With the IndiaAI mission gaining traction, investing in AI across different layers, particularly the foundational layer, has come under a tighter spotlight. As the debate around foundational models becoming commoditised increase, many contend that the edge shifts to those who can execute quickly, integrate deeply, and build for real-world use cases.
In a recent conversation with AIM, Prayank Swaroop, a partner at Accel India, spoke about the VC fund’s investment thesis in India. He believes that India’s most promising opportunities in AI lie elsewhere.
“While we’re closely monitoring developments in foundational models, the highest upside right now is clearly in applied AI,” he said.
Recently, IndiaAI Mission announced the selection of three more startups—Soket AI, Gnani.ai, and Gan.AI—to develop indigenous foundation models. This brings the number of startups under the foundation model development initiative to four, including the previously announced Sarvam AI.
Sarvam’s funding comes from investors like Lightspeed India Partners, Peak XV Partners, Lightspeed Venture Partners, and Khosla Ventures, among others.
The announcement drew a slew of criticism on social media.
Swaroop asserted that one doesn’t need to be OpenAI to create a transformative AI product. This highlights the fact that many of the most innovative developments are coming from companies that leverage existing models like GPT and Claude.
Inside Accel’s Investment Thesis
According to Swaroop, Accel is increasingly concentrating on AI across three main areas: agentic enterprise platforms, vertical AI products and AI-enabled services.
“This surge [of AI adoption] is driven by India’s unmatched engineering talent pool, cost efficiency, and access to domain-specific datasets,” Swaroop said.
The firm’s latest $650 million fund is designed to support startups that bring clear use cases. A key part of its strategy is product-led growth, backing companies that can scale through user demand rather than relying heavily on sales teams, which they believe gives these startups a stronger foundation for long-term success.
Why the Application Layer?
However, this raises the question: why is the application layer considered India’s stronghold in AI? Swaroop makes the case for an AI surge in India driven by its engineering talent pool, cost efficiency, and access to domain-specific datasets.
Swaroop believes that the bottleneck has shifted from engineering resources to product thinking and strategic depth. “AI is making software development easier, but the winners will be those who build durable products that solve high-value pain points,” he said.
According to him, founders who adopt a “build for India, scale for the world” mindset, stay user-obsessed, and execute with speed are best positioned to create globally relevant solutions.
He argued that the “services as software” model, where India’s traditional BPO strength is transformed through AI automation, is gaining significant traction. Furthermore, India’s unique advantage lies in the application layer as startups are leveraging open-source and accessible foundational models to build verticalised solutions in healthcare, legal operations, and financial services.
What’s Everyone’s Missing About AI?
“If anything, AI remains underhyped in terms of its true potential,” he said, talking about how the breadth and depth of AI’s impact across sectors are only beginning to be realised.
Swaroop explained that, contrary to common belief, AI’s potential remains underhyped in several key sectors. Agentic AI is just beginning to demonstrate its power with examples like Genspark (coding), Manus (enterprise workflows), and August.ai (preventive healthcare), showing early signs of a broader shift in productivity tools.
Another underrated opportunity, he believes, is India-first consumer AI: products built for local languages, cultural contexts, and price sensitivity. AI-powered regional entertainment and Bollywood content generation are in their early days, gaining traction.
Beyond Foundational Models
The recent acquisition of Windsurf by OpenAI and the appointment of a CEO of applications signal a clear convergence between foundational model development and AI applications.
“You don’t have to be OpenAI to build a transformative AI product, as many successful startups, such as Cursor, RapidClaims, Chronicle and Rocket.new from our portfolio, are building on top of existing models like GPT and Claude,” Swaroop said.
However, this doesn’t close the door for startups, he argues. The above-mentioned startups are creating meaningful solutions by leveraging existing models, such as GPT or Claude. The competitive edge lies in addressing deep user pain points, maintaining control over the product stack, and optimising for performance and cost.
Swaroop argued that localisation and sensitivity to price and context are essential for capturing the Indian market.
He talks about how while for most Indian founders, building on top of open-source or accessible LLMs offers better returns. However, there is growing interest in building India-specific models that account for local regulation, language diversity, and cost constraints.
