The world's back office is watching its business model get tokenized in real time.
The Summary
- India's market sentiment is cooling as investors redirect capital toward AI infrastructure plays, leaving software services and outsourcing exposed
- Ruchir Sharma points to chronically low R&D spending as India's structural disadvantage in the AI value chain
- The upside: India could capture value later when AI adoption shifts from infrastructure build-out to productivity implementation
- Near-term pain, long-term option value, if the country can close the innovation gap
The Signal
India built a $250 billion software services industry on one premise: labor arbitrage. Western companies pay Indian engineers 30-40% less to do the same work. That worked when the work was coding, testing, maintaining legacy systems. It breaks when the work is training models, building inference infrastructure, and designing agent architectures.
Sharma's thesis is straightforward: India is on the wrong side of the AI trade because it optimized for services, not R&D. The country spends roughly 0.7% of GDP on research and development. South Korea spends 4.8%. China spends 2.4%. The U.S. spends 3.5%. You don't build frontier AI capability at 0.7%. You maintain it.
The immediate threat is agent displacement. The coding work that powered Infosys, TCS, and Wipro is exactly what LLMs eat first. Junior developer roles, QA testing, documentation, basic system integration. These aren't jobs that require cultural context or relationship management. They're structured tasks with clear outputs. Perfect agent territory.
"Low R&D spending and exposure of software and outsourcing jobs are weighing on sentiment."
But Sharma sees a second-order opportunity. Right now, capital is flooding into picks-and-shovels plays:
- Nvidia and compute infrastructure
- Data center buildouts in the U.S. and Middle East
- Chip design and manufacturing capacity
India has none of that. What it does have is 500 million English speakers, strong nominal GDP growth continuing through the 2020s, and a domestic market large enough to generate proprietary data at scale. The question is whether Indian firms can pivot from labor arbitrage to platform creation before the arbitrage window closes completely.
The valuation reset Sharma mentions is real. Indian tech stocks traded at a premium for two decades because they were seen as stable, high-margin service providers with dollar-denominated revenue. That premium is compressing. If you believed services margins would hold, you're repricing. If you believe India can build AI-native companies that serve its domestic market first and export second, you're accumulating.
The Implication
Watch where Indian capital goes next. If it flows into more service delivery with AI tooling bolted on, that's a managed decline. If it flows into foundational model research, agent frameworks for non-English languages, or vertical AI applications for India's agriculture and healthcare sectors, that's a real pivot.
For Web4 builders, India's challenge is instructive: you can't rent your way into the agent economy. You either own the infrastructure, own the models, or own distribution into markets no one else can reach. India has a shot at the third path, but the clock is running.