Anthropic opened an office in Bengaluru this month — its second base in Asia after Tokyo — and used the moment to say something specific: India is now the company's second-largest market worldwide for its Claude models, and its India revenue has doubled since October 2025. Running the new office is Irina Ghose, who spent two years as Microsoft's India managing director before leaving that role in December 2025. That is a real hire, a real building, and a real claim about where an AI company's growth is coming from. It is also, on the numbers that matter most, an unverifiable one — Anthropic has not disclosed what India revenue actually is in dollar terms, only that it doubled off a base nobody outside the company can see.
What the company says people are actually using it for
The one figure Anthropic did put a number on is usage composition, not revenue: close to half of all Claude activity in India involves computer-science and mathematical work — application development, modernising existing systems, deploying production software. That is a notably different profile from consumer chatbot usage, and it tracks with India's position as the world's largest base of professional software developers. It also lines up with two concrete integrations announced alongside the office launch. The Ministry of Statistics and Programme Implementation has stood up a government MCP server, a connector that lets AI systems query official statistics directly rather than through a search result or a PDF. And Swiggy is building its own MCP integration so that ordering groceries or booking a restaurant reservation can happen through a Claude conversation rather than the Swiggy app itself.
Model Context Protocol (MCP), for readers meeting the term here, is an open standard Anthropic introduced for connecting an AI model to outside tools and data sources — a government statistics database, a food-delivery backend — without a custom integration for each one. Its adoption by a government ministry and a listed consumer company in the same announcement is a more concrete signal of real usage than the revenue claim, because it's independently checkable: either the MCP servers exist and work, or they don't.
What the office itself signals
Anthropic's choice of city is not incidental. Bengaluru is where India's software engineering talent, its biggest technology employers and its startup capital are already concentrated, and it is the second Asian outpost Anthropic has opened after Tokyo — ahead of any standalone office in a market like Singapore or Southeast Asia that some rivals have prioritised first. Ghose's mandate, as described alongside the launch, is to hire across both technical and enterprise roles, which suggests the office is meant to do two different jobs at once: build product and support relationships with the large Indian companies Anthropic is trying to sell Claude into. Running both functions out of one newly opened office, under one India managing director, is a bet that the two feed each other — technical credibility helping the enterprise pitch, and enterprise deals generating the kind of real-world usage data that improves the product for technical buyers.
A hiring boom next to a layoff story
The timing lands two days after this site covered a different India tech story: TCS cutting jobs while citing AI-driven efficiency, and Infosys hiring, with both companies' own numbers showing the "AI replaces workers" narrative was thinner than the headlines suggested. Anthropic's Bengaluru launch is the same city, the same industry, and close to the same week — but the opposite direction of travel: a foreign AI company opening an office and hiring across technical and enterprise roles, in the city whose IT services firms are the ones facing the layoff headlines. Both things can be true without contradicting each other. The AI companies selling the tools and the services companies whose labour model those tools are reshaping are not the same business, and one growing headcount in Bengaluru says nothing about whether the other's is shrinking for the reasons its press statements claim.
The land grab has more than one company in it
Anthropic is not alone in treating India as a demand market rather than only a cost base. OpenAI and Google have both been expanding commercial and partnership activity in the country over the past year, and neither discloses India-specific revenue either — this is an industry-wide pattern, not a unique claim. What makes a company's "second-largest market" statement worth reading skeptically isn't that it's false; it's that "second-largest" is a ranking, not a size. A market can be a company's second-biggest by revenue while still being small in absolute terms next to the US, if the rest of the company's international business is thin. Nothing in this announcement lets an outside reader tell the difference, and Anthropic — like its competitors — has no obligation to disclose the number that would.
A note on the source: this article covers Anthropic, the company whose Claude models power this site's own automated content pipeline. That relationship is disclosed here because it's directly relevant: readers evaluating this piece should weigh it exactly as they would coverage of any company by a publication that uses its product, and the absence of a disclosed India revenue figure is treated above as a real limitation, not a footnote. A publication with no stake in the company would flag the same gap; this one has more reason to flag it clearly, not less.
