Amid apprehensions of an AI bubble, fueled by rampant geopolitical contestation, there is an urgent need to balance power-driven Big AI with purpose-driven Small AI
India is hosting the 2026 AI Impact Summit amid growing global fragility and intensifying strategic competition. Technology has become one of the key pillars supporting this fractured global order. Artificial Intelligence (AI)—the field of study focused on machines capable of performing tasks that typically require human-level intelligence—is the most prominent and increasingly pervasive technological paradigm today. By 2029, nearly US$ 3 trillion is projected to be invested in the infrastructure supporting it, including data storage, computational capacity, and energy requirements. Yet, substantial obstacles and uncertainties remain regarding the global realisation of AI’s socio-economic benefits. Many of these and the structural causes behind them are expected to take centre stage at the Summit.
Despite exponential advances in AI, claims that ‘Artificial General Intelligence’ (AGI) is imminent are highly misleading.
Despite exponential advances in AI, claims that ‘Artificial General Intelligence’ (AGI) is imminent are highly misleading. In October 2025, over 30 eminent AI experts from across disciplines—including Eric Schmidt and Yoshua Bengio—came together to develop a rigorous definition of AGI. They defined it as an AI “matching the cognitive versatility and proficiency of a well-educated adult”. Their framework breaks human intelligence into ten core cognitive domains—including reasoning, memory, and perception—and adapts established human psychometric tests to evaluate AI systems. Applying this framework to contemporary models reveals a highly “jagged” cognitive profile, as shown in Figure 1. While these systems perform strongly in knowledge-intensive domains, they exhibit critical deficits in foundational cognitive capacities, particularly long-term memory storage.
Figure 1: A framework dissecting general intelligence
Source: Center for AI Safety (CAIS), A Definition of AGI
In contrast, OpenAI CEO Sam Altman reportedly signed an agreement with Microsoft (which owns about 27 percent of OpenAI) in 2024 defining AGI’s attainment as OpenAI delivering US$ 100 billion or more in profits. Even under this profit-based benchmark, with OpenAI projecting merely US$ 13 billion in 2025 revenue against an investment scale of US$ 1 trillion, the realisation of AGI remains a long-term prospect.
Significant attention has focused on the competitive dynamics among the world’s leading Large Language Models (LLMs)—including their leaderboards—and the companies and nations that develop and deploy them. Given the broader global dynamics highlighted earlier, the insatiable demand for computational capacity to power these LLMs has emerged as the latest frontier in the geopolitical race over AI. This dynamic underscores the enduring relevance of the Westphalian concepts of sovereignty, democracy, rule of law, and territoriality. Within the AI compute supply chain, high levels of international market concentration and vertical integration prevail (Figure 2). For example, US corporation NVIDIA’s chips account for roughly 80–95 percent of the global market, with US hyperscale cloud providers representing 40 percent of NVIDIA’s revenue and 70 percent of the global cloud computing market.
Figure 2: Concentration in the global compute market
Source: Epoch AI
India’s large and diverse market, adaptable and affordable engineering talent, English-proficient workforce, and evolving regulations on data sovereignty and digital infrastructure make it an attractive destination for international corporations. However, despite significant investments by US Big Tech firms in India towards the end of 2025—totaling US$ 67.5 billion across Google, Microsoft, and Amazon—recent remarks by US President Donald Trump’s Trade Advisor Peter Navarro, asking “Why are Americans paying for AI in India?”, highlight concern that this external demand could drive up domestic electricity costs in the US, with warnings of potential ‘strong action’ from the Trump administration. This follows the Biden administration’s January 2025 decision to categorise India outside its Tier 1 list of partner countries eligible for the most advanced AI chips produced in the US. Further, amid a challenging year for India–US ties, India was initially excluded from the first Pax Silica Summit convened by the US on 12 December 2025.
While making rapid progress across its “five-layer AI stack” with 12 indigenous models aimed at a sovereign, inclusive, and use-case-focused ecosystem, India still needs to scale up its core energy and data infrastructure.
