India’s sovereign AI and the perilous architecture of Nvidia’s boom

The modern gold rush is not taking place in physical mines, but within the humming, temperature-controlled aisles of hyperscale server farms. Nowhere is this shift more evident than in India, where the National Payments Corporation of India (NPCI) has recently partnered with Nvidia to construct a sovereign AI model explicitly tailored for the country’s vast digital payments ecosystem. Yet, this ambitious project is merely one node in a sprawling, highly lucrative, and potentially fragile web of investments underpinning the global AI boom—a boom of which Nvidia is the undisputed architect.

India’s digital payments infrastructure already operates at a population scale where performance and resilience are paramount. The NPCI’s collaboration with Nvidia marks a strategic pivot away from fragmented, use-case-specific AI tools. Instead, the goal is to build a scalable, payments-native foundational layer designed for high-volume, low-latency environments.

• NPCI will leverage the Nvidia Nemotron family of open models, incorporating open weights, training data, and recipes into its development approach.
• The initiative is strictly aligned with India’s regulatory and data sovereignty requirements.
• By combining accelerated computing with operational intelligence, the platform aims to support trust frameworks and grievance redressal while maintaining data security.

As Vishal Kanvaty, NPCI’s Chief Technology Officer, notes, the objective is to create a sovereign foundation that bolsters trust and security within the broader ecosystem. This is not merely about making payments smarter; it is about treating artificial intelligence as critical national infrastructure.

If AI is the new infrastructure, Nvidia is effectively the sole provider of the concrete. The chipmaker’s latest quarterly earnings released this February demonstrated an astonishing immunity to recent market fears regarding an AI bubble.

• Nvidia posted an enormous total profit of US$120bn for the fiscal year.
• Quarterly revenue reached US$68.13bn, comfortably exceeding analysts’ predictions of US$66.2bn.
• The core of this financial engine—the data center business—experienced a 75% year-over-year growth, hitting US$62.3bn for the quarter.

The sheer momentum is staggering. CEO Jensen Huang recently highlighted that customers are racing to invest in AI compute, describing these data centers as the “factories powering the AI industrial revolution”. The company commands an estimated 90% share of the AI chip market. Furthermore, Nvidia has secured over US$500bn worth of orders for its current Blackwell processors and the upcoming Rubin graphics processing units. If data center capital spending grows as anticipated, Nvidia’s revenue from this segment alone could soar by 165% to nearly US$450bn by 2027.

However, beneath the eye-watering revenue figures lies a complex financial architecture that warrants strict scrutiny. An increasingly interconnected web of dependencies has formed between technology manufacturers and AI startups.

• Nvidia has emerged as a key financial player, investing heavily in leading AI startups like OpenAI, xAI, and Mistral, which in turn rely on Nvidia’s chips.
• The company also took a 7% stake in CoreWeave, a “neocloud” provider, and later agreed to buy $6.3bn worth of cloud services from the firm.
• These “circular deals” effectively bind the fortunes of these businesses tightly to one another.

While industry leaders defend these interconnected arrangements as the necessary approach to financing a technological revolution, skeptics see troubling historical parallels. The current dynamic echoes the telecom boom of the late 1990s, where equipment makers fueled relentless expansion through vendor financing. When demand fell short of forecasts, the model broke, leaving heavily leveraged carriers bankrupt and capacity severely underused.

If revenue from AI products fails to grow as fast as today’s lofty expectations demand, the risks of these circular deals will be laid bare. A company might find itself staring at untenable bills for data center capacity, while the chipmaker loses twice: the customer stops buying its products, and the value of its equity stake tumbles.

For now, the flywheel spins ever faster, powered by sovereign ambitions in India and an insatiable global appetite for computing power. Yet, as the lines between investor, supplier, and customer continue to blur, the AI industry’s foundations may be more intertwined—and potentially more fragile—than its backers care to admit.