A Scenario — 2026 to 2035
The US and China are racing toward artificial general intelligence. India is watching from the stands. This is a scenario — two possible futures — for what happens next.
For a thousand years, a pattern has repeated on the Indian subcontinent. A civilizational power — the wealthiest, most intellectually fertile society on Earth — failed to match the technological and organizational capacity of smaller, hungrier forces. The result was not defeat in a single battle. It was centuries of systematic extraction, first by Islamic sultanates, then by a British trading company that arrived as a partner and stayed as a sovereign.
Artificial intelligence is the defining technology of the next century. The US and China are building it. India is not even in the running. This scenario traces how that gap unfolds — and what it would take to close it.
This is not prediction. It is a scenario — one of many possible futures, written to be specific enough to be falsifiable, provocative enough to force a response. At the end, the path splits. You choose how it ends.
This scenario is directly inspired by AI 2027 — Daniel Kokotajlo, Eli Lifland, Thomas Larsen, and Scott Alexander's research-backed forecast of how AGI unfolds over the next few years. Read theirs first. This is India's chapter of the same story.
We used fictional company names to avoid singling out any one organization. "OpenMind" refers to the leading US frontier AI company. "DragonMind" refers to China's state-coordinated AI effort. "BharatMind" is India's would-be sovereign AI — existing, in 2026, mostly as aspiration. "Relion" refers to India's largest digital and telecom conglomerate. "Arkani Group" refers to India's largest infrastructure conglomerate.
Act I
The race is already underway. India celebrates a third-place ranking that masks a gap so large it has ceased to be a gap — and become a chasm.
In early 2026, India's government releases a glossy report celebrating the country's number three position on the Stanford Global AI Vibrancy Index. The headline travels fast on Indian news channels. What travels less far is the score: US: 78.6. China: 36.9. India: 21.6. Third place, yes. In the same race, no.
The United States alone will spend an estimated $660–690 billion on AI infrastructure in calendar year 2026 — more than India's entire defence budget multiplied by twelve. China, through state-directed capital and private megafunds, is spending another $70 billion. India's IndiaAI Mission, announced with fanfare, carries a five-year budget of $1.25 billion.
The silicon gap is physical. The US has roughly 14 million AI-grade GPUs. China has 9.5 million. India — the world's most populous nation, the "IT superpower" — has approximately 58,000. Every Indian AI startup, every government analytics initiative, every ambitious engineer in Bengaluru runs their models on American cloud infrastructure or, increasingly, on platforms quietly routed through Chinese-backed servers.
"The British didn't arrive with tanks. They arrived with a trading company. By the time India understood what was happening, the extraction was already structural."
Historical parallel, 1757Meanwhile, China's Digital Silk Road moves quietly but relentlessly. In 2026, Beijing signs AI cooperation frameworks with Pakistan, Bangladesh, Myanmar, Nepal, and Sri Lanka — embedding Chinese technical standards, Chinese surveillance systems, and Chinese data pipelines into India's immediate neighbourhood. India notices. India does not respond at the required scale.
The talent picture is equally sobering. India's universities produce extraordinary AI researchers — and then exports the best of them to OpenAI, DeepMind, Anthropic, and Meta. 45% of India's top AI talent works abroad. China retains 95% of its equivalent cohort. India is subsidizing the R&D budgets of its competitors with its own education spending.
AI Power Index — 2026
1 GW ≈ 100,000 H100 GPUs running continuously
To close the gap with China, India needs ~11 GW of sovereign AI compute. At current rates: 80 years.
Act II
OpenMind releases Agent-4. The world's largest economy begins to automate knowledge work at scale. India's IT sector — built on human cognitive labour — stares into an existential void.
By early 2027, OpenMind's Agent-4 achieves what researchers had called "software engineer parity" — the ability to complete any coding task a senior engineer could, at 30x the speed, for a fraction of the cost. (Benchmark models had already crossed 80% on standard coding tests in late 2025; Agent-4 is the first system to close the remaining gap on novel, unseen codebases.) DragonMind's comparable system follows within seven months, having benefited from a model weight acquisition that US intelligence describes — in classified memos — as "the most consequential technology transfer in history."
The global impact is immediate and brutal. The market for outsourced software development — a market India has dominated for two decades, generating $250 billion annually and employing 5.4 million people — begins contracting at 20% annually. Not because Indian engineers are bad. Because AI engineers don't need salaries, visas, or sleep.
Infosys, TCS, and Wipro collectively announce workforce restructuring affecting 340,000 employees over 18 months. Their framing: "AI-first transformation." The financial press calls it a pivot. Former employees call it something else.