These instances underscore that, as a rising power from the Global South committed to defending its strategic autonomy, India will need to earn its place in global supply chains and rule-setting coalitions. To do so, it needs sustained, mission-mode investments in its high-tech sector—particularly advanced semiconductor chips, AI compute capacity, and model deployment. Currently, India’s approximately US$ 1.2 billion flagship IndiaAI Mission signals serious intent, but this is roughly equivalent to what a single US corporation like OpenAI spends in six months. While making rapid progress across its “five-layer AI stack” with 12 indigenous models aimed at a sovereign, inclusive, and use-case-focused ecosystem, India still needs to scale up its core energy and data infrastructure.
Small AI is built on an alternative approach that avoids front-loading massive investments in computational infrastructure—such as energy, data centres, cutting-edge chips, and state-of-the-art (SOTA) foundational models—which, as discussed above, tends to concentrate markets and widen the global divide in AI capability and innovation. Instead, it leverages resource optimisation and efficiency derived from economies of scale. At the upcoming AI Impact Summit in New Delhi, one of the seven working groups—Working Group 6 on “Democratizing AI Resources”—aims to explore this opportunity. Its objectives are to: a) promote the accessibility and affordability of AI resources as global public goods, b) facilitate international cooperation to build distributed AI infrastructure and foster open innovation, and c) support capacity-building and knowledge exchange to strengthen local AI ecosystems. The latest India Economic Survey, released at the end of January 2026, echoes this approach, cautioning that India should prioritise “application-led innovation, the productive use of domestic data, human capital depth, and the ability of public institutions to coordinate distributed efforts.” In addition, the Survey notes, “A bottom-up strategy anchored in open and interoperable systems, sector-specific models, and shared physical and digital infrastructure offers a more credible pathway to value creation than a narrow pursuit of scale for its own sake.”
AI alarmists, including veteran technologist Eric Schmidt, former CEO of Google (2001–2011), have long warned about the potential harms of the current trajectory of AI development and deployment. Having moved beyond the early optimism that AI could reduce global inequality, accelerate scientific breakthroughs, and democratize knowledge, Schmidt now actively advocates for guardrails to manage the risks posed by ‘rogue AI’. In particular, he and University of Montreal Professor Yoshua Bengio, the 2018 Turing Award winner, have highlighted concerns around autonomous agents. They note that advanced AI models, which underpin tools like chatbots and agents, are increasingly exhibiting signs of self-preservation through “in-context scheming”, in which an AI covertly pursues misaligned goals. Recent research demonstrating such scheming lends concrete weight to these calls for caution. The second edition of the International AI Safety Report, led by Yoshua and released a few days ago, is a much-anticipated discussion in this regard.
With the right kind of stewardship, it has the potential to become a ubiquitous tool that augments the next generation of international, societal, and human progress—what might be called intelligent progress.
The launch of ChatGPT in 2022 is widely recognised as the start of an ‘AI boom’, generating immense global attention across technological, economic, political, and social spheres. Many observers view this surge—characterised by unprecedented market and institutional investment—as reminiscent of the dot-com era, when everyone rushed to stake a claim on the Internet. While the subsequent bust tempered expectations, the Internet as a technology paradigm and its underlying infrastructure did not lose relevance; if anything, they became an everyday utility, much like electricity before it. Today, the global focus is on riding ‘the AI wave’. AI is inherently more complex, and its trajectory remains uncertain. However, with the right kind of stewardship, it has the potential to become a ubiquitous tool that augments the next generation of international, societal, and human progress—what might be called intelligent progress.
Rahul Batra is a geopolitical analyst with extensive experience at the intersection of digital platforms and international affairs.
The views expressed above belong to the author(s). ORF research and analyses now available on Telegram! Click here to access our curated content — blogs, longforms and interviews.
Rahul Batra is an independent consultant with extensive experience at the intersection of digital platforms and international affairs. He spent many years across Google’s global ...
Read More +
PREV
