India's AI Funding Boom in H1 2026 Left Most Startups Starving

The Indian AI ecosystem was abuzz with activity in the first half of 2026, with startups raising nearly triple the venture capital they bagged in the comparable period last year. However, the momentum is being led by select large players as investor interest consolidates.

In H1 2026, AI-native startups in India raised $981 million, a nearly 3x jump from $339 million raised between January and June 2025, according to Tracxn data analysed by AIM.

However, the funding boom was highly concentrated among a small group of companies building the infrastructure that could underpin the country’s AI ambitions. Neysa’s $600 million Series B haul and Sarvam AI’s $234 million pickup together accounted for nearly 85% of disclosed funding in the latest period.

Moreover, the number of deals also declined significantly, from 54 in H1 2025 to just 32 this year.

Geography-wise, Bengaluru continues to attract the largest share of investments, reinforcing its position as India’s AI capital. But that inflow is now becoming increasingly selective.

Venture capital is headed towards companies building foundational layers, from models and cloud capacity to semiconductors, data pipelines, and orchestration platforms.

The trend is visible beyond the two mega rounds. Startups including Nava, Portkey, HrdWyr, Deccan AI, Cluster Protocol, and Tsavorite raised fresh capital across AI cloud infrastructure, enterprise AI, governance, data infrastructure, chip design, and compute. 

Why Infrastructure?

Over the past 18 months, the global AI conversation has evolved from experimenting with generative AI applications to who owns the infrastructure powering AI.

The Indian government, through the IndiaAI Mission, is investing in sovereign AI capabilities, building domestic compute capacity that gives greater control over the data, models, and hardware. 

At the same time, enterprises are moving beyond pilots and looking to deploy AI in production environments, creating demand for scalable infrastructure that meets requirements for cost, latency, data residency, and governance.

For Deepak Gupta, General Partner at WEH Ventures, the move towards infrastructure is a classic “picks-and-shovels” opportunity. During the first wave of the generative AI boom, investors poured money into application-layer startups and wrappers. Many of those companies, however, faced low barriers to entry, intense competition, and high API costs.

“No matter which app wins the consumer race, everyone needs compute power, localised data centres, and clean data pipelines,” Gupta tells AIM. He adds that tightening data-residency requirements could make local AI infrastructure especially important for sectors such as banking and defence, where sensitive data cannot simply be routed through offshore servers.

That logic is increasingly shaping the way capital is deployed. Rather than backing another application built on top of someone else’s LLM, investors are looking for companies that own critical parts of the AI stack. That’s how Sarvam was able to bag $234 million in its Series B round led by HCLTech.

“Having an infusion of capital allows us to take more and bigger bets,” Vivek Raghavan, Co-founder of Sarvam AI, earlier told AIM. “We want to train larger models, host models at a very large scale, and build agentic orchestration layers.”

The Next Frontier

A notable deeptech deal in H1 2026 came from HrdWyr, a semiconductor startup, which raised $13 million in a Series A round led by IdeaSpring Capital. HrdWyr is betting that the next phase of AI will be shaped by where intelligence runs, not just what it generates.

“The first generation of AI happened largely in the cloud. That is shifting massively towards computing on the edge,” Shiva Ramamurthy, Co-founder of HrdWyr, notes. 

The company is building intelligent edge semiconductor chips designed to process AI workloads closer to devices, reducing dependence on hyperscale cloud infrastructure. 

Ramamurthy believes the industry is at the start of a much larger transition. “Virtually every chip that has existed over the last 30 years will get reinvented,” he says, referring to the move towards intelligent edge computing, robotics, and physical AI systems.

For IdeaSpring Capital, the investment aligned with the firm’s thesis of backing companies building foundational technologies for India. “We wanted companies building for the Indian market first,” Naganand Doraswamy, Managing Partner at IdeaSpring Capital, says.

Such bets come with longer timelines than traditional SaaS investments. “For semiconductor companies, you’re talking three to four years before you see the light of day.”

Historically, that gestation period kept many venture investors away from semiconductor startups. AI, however, is changing the equation, and semiconductor startups are now being treated as strategic AI infrastructure plays.

Two-Speed AI Market

While the investor focus in H1 has clearly been on infrastructure, it does not necessarily mean application-layer AI has lost relevance. 

Gupta expects the next breakout AI startup will be determined by enterprise adoption, as building foundation models or raw hardware infrastructure from scratch is capital-intensive and difficult for most startups to sustain.

“The true moat isn’t the underlying math of the AI; it is the proprietary enterprise data and domain-specific context it solves,” Gupta notes.

Pankaj Makkar, Managing Director at Bertelsmann India Investments, shares a similar view. India’s services DNA, he points out, makes the country well placed to enable enterprise adoption, much as it did during earlier SaaS and internet cycles.

At the same time, Makkar notes that the infrastructure story could become more viable as Indian corporates begin using their balance sheets to fund startups. HCLTech’s investment in Sarvam AI is an important proof point. Infrastructure companies require much heavier capital commitments than enterprise AI adoption companies, and strategic corporate backing could help bridge that gap.

The result may be a two-speed AI market. A small number of capital-intensive companies building sovereign models, compute infrastructure, and chips will attract outsized rounds. A broader group of startups will compete to translate those capabilities into production-grade enterprise products across sectors such as financial services, healthcare, manufacturing, cybersecurity, and customer support.

Looking ahead, Gupta expects funding in the second half of the year to remain concentrated around a few category leaders rather than broadening evenly across the ecosystem. The “spray and pray” days of venture capital, he says, are being replaced by a flight to quality, with investors looking for technical moats, strategic partnerships, and access to scarce compute capacity.

Makkar, however, notes that while mega deals have captured attention, the long tail of AI investments is still growing as companies test use cases across enterprise, consumer, and SMB markets. Over time, he expects consolidation as winners emerge and the technology matures.