The same week, OpenMind announces 200,000 "AI software agents" available for enterprise licensing at $8 per agent per hour. Tata Consultancy Services' average billing rate: $28 per human hour.
What India does not fully reckon with in 2027 is the second-order effect. India's IT sector was not just a source of foreign exchange. It was the country's primary pipeline for producing a technically sophisticated middle class — engineers, managers, and founders who would eventually build the next generation of Indian industry. That pipeline is rupturing.
India's response, in 2027, is to announce an "AI Services" industry — essentially positioning India as a premier destination for deploying and customizing foreign AI systems. Relion — India's largest conglomerate, controlling telecom, retail, and green energy under its digital arm — launches "JioSphere AI." It runs on Meta's open-source LLaMA architecture. The press release calls it a breakthrough. What it is, precisely, is a foreign model with an Indian brand painted on top.
Meanwhile, Arkani Group — the infrastructure behemoth that controls India's largest ports, airports, and power transmission networks — announces a $14 billion AI data centre investment across seven campuses. This is celebrated as India's entry into serious AI infrastructure. The details are less celebrated: the servers will run Nvidia chips on foreign supply chains, the cooling systems are engineered by American firms, and the primary clients are foreign hyperscalers looking for cheap land and power.
The pattern is identical to the diwan class of the Mughal and British eras — Indian intermediaries who prospered enormously by facilitating foreign power, accumulating personal wealth while the structural sovereignty of the civilization eroded around them. India is, in this moment, building hotels in a city whose roads, electricity, and water are all owned by someone else.
India's IT Sector — Crisis Indicators 2027
Every Indian user interacting with OpenMind's AI enriches OpenMind's model with Indian behavioral data, Hindi linguistic patterns, and Indian cultural context. India is exporting raw data so American companies can weave it into intelligence — then selling that intelligence back to India as a service.
In the 19th century, India exported raw cotton to Lancashire. Lancashire sold manufactured cloth back to India. The mechanism is identical. Only the commodity has changed.
Act III
The architecture of a new colonialism takes shape — not through conquest, but through contracts, APIs, and switching costs that compound quietly until they become inescapable.
By 2028, four distinct dependency chains have formed around India's digital economy, each reinforcing the others. Together, they constitute what historians will later describe as the period when India's digital sovereignty was quietly transferred — with India's enthusiastic participation.
Infrastructure: 75% of India's cloud workloads run on AWS, Azure, and Google Cloud. Migrating would cost Indian enterprises an estimated $18 billion and 3–5 years of disruption. Nobody migrates.
Data: India's agricultural yield data, patient health records, financial transaction patterns — the most intimate data about 1.4 billion people — flows daily into models trained in California. India has a data protection law. It has no mechanism to retrieve what has already left.
Platform: 90% of Indian government AI initiatives now run on foundations built by OpenMind, Google, or Meta's open models. The systems that assess tax compliance, flag welfare fraud, screen visa applications, and monitor state borders increasingly run on foreign intelligence with foreign supply chains.
Talent: 45% of India's top AI talent is abroad. Those who return increasingly join Indian subsidiaries of American and Chinese firms — technically in India, strategically elsewhere.
The power dimension of this dependency is rarely discussed in India's English-language press, but it is the most revealing. The US dedicates 33 GW of electricity to AI compute — approximately the output of 25 large nuclear power plants, running continuously, purely to power AI training and inference. China: roughly 11 GW and rising at 40% annually. India's dedicated AI compute draws approximately 0.4 GW — less than a single mid-sized Google data centre campus in Iowa. Arkani Group's announced 7 GW of data centre capacity sounds impressive until you learn that most of it is contracted to foreign hyperscalers seeking cheap land and subsidised power — not building Indian AI. India is becoming the landlord for the data centres of its competitors.
In October 2028, India's Ministry of Defence tables a classified assessment. Its finding: in the event of a major geopolitical crisis, India's military decision-support systems — logistics, target assessment, satellite analysis — carry a meaningful dependency on foreign AI infrastructure components. The assessment recommends immediate action. It is marked "restricted." It does not generate immediate action.
The year ends with a number that should be burned into the consciousness of every Indian policymaker: ₹4.8 lakh crore ($58 billion) in annual digital economy value is effectively accruing to foreign platform owners. Not through unfair trade. Through agreements India signed freely, because the alternative — building it yourself — seemed too expensive and too slow. It always does, until it doesn't.
Dependency Depth — 2028
| Domain | Foreign Dependency | Trend |
|---|---|---|
| Cloud Infrastructure | 75% | ↑ Rising |
| Foundational AI Models | 90% | ↑ Rising |
| GPU / Compute Hardware | >95% | ↑ Rising |
| Semiconductor Chips | 95% | → Flat |
| Defense AI Systems | ~40% | ↑ Rising |
| Government AI Platforms | 90% | ↑ Rising |
By end 2028, Chinese AI infrastructure (hardware, platforms, governance systems) is the primary framework in Pakistan, Bangladesh, Myanmar, Nepal, and Sri Lanka. The Maldives has signed a DragonMind AI partnership. Bhutan is negotiating one.
