Indian startups hit $9.71 billion in H1 2026. The real story is compute, not code.

2026-07-11 — business growth strategy India

I was talking to an investor last week about why Indian AI funding suddenly spiked. They said "infrastructure". I asked why that mattered. They looked at me like I'd missed something obvious.

So I pulled the numbers. H1 2026: $9.71 billion across 877 rounds. That's 21% up year-on-year. India's third-largest ecosystem now, behind the US and China. Fine. But the real move is inside that figure.

AI startups alone pulled in $676 million across 57 deals. Compare that to $162 million across 30 deals in the same half last year. That's fourfold growth, more or less. Not gradual. Not expected.

Two rounds did most of the heavy lifting. Neysa—sovereign compute infrastructure—closed $1.2 billion in Q1. One of the biggest raises in Indian startup history. Then Sarvam AI, the Bengaluru team building foundation models for Indian languages, hit $234 million in their Series B and became a unicorn at $1.5 billion valuation. Those two deals alone rewrote what was possible.

Here's where it gets specific, though.

The IndiaAI Mission sanctioned Rs 10,371.92 crore back in early 2024. That's $1.25 billion. They deployed 34,000 GPUs across Indian data centers and made them available to registered startups at Rs 65 to Rs 115 per GPU-hour. About 42% below market rates. Another 20,000 GPUs are coming before year-end. They're targeting 100,000 public GPUs by December 2026.

Actually, that's not quite right—I should say those are the announced targets. Whether they hit them is another question. But the intention is clear: the government is removing the infrastructure barrier entirely.

Sixty-six percent of institutional investors Inc42 surveyed said the IndiaAI Mission actively shaped their AI investment thesis. That's not a nice-to-have stat. That's 2 out of 3 VCs changing their portfolio because of government compute availability.

Most consultants get this part wrong, including us sometimes. They treat government subsidies as a discount problem. "Startups save money, that's nice." But what actually happened was different. The talent pool was always there. The market was always there. What didn't exist was the ability to train models at scale without spending like a US venture firm. The IndiaAI Mission closed that gap. Capital followed.

Beyond AI, nothing much shifted. Fintech, healthtech, agritech, SaaS—they're still the backbone of the ecosystem. But something did change in how the money flows. Fewer companies got funded. Larger checks went to the ones that did. Investors are being more selective. Or maybe just wiser about which bets matter.

Sarvam's story is worth sitting with separately. India has 22 officially recognized languages. The market for foundation models that actually work in Hindi, Tamil, Telugu, Kannada isn't a side project. It's an enormous opportunity that Western models don't touch and probably can't solve. Investors stopped treating vernacular AI like a curiosity. They're treating it like a real technical problem with a real addressable market. That's a meaningful shift in how the ecosystem thinks.

The 21% growth also came with context I'd mention. Global venture markets were cautious through most of 2025. India's numbers partly held because domestic institutional investors stepped in when foreign VCs pulled back. The ecosystem is more resilient now, sitting on a wider base than it was five years ago. Less dependent on Silicon Valley appetite.

If you're deciding where to build or where capital should flow in the next wave, the answer from the data is clean: AI infrastructure and vernacular models. Not another payments app. Not another SaaS clone. The government put down compute. Private capital noticed. The question now isn't whether India has an AI sector.

It's whether what gets built here stays local or goes global.