The browser isn’t just for humans anymore. It’s becoming a workplace for AI agents, and whoever controls it wins.
Google dodged a bullet this week. They don't have to sell Chrome after all.
This is good news for Google. But it’s also a reminder of why every AI company suddenly wants its own browser. Perplexity has Comet, Anthropic is piloting a Chrome extension, and OpenAI is rumored to be working its own AI browser too.
Why does a browser matter so much to AI companies?
Here are the top three reasons (especially the last one):
1. Data advantage
Browsers see a) what users search for, b) the sites they visit, c) what web apps they use. For AI companies, this data leads to better search results, smarter personalization, and the ad and AI subscription revenues.
For context, Google Search generated $198.1 billion in 2024, 57% of Google's total revenue. Chrome plays a major role in driving that revenue (check out my previous post on this topic).
2. Ecosystem control
If you control the browser, you control the starting point of every online journey: the address bar, new tab page, search box. It decides whether you land on Google, ChatGPT, or Perplexity first.
Google does this by making Google Search the default in Chrome. People who switched to Chrome from Microsoft’s Internet Explorer did 48% more Google searches, and from Firefox 27% more. And once people are on Google, they tend to use more Gmail, Calendar, and Docs too.
Browsers are sticky too. They hold your bookmarks, history, passwords, and extensions. Once you're locked in, switching feels like moving houses. That's exactly what AI companies want: higher user switching cost.
3. The real prize: AI agent computer use
The most important reason is that browsers are becoming the operating system for AI agents.
Picture this future: You're working on strategic brainstorming in one tab. Meanwhile, ten other tabs have AI agents running, one doing market research, another booking your trip, another updating Salesforce CRM, another scheduling meetings.
In other words, in the future the browser stops being a window for humans, and becomes a workplace for AI agents.
AI browsers address two key issues:
No more integration nightmare: Many enterprise software is ancient. No modern APIs. Limited documentation. Just clunky interfaces built in 2001 (think government websites, legacy enterprise tools). But AI agents using browsers can work with anything that loads in Chrome. Since no integration is needed, this saves a lot of engineering time and effort.
Platform independence: Unlike mobile apps tied to iOS or Android or whether you’re using a PC or MacBook, browsers work everywhere. So for these AI companies, winning the browser means they don’t need to win the operating system war, they can just sit on top of it.
Counter-argument: But can't chat interfaces just replace browsers?
Something I've been thinking is: Why not just tell your AI chatbot "book me a flight" and skip the browser entirely? Even then, the browser is where the actual work happens. The AI still needs to navigate Expedia, fill out forms, click buttons. The browser is the execution layer, whether you're watching or not.
There's also AI-to-AI communication to consider. If AI agents start talking directly to each other through APIs or specialized protocols like MCP (which connects AI models to external systems), they could bypass browsers entirely. But most systems today still require human-like interaction, and legacy software won't get modern APIs anytime soon. Also, users may prefer AI agents that work through interfaces they can see and understand. Full AI-to-AI automation is likely still years away.
That's why AI companies can't settle for being just the chat interface. If they own both the conversation layer (where you talk to AI) AND the browser layer (where AI does the work), they own the whole stack: how you interact with AI and how AI interacts with the world.
The race is on
The browser wars of the 2000s were about human eyeballs. The browser wars of the 2020s are about AI agents. So winning the browser means owning the future interface of work itself.
That's a prize worth racing for.