India is surrounded — not by armies, but by dependency architectures that point toward Beijing. Every data centre built by Huawei in Dhaka is a node in a network that treats India as a geopolitical problem to manage, not a partner to serve.
2029 — The Fork in the Road
The dependency chains are visible. The encirclement is documented.
The GDP cost is calculable. The historical parallel is unmistakable
to anyone willing to look.
The question is whether India — its government, its industry, its citizens —
responds at the scale the moment demands.
Choose the path India takes.
Bear Case
The New
Dependency
India continues on its current trajectory. Good intentions, incremental funding, and the wrong bets.
Bull Case
The Third
Pole
India declares AI sovereignty a national emergency and mobilises at a scale matching the threat.
Bear Case — 2029 to 2035
India chose incrementalism. History, as it has before, chose India.
Facing an economy with 1.2 million IT workers displaced and an AI sector generating revenue almost entirely for foreign shareholders, India's government negotiates a landmark deal: the US–India AI Framework Agreement. The terms, dressed in the language of partnership and co-investment, read differently on close examination.
India receives preferential API access to OpenMind models. Priority compute allocation during non-peak hours. A joint research center in Hyderabad — staffed largely by Indian researchers building American intellectual property on American infrastructure. In exchange: India agrees to data localisation carve-outs for American firms, alignment with US AI governance standards, and a pledge not to subsidise "competing" Indian foundational model efforts classified as creating "market distortions."
By 2031, every country bordering India except Bhutan is running Chinese-built AI infrastructure at the state level. DragonMind's governance AI — processing judicial recommendations, social credit inputs, and predictive policing models — operates in Dhaka, Islamabad, Kathmandu, and Naypyidaw. The Colombo Port City, financed by Chinese capital and governed by Chinese-standard smart city systems, has become a data node with direct pipelines to Beijing.
India's intelligence services flag the architectural implication: any future regional crisis will involve adversaries with AI systems that have been trained on decades of South Asian data, in languages India's own AI systems still struggle to parse fluently. India is now fighting a potential war with tools its adversaries helped design.
An internal Ministry of External Affairs document, leaked to Indian media, contains a single paragraph that becomes India's most-shared political text of the decade:
"In the event of armed conflict with a major neighbouring power, India's decision-support systems, logistics AI, and satellite analysis platforms carry estimated foreign dependency components of 35–60%. The structural nature of this dependency means it cannot be resolved within a conflict timeline. We are, in the relevant sense, not self-sufficient in the tools of modern warfare."
By 2035, the structural extraction has become self-sustaining. India's digital economy generates ₹38 lakh crore ($460 billion) annually. Of this, an estimated 12–15% flows to foreign platform owners as API fees, cloud margin, licensing, and data rents — a permanent tax on Indian economic activity paid to foreign sovereigns.
Indian languages have degraded in frontier AI systems. The models trained on English, Mandarin, and Spanish have no commercial incentive to achieve true fluency in Tamil, Telugu, or Odia. India's cultural and linguistic diversity — the raw material of its civilizational identity — is underrepresented in the systems that increasingly mediate how Indians access information, services, and opportunity.
A parliamentary committee report, tabled in late 2035, draws a comparison nobody in the government wants to hear but none can honestly refute: India's digital economy in 2035 bears structural resemblance to its agrarian economy in 1850 — highly productive for the primary extractor, generating local employment but not local wealth, dependent on external capital for every meaningful upgrade, and offering no plausible exit without a disruption equivalent to Independence.
Bull Case — 2029 to 2035
India chose urgency. What followed was the most consequential industrial mobilisation in the country's modern history.
In March 2029, India's Prime Minister addresses a joint session of Parliament. The speech is 47 minutes long. It contains no euphemism. It acknowledges, with an explicitness that stuns the diplomatic community, that India is on the verge of a structural dependency that will take a generation to reverse if it is not arrested now. The speech is called the "AI Swaraj address" — invoking Gandhi's demand for self-rule not as nostalgia but as a living imperative.
1. AI Sovereign Fund: ₹5 lakh crore ($60B) over 10 years, ring-fenced from the Union Budget, for compute, talent, and foundational model development.
2. National AI Service Act: India's top 2,000 AI researchers — wherever in the world they work — are offered a "National AI Service" package: ₹5 crore tax-free return grant, world-class facilities, and equity in national AI projects.
