# Why is Google a Threat to Open AI Recent headlines (Dec 3, 2025) such as [OpenAI’s “Code Red”](https://garymarcus.substack.com/p/openais-code-red) by Gary Marcus indicate that something important may have changed in AI. But what? Open AI had competitors such as DeepSeek and Anthropic before. **Why is the threat posed by Google different?** I start with an answer by Claude, add some observations and finish with a note on the future of AI. ## Claude's Answer **Google's Different Business Model** - *OpenAI/Anthropic/Others*: Must charge users because AI is their product - *Google*: AI is just a feature to protect their real business - Makes ~$300 billion/year from ads - AI enhances their core products (Search, Gmail, Docs) - They make money from your attention and data, not from AI subscriptions **Historical Precedent**: - *Gmail (2004)*: Killed paid email services by offering 1GB free - *Google Maps*: Destroyed companies like Garmin by making navigation free - *Android*: Given away free to compete with iOS - *Chrome*: Free browser that killed Internet Explorer - *Google Photos*: Killed many photo storage services **The Economics**: Google can lose $10 billion/year on AI and barely notice - that's just 3% of revenue. For Google, AI is a: - *Loss leader*: Like free samples at Costco - *Defensive move*: Protects search from ChatGPT - *Data collection tool*: More user interactions = better ad targeting - *Moat strengthener*: Keeps users in Google ecosystem **Real-World Impact** - Google's Gemini integrated free into Gmail, Docs - Free AI features in Google Search - Gemini Flash: extremely cheap/free API tier - NotebookLM: Powerful AI tool, completely free **The Nightmare Scenario for OpenAI** isn't Google having better AI. It's Google having "good enough" AI that's free. Because: - 90% of users only need "good enough" - Free beats paid for most consumers - Convenience (built into existing tools) beats quality ## Additional Observations - Gemini is uptodate, even conerning the events of the day in news and politics. - Gemini can be used to summarize and analyze tabs open in your browser. ## The Future of AI? What do these developments mean for the future of AI? While Gemini browser integration is a huge step forward in terms of UX, the next major improvements will come from integrating AI into non-browser based workflows. For privacy and security reasons this will require the huge of LLMs that run locally on each device and are communicating with the "wild web" only in ways controlled by each user. There are already many efforts in this direction labelled variously - On-device AI - Edge AI - Client-side AI - Embedded AI - Offline AI - Private AI - Self-sovereign AI - ... I like **local-first AI**. [Will **local-first AI** do for Google what Google AI-integration now does for OpenAI?](https://hackmd.io/@alexhkurz/ryJ6QZ0bWg)