When a company's market value crosses $3 trillion, the temptation is to treat the number as spectacle. Amazon's crossing of that threshold, reported by NDTV on August 3 after shares surged five percent, warrants closer examination. The profits that drove the rally — more than $62 billion in a single quarter — did not come from selling books or same-day groceries. They came from two artificial intelligence divisions: cloud computing and proprietary AI chips. That is a fact about the geography of the AI economy, not just about one company's balance sheet.
The geography runs like this. A small number of US technology firms — Amazon, Microsoft, Google, Nvidia — now control the compute infrastructure on which the rest of the world's AI ambitions depend. They own the data centres, the chip architectures, the API layers, the foundational models. Every government running a smart-city pilot, every hospital deploying diagnostic AI, every startup building a language model for a non-English market, pays rent to this infrastructure. Amazon's Trainium and Inferentia chips, designed in-house to reduce dependence on Nvidia and run AI training workloads at scale, are the latest expression of this vertical integration. The profits are a symptom; the structure is the story.
The Compound Engine
What makes Amazon's $3 trillion crossing significant is not the valuation per se but what produced it. AI cloud revenue and chip sales are not ancillary to Amazon's business — they are now its primary profit engine. The e-commerce and logistics operations that made Amazon a household name are, in earnings terms, increasingly the loss-leader that keeps Prime subscribers inside the ecosystem. AWS, the cloud division, cross-subsidises everything else while generating the margins that Wall Street rewards. When those margins are further amplified by proprietary AI chips — cutting Amazon's own inference costs while selling compute to external customers — the compounding effect becomes geometric.
This matters for any economy trying to build AI capacity from outside the US hyperscaler club. The more Amazon internalises its AI chip supply chain, the higher the barriers to entry for alternative compute providers. A government that wanted to build sovereign AI infrastructure in, say, 2018 faced a different competitive landscape than one attempting the same in 2026. The gap has widened by a $3 trillion market cap and sixty-two billion dollars in quarterly profits.
India's Asymmetric Position
AWS has committed $12.7 billion in India cloud infrastructure investment through 2030, making it one of the largest single technology investment pledges in the country's history. The investment funds data centres, local availability zones, and the compute backbone that powers everything from government portals to fintech unicorns. Amazon's India logistics network employs hundreds of thousands directly and indirectly. By any conventional measure, Amazon is a committed India partner — its cumulative investment commitment in the country runs to $26 billion.
But commitment and dependency are not opposites; they often reinforce each other. The same infrastructure investment that builds Indian digital capacity also deepens India's structural reliance on a single foreign hyperscaler for AI workloads. When a government ministry deploys an AI-powered citizen service on AWS, it is not merely using cloud storage — it is running its inference workloads on Trainium chips, calling APIs built on Amazon-trained models, and storing the resulting data in Amazon-managed availability zones. Each layer of that stack is owned, priced, and ultimately controlled from Seattle.
India's Ministry of Electronics and IT has articulated this concern without naming it so bluntly. The IndiaAI Mission, backed by a ₹10,371 crore outlay, seeks to build domestic compute capacity — GPU clusters procured for Indian researchers and startups, foundational model development, and a compute-sharing framework that gives smaller players access to hardware they could not individually afford. If India does not own some meaningful share of the AI compute stack, the digital sovereignty assertions made at the Voice of Global South summits remain aspirational rather than structural.
The Valuation Gap as Diagnostic
Here is where Amazon's milestone becomes most instructive. The market capitalisation of India's largest IT services firm, Tata Consultancy Services, runs to roughly $160 billion — a figure that represents decades of building the world's most efficient software-delivery machine, employing over six hundred thousand people, and generating consistent free cash flow. Amazon, on the day it crossed $3 trillion, was worth nearly nineteen times that. The comparison is not meant to diminish TCS or the Indian IT sector; it is meant to identify the fault line precisely.
Indian IT built its global position by arbitraging labour costs against Western demand for application development and maintenance. That model worked brilliantly for three decades. What Amazon's profit structure reveals is that the next layer of value — the infrastructure on which applications run, the chips on which models train, the APIs through which intelligence is accessed — does not yield to labour arbitrage. It yields to capital concentration and chip design. India has deep software talent. It does not yet have a domestic hyperscaler, a competitive AI chip programme, or foundational model infrastructure at the scale that Amazon, Microsoft, and Google have built.
Nasscom has flagged this risk repeatedly: Indian IT firms risk commoditisation unless they embed AI capabilities at the platform level rather than offering labour-intensive integration services around foreign AI stacks. Amazon's quarterly profits validate that warning with arithmetic. If the margin in AI accrues to whoever owns the chip and the model, not to whoever deploys the application, then India's IT sector faces a structural squeeze — good revenue, thin margins, no pricing power — unless it moves up the stack.
Managing the Partnership Without Surrendering the Leverage
The correct response is neither to reject Amazon's investment nor to passively accept the infrastructure dependency it creates. India sits in a rare position: it is simultaneously Amazon's largest growth frontier outside North America and a market large enough that losing it would materially damage AWS's global expansion story. That asymmetry — India needs Amazon's compute; Amazon needs India's scale — is leverage, and it should be used deliberately.
The IndiaAI Mission's procurement strategy becomes the instrument. Analysts at institutions like the Takshashila Institution have argued that India's compute procurement must combine AWS and Azure capacity leasing with domestic chip design incentives — not one or the other. The goal is not autarky, which would be both expensive and counterproductive. The goal is a negotiating floor: a defined proportion of government AI workloads running on India-owned or India-operated infrastructure, combined with technology transfer clauses and training partnerships as conditions for continued market access expansion. Amazon's lobbying weight in Washington, and its genuine strategic interest in the Indian market, make it a tractable partner for exactly this kind of structured conversation.
India's Voice of Global South articulation — that AI governance must not replicate the asymmetries that disadvantaged the developing world in earlier technological transitions — is the right framing at the multilateral level. But framings require backing infrastructure. The $3 trillion moment quantifies, in market terms, what it costs to cede the AI infrastructure layer entirely to foreign capital. The IndiaAI Mission's ₹10,371 crore outlay is not Amazon-scale money. It does not need to be — it needs to be enough to establish the credible capability that turns India from a pure customer into a partner with genuine optionality. That is a different target, and it is achievable.
