Here is a number that does not get the attention it deserves: over 12 billion transactions per month. That is UPI in 2024. Not a pilot. Not a “digital payments initiative.” Twelve billion monthly transactions running on infrastructure that Indian engineers designed, built, and scaled to population level — in a country where two decades earlier, a majority of the population was unbanked.
India did not build a payments app. India built a payments rail. The distinction matters for everything that follows.
PRED-011 — India will rank number two globally in agent infrastructure company revenue by December 2030 — behind only the United States. India will surpass both China and the EU in total revenue from agent infrastructure companies: orchestration platforms, trust and attestation services, agent identity systems, monitoring tools, and inter-agent communication protocols.
Confidence: 3 out of 5.
The chapter builds the arithmetic case. Over 17 million developers on GitHub as of 2024 — the second-largest developer community in the world and the fastest-growing, per GitHub’s Octoverse 2024 report, which projects India will surpass the United States by 2028. India built UPI at 12 billion-plus transactions per month. India built Aadhaar — the world’s largest digital identity system — and identity is one of the five components of the dependency layer. India built ONDC, an open commerce protocol demonstrating protocol-level infrastructure capability. Indian startups run on burn rates three to five times lower than Silicon Valley equivalents — structural profitability at revenue levels where American competitors are still burning cash.
The arithmetic is real. But arithmetic is not why I believe PRED-011.
I believe it because of the infrastructure instinct.
Indian developers have a disproportionate historical bias toward infrastructure problems over consumer applications or model training. This is not a cultural preference expressed in surveys. It is a pattern observable in what India actually ships at scale. UPI is a payment rail, not a payment app. Aadhaar is an identity protocol, not an identity product. ONDC is a commerce protocol, not a marketplace. Freshworks, Zoho, and Postman are infrastructure for other businesses to run on. The pattern is consistent across two decades: when Indian engineers build at population scale, they build the plumbing.
That same orientation — solve the infrastructure problem first, then let applications grow on top — is the orientation most likely to produce the agent trust layer, the agent identity system, the inter-agent communication protocol. Not because India has more developers (it does, but that is the arithmetic argument). Because the instinct to build the rail instead of the app is culturally embedded in how Indian engineering teams operate.
I see this in my own hiring and vendor decisions. When I evaluate Indian engineers for infrastructure roles across my ventures, the default mental model is “build the platform, not the feature.” When I evaluate engineers from ecosystems oriented toward consumer applications, the default is inverted. Neither instinct is better in the abstract. For agent infrastructure specifically, the Indian instinct is the right one.
This prediction is explicitly not about AI model training. India will not outcompete the United States or China on building the largest language models. The capital requirements, the compute access, and the research talent concentration all favour the US and China for foundation models. PRED-011 is about the layer beneath the models — the plumbing that agents need to function, verify, communicate, and be trusted. That is a software problem. India went from IT services to SaaS products in fifteen years, producing Freshworks, Zoho, and Postman. The next evolution — from SaaS products to agent infrastructure — aligns with India’s demonstrated strength: building reliable systems at population scale, with the world’s fastest-growing developer talent pool, at structural cost advantages that compound over time.
The instinct and the economics point in the same direction. That is why the confidence is 3, not 2.
The published falsification trigger:
If by December 2030, India’s share of global agent infrastructure revenue is below 5%, or if fewer than three Indian agent infrastructure companies reach $50M+ in annual recurring revenue, this prediction is wrong.
The realistic failure mode is not a lack of talent or cost advantage. It is a routing problem. If Indian founders overwhelmingly build agent applications — chatbots, vertical SaaS agents, consumer AI — instead of agent infrastructure, the instinct thesis is wrong and PRED-011 reverts to a weaker cost-arbitrage argument. Cost arbitrage alone does not get you to number two globally. The instinct has to be real.
If you are at an Indian AI startup — or an India-based team inside a global company — I want to know: does your current roadmap lean toward infrastructure or applications? Infrastructure means: orchestration, trust, identity, monitoring, inter-agent communication, developer tools for agents. Applications means: vertical agents, chatbots, consumer-facing AI products.
One line. Infrastructure or applications. Send it to me.
I will publish a distribution on the public PRED-011 tracking page at atin-agarwal.com/predictions/pred-011-india-agent-infrastructure/, updated quarterly as data comes in. If a clear majority of Indian AI startups surveyed are building consumer and application-layer agents rather than infrastructure, the instinct thesis is wrong. I want to know.
If you are an Indian founder: the infrastructure lane is the one that matches both your cost structure and your historical strengths. The application lane is crowded with global competitors who have distribution advantages you do not. The infrastructure lane has fewer competitors, higher defensibility, and aligns with what Indian engineering teams have proven they can build at scale. Pick the lane that compounds.
If you are a non-Indian founder building on agent infrastructure: the India ecosystem you compete with in 2029 is infrastructure, not applications. Your orchestration vendor, your attestation provider, your monitoring stack — the companies building those layers will increasingly be Indian. Evaluate accordingly.
If you are an Indian investor: the infrastructure thesis is an alpha source. The global investor consensus on Indian AI is still pattern-matching to consumer SaaS and IT services. The infrastructure play is under-covered and under-funded relative to the opportunity. Three to five Indian agent infrastructure companies at $50M-plus ARR by 2030 is the benchmark. Start funding for it now.
This issue is drawn from Chapter 9 of The AI Agent Economy — 15 falsifiable predictions with dates, numbers, and explicit triggers for being proven wrong. Pre-order on Kindle — $9.99. Release July 1, 2026. atin-agarwal.com/books
Read the full PRED-011 entry on the public tracking page → atin-agarwal.com/predictions/pred-011-india-agent-infrastructure/
Previous issue: Issue 10 — The $50B trust-layer company will be acquired before it IPOs
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.