.

• Developer-first vs enterprise-first architectures: Supabase's approach reveals a fundamental fork in platform strategy, build for the individual developer or build for the procurement committee. Traditional wisdom says you eventually need enterprise deals to scale revenue, but Supabase is betting that AI-assisted coding creates such explosive demand for backend infrastructure.

• Why this works now (maybe): Supabase believes "the death of Oracle won't take a generation," which is a bold claim that depends on AI dramatically accelerating application development. If Cursor, Replit, and similar tools genuinely enable 10x more developers to ship production apps, those apps need databases, and developers prefer Postgres over Oracle.

💡 Takeaway: For engineering leaders building platforms, Supabase's strategy offers a counterintuitive model, maintain strict product vision even when it costs you immediate revenue. The key distinction is they're not just focusing on product, they're making specific bets about AI tooling creating database demand at unprecedented scale.

Together with DESelect

2025 State of Marketing Operations Report

Stay ahead of industry shifts with the definitive 2025 State of Marketing Operations Report. This all-new edition reveals what 34 marketing ops leaders are prioritizing now: centralized audience management, AI progress, platform-agnostic strategies, and more. Don't miss the action-ready recommendations and firsthand perspectives powering high-performing teams in SaaS, retail, healthcare, and beyond. Get your copy to strategize smarter!

Download the 2025 State of Report

🥇 DeepSeek Beats OpenAI for $294K

🐝 The Buzz: China's DeepSeek just dropped DeepSeekMath-V2, scoring IMO 2025 gold medal performance (5 of 6 problems solved) exceeding the best human score of 90 for a training cost of $294,000. Compare that to OpenAI's hundreds of millions per model, and you see why US startups are increasingly bypassing American providers for Chinese open-source alternatives. The strategic message is unmistakable: China is challenging the narrative that frontier AI requires Silicon Valley budgets, and they're doing it with full technical transparency while Western labs stay proprietary.

• Transparency as competitive weapon: Unlike OpenAI and Anthropic keeping architectures proprietary, DeepSeek published full technical documentation for DeepSeekMath-V2 and released MIT-licensed open weights. By proving Chinese labs can match Western frontier performance while remaining open, DeepSeek undermines the "we need to stay closed for safety" narrative that justifies Western AI companies' moats.

• The geopolitical AI narrative shift: DeepSeek's releases reframe US-China AI competition from "who builds the best models" to "who builds the most cost-efficient path to capability parity." When DeepSeek R1 launched in January 2025, Nvidia lost $593 billion in market value, the largest single-day market cap loss in US history as investors realized AI compute demand might not scale as expected if Chinese labs achieve comparable results on smaller budgets.

• What this means for engineering teams: If Chinese open-source models genuinely approach frontier capabilities at 1% of training costs, do you pay OpenAI's API premiums, or integrate DeepSeek's MIT-licensed weights and fine-tune locally? The tradeoff isn't just cost - it's control, latency, data sovereignty, and whether you believe US labs' proprietary advantages justify 100x price differences. DeepSeek's bet is that most engineering teams will eventually choose efficiency over brand, especially as Chinese models keep publishing benchmarks that close the gap.

💡 Takeaway: DeepSeek's strategy challenges engineering leaders to question whether frontier AI requires frontier budgets, or whether architectural innovation and transparency can compete with closed, capital-intensive approaches. Whether this holds for general-purpose models remains unproven, but the geopolitical and market implications are already reshaping vendor strategy.

📰 Like newsletters? Here are more our readers enjoy.

💻️ How to Turn Documents Into Illustrated Stories

• Go to Google Gemini

• Click the Explore Gems link in the left menu

• Select Storybook Gem

• Upload your policy document like a Privacy Policy

• Prompt the following:

Turn this privacy policy into a story about an employee learning to handle customer data securely. Make it engaging and include illustrations

🐝 AI buzz bits

🤖 Figure AI faces whistleblower lawsuit claiming F.02 humanoid robot hands generate 2x the force required to fracture human skulls, based on internal impact testing before the company's valuation jumped from $2.6B to $39B.

🔐 OpenAI terminated Mixpanel analytics after API user data breach via SMS phishing that exposed names, emails, locations, and browser metadata for OpenAI API users through an unauthorized data export.

🤖 China's economic planning agency warned of bubble risk in humanoid robotics as 150+ companies produce remarkably similar models. Beijing designated humanoid robotics as a key growth driver through 2030, but widespread factory or household adoption remains years away.

⚠️ Google slashed Gemini 3 Pro free tier limits days after launch, cutting image generation and reducing thinking mode access to vague "basic access" with fluctuating daily limits. The rapid restriction post-launch signals infrastructure strain under viral adoption, contrasting with the optimistic benchmarks and pricing announced just weeks earlier.

🔥 AI productivity tools

For a full list of 1000+ AI tools, visit our Supertool Directory

🤘 Meme