OpenAI plans Chrome competitor browser as Perplexity's Comet and The Browser Company's Dia heat up browser wars

Summary

OpenAI is planning to launch a web browser built on Chromium within weeks, according to a Reuters report that broke the same day Perplexity announced Comet, its own AI-powered browser. The Browser Company's Dia browser entered beta roughly a month earlier, so none of these entrants will be first to market.

Both Perplexity's Comet and OpenAI's browser run on Chromium, the open-source foundation underlying Chrome and Edge. That choice grants both access to Chrome's extension ecosystem, a meaningful advantage over building from scratch.

The strategic move follows the oldest playbook in tech: find product-market fit with a killer use case, then vertically integrate and horizontally expand until you control the interface layer. Once you own the defaults, you own the user. This is the third browser war, distinct from the Netscape-versus-IE fight of the 1990s and the Chrome-versus-Firefox scramble of the 2000s.

Google's dominance in the search era made browser distribution nearly frictionless. The company could pitch a faster browser to billions of daily Google visitors. OpenAI's distribution advantage is less obvious. ChatGPT runs in the browser but sits inside a tab. Getting users to download and daily-drive a separate app is harder than embedding a prompt bar in an existing interface.

Perplexity's Arvin frames Comet as a solution to the MCP and connector problem. Rather than waiting for third-party integrations to surface context from other apps, users download Comet and let the agent browse tabs and pull relevant information directly. The value proposition is cleaner agent execution.

Browser lock-in is real. Users accumulate tabs, some pinned for years, that function as a makeshift operating system of open documents, spreadsheets, and services. Switching requires not just importing tabs but re-authenticating across dozens of services, a friction point neither Perplexity nor OpenAI has fully solved.

Owning the browser also means owning training data. Every click, form entry, and page interaction becomes a signal for training agents on how humans navigate the web. OpenAI's Operator already automates browser-based tasks on Chromium. A native browser surface would multiply that feedback loop.

Whether these ventures will mature into dominant products remains uncertain. Google bet on social networking and lost. OpenAI exploring browsers, devices, and chips is rational experimentation, but exploration does not guarantee victory across every vertical. The surface area of integration is wide, but so is the risk of dilution.