3. BharatCompute Authority: Target of 2 million sovereign GPUs (~20 GW of dedicated AI compute capacity) within 36 months. No foreign cloud provider permitted to handle classified or critical-infrastructure data.
4. BharatMind Initiative: Ten competing foundational model projects, each funded at ₹2,000 crore. First competitive frontier model target: 36 months.
5. IndiaAI Global Alliance: Offer India's AI capabilities, infrastructure, and governance framework to 50 partner nations — positioning India as the democratic, non-predatory alternative to US and Chinese AI dominance.
The global reaction is immediate and divided. OpenMind's stock drops 3% on fears of a lost market. China's Foreign Ministry issues a statement calling India's data sovereignty provisions "protectionist." These reactions are, in their own way, confirmation that India has finally done something that matters.
The first 18 months are brutal. Building a frontier AI lab from near-scratch, in a country whose best AI talent has spent the last decade working in California and London, is not elegant. The early BharatMind prototypes are embarrassing by Silicon Valley standards. The engineers know it. They build anyway.
What changes the trajectory is not a single breakthrough but a culture shift. The National AI Service Act brings back 8,400 researchers in its first year — more than expected, because many had been waiting for exactly this signal: that India was serious. The IIT-AI campuses in Bengaluru, Hyderabad, and Chennai begin operating at a standard that can recruit internationally. For the first time, some researchers choose India over MIT.
The IndiaAI Global Alliance, launched in 2029 as aspiration, has by 2032 become a genuine multilateral institution. 62 nations — primarily across Africa, Southeast Asia, South America, and the Middle East — have signed partnership agreements. The terms are notably different from China's model: no equity in critical infrastructure, no governance system dependencies, no embedded surveillance architecture. India offers capability and asks for cooperation.
BharatMind-2, released in early 2033, achieves what independent benchmarks describe as "genuine competitive performance" with the leading American models — particularly in multilingual tasks, medical diagnosis in low-resource settings, and agricultural yield prediction for tropical climates. It is the first Indian technology product that the world's most sophisticated users choose not because it is Indian, but because it is better for their specific needs.
India's Semicon 2.0 program, running since 2029, achieves its first 7nm fabrication run at the Dholera campus — four years ahead of the schedule that critics called "hopelessly optimistic." The breakthrough is attributed to a technology transfer partnership with a Taiwanese foundry, negotiated in exchange for India's strategic neutrality guarantees — and the credible alternative India could now offer China's Belt and Road digital partners.
India is not yet at the frontier of semiconductor fabrication. But it has broken the most dangerous dependency: the assumption that it never could be.
The phrase "Third Pole" enters the lexicon of international relations in 2034, when a Foreign Affairs essay notes that the AI world order — assumed since 2025 to be a bipolar US-China contest — has a third centre of gravity. India does not control the frontier. It does not need to. What India controls is something more durable: the trust of the non-aligned world in a digital era where alignment means data dependency, surveillance infrastructure, and platform lock-in.
India's AI sector reaches $420 billion in annual revenue by 2035. BharatMind-class models are used operationally in 74 countries. India sets the AI governance agenda at the UN's AI Safety Council — not because it has the most powerful models, but because it represents the largest coalition of nations that have chosen neither Silicon Valley nor Beijing as their digital sovereign.
None of this was inevitable. It required a political decision, made in 2029, to treat the threat with the seriousness it deserved. It required a generation of engineers who chose to stay and build rather than leave and build for others. It required India to remember what it once was, understand clearly what it risked becoming, and decide — without ambiguity — that those were not the same thing.
This scenario was written in 2026. The data in it is real. The investments cited are real. The dependency chains described are forming right now. The historical parallels are not metaphors — they are structural descriptions of the same mechanism operating in a new medium.
The choice described at the fork point is not fictional. It is the choice India's government, industry, and citizens are making today — through budgets, through talent policy, through procurement decisions, through what they choose to build and what they choose to buy.
Civilizations are not conquered. They are abandoned by their own complacency.
Sources: Stanford AI Vibrancy Index 2025 · Futurum AI Capex Report 2026 · NASSCOM IT Sector Report · World Financial Review: Economic Drain from Colonial India · Carnegie Endowment for International Peace: China's Digital Silk Road · Digital Republic: India AI Ambitions · Angus Maddison: Economic History of India
Hi, I'm Rahul Rai.
IIT Bombay dropout. Wharton undergrad. I previously co-founded a $100M+ quant hedge fund and a Google-backed edge AI startup.
Outside of markets and technology, I love reading about history, Indian mythology and Buddhist philosophy. The longest lenses give the clearest views.
Follow me on X / Twitter or reach out at rahulplus@gmail.com.