New Delhi just hosted the most consequential AI gathering the Global South has ever seen. About USD 210 billion in infrastructure commitments. Every major AI CEO in one room. The New Delhi Declaration endorsed by 88 nations. India formally joining Pax Silica, aligning with a US-led coalition securing the full silicon stack from critical minerals to semiconductor fabrication.
The infrastructure story is real. The momentum is genuine. But infrastructure alone does not build winners.
What actually determines whether India emerges from this moment as a platform builder or a sophisticated tenant is happening one layer above the data centres. It is happening in the decisions founders are making right now about what to build, who to build for, and whether to optimize for acquisition or category ownership.
The Question the Summit Raised But Didn't Answer
Vinod Khosla said the part everyone was thinking but few wanted to hear: traditional IT services and BPO are structurally finished by 2030. Not disrupted. Gone. The USD 250 billion engine that built India's modern middle class is being disassembled by the same AI wave creating the new opportunity.
Most people in that room knew he was right. The discomfort came from what follows.
The transition from services to platforms is not a pivot. It is a reinvention. The timeline is shorter than this industry admits. And the opportunity is not in the infrastructure layer everyone is announcing. It is in the application layer almost nobody is funding.
Infrastructure becomes a knife fight. Whoever has scale undercuts. Whoever has distribution bundles. Margins compress. Differentiation narrows. The durable white space remains at the application layer.
What India Can Actually Own
I left Delhi with a stronger conviction in two specific opportunities.
The first is consumer AI built for India's actual shape. Sam Altman noted India has over 100 million weekly ChatGPT users. That is not a threat. It is a demand signal. The appetite is enormous. The openness to behavior change is real.
What has been missing is the layer that earns delegation rather than experimentation. Users do not pay for cleverness. They pay when something reliably disappears from their mental load. Health management. Financial planning. Education continuity. Vernacular access that works at low bandwidth with cultural context built in. That product is now buildable in India, at Indian price points, in Indian languages.
The indigenous models launched at the summit – Sarvam's LLMs up to 105 billion parameters, BharatGen's Param2 supporting 22 Indian languages, Gnani's voice cloning across 12 languages – are production systems making the application layer viable for Indian consumer complexity.
A foundational model that works across Indian languages, at low bandwidth, with cultural context embedded, is not just domestic infrastructure. Every market that looks like India – most of the world – needs this. The model that succeeds here travels.
The second opportunity is physical AI. India's manufacturing cost advantage, hardware-literate founders, Make-in-India policy support, and supply chain alignment through Pax Silica compound. AI-native hardware design. Robotics for manufacturing. Edge intelligence for constraint-heavy environments. The founders building here are earlier than most investors think and further along than most say publicly. Indian constraints become product advantages. That is the bet.
The Capital Gap Nobody Is Talking About
India now has world-class infrastructure being built and world-class talent coming online, but the early-stage capital funding application companies is dangerously thin.
GCC investments are real. Data centre commitments are locked. Enterprise software gets funded every cycle. But consumer-facing platforms, workflow automation companies, AI-native category creators- the companies that will employ talent transitioning out of services and capture value users will pay for - are being underfunded or pushed to raise internationally before achieving local product-market fit.
This requires investors who understand the next decade of value creation will not look like the last. Winners will not be services plays repackaged as SaaS. They will be platforms that own workflows, capture behavioral data, and build proprietary feedback loops that compound over time.
The Generation That Changes Everything
The image I keep returning to is eight-year-old Ranvir Sachdeva on the keynote stage, talking about linking ancient philosophy to modern AI, then engaging with Sam Altman and Sundar Pichai as a peer.
His generation is native to AI. That is the real advantage: builders for whom India-first AI is not a constraint but a default, who see vernacular interfaces and low-bandwidth optimization as features, not bugs.
What Needs to Happen Next
The infrastructure is being laid. The models are being built. The sovereign positioning is happening. What is needed now is execution.
The platform layer – companies that own workflows, not headcount – is still largely unbuilt. The consumer layer is still largely unfunded. That is where the next decade of value gets created.
People do not pay for intelligence. They pay for outcomes they can trust. India now has the infrastructure to deliver those outcomes at scale. The question is who builds the products on top, and whether they get funded here or somewhere else.
The window is open. The timing is real. The question is whether India's ambition matches the moment.
Disclaimer: The views expressed are solely those of the author and do not necessarily reflect the official position of the publication. |
