On the second day of India’s AI Impact Summit in New Delhi, a robotic dog trotted across the exhibition floor. It had been introduced on national television as a product of Indian ingenuity, a symbol of the country’s arrival as a “principal AI system-builder”. Within hours, social media users had identified it as a Unitree Go2—manufactured in Shenzhen and retailing globally for roughly Rs 200,000 (approximately £1,625). The robot was Chinese. The symbolism was not.
The AI Impact Summit, held from February 14th to 16th and extended through the 21st, has been designed to showcase India as a rising artificial intelligence (AI) power. Nearly 20 world leaders attended. Sundar Pichai, Sam Altman and Dario Amodei appeared on stage. The government cited the Stanford Global AI Vibrancy Tool, which ranks India third globally behind the US and China. The underlying message is clear: India has moved beyond aspiration.
Yet the episode of the imported robodog was not merely a public-relations mishap. It was an inadvertent metaphor. India’s AI ambition is vast. Its institutional, financial and infrastructural base remains narrower. The country excels at deploying technology built elsewhere. It has not yet demonstrated that it can build frontier systems at scale. The gap with the US is obvious. The gap with China—closer, geographically and strategically—may prove more consequential. That distance is structural, not cosmetic. No summit can compress it.
India’s AI Impact Summit: Theatre and technological state
Indian Prime Minister Narendra Modi has long understood the politics of spectacle. From the Central Vista redevelopment to the abrupt demonetisation of 2016, his government has paired grand announcements with national choreography. The AI Impact Summit followed that pattern. The staging has been meticulous. The language was expansive. India, the consul-general in Shanghai wrote in an op-ed, was emerging as one of the world’s leading AI “system-builders”.
The term invites scrutiny. What constitutes a system-builder? In AI, it typically implies the capacity to design and train large foundation models, manufacture or procure advanced chips, operate hyperscale data centres, and sustain an ecosystem of research and venture capital capable of iterative innovation. It implies ownership over the core stack, not merely fluency in its application.
India’s strengths lie elsewhere. It has produced generations of software engineers and technology executives. It operates digital public infrastructure—Aadhaar, UPI and DigiLocker—at an extraordinary scale. It has shown unusual administrative creativity in deploying digital tools for welfare distribution and payments. These are real accomplishments. They do not yet amount to control over the frontier of AI development.
Third place, with context
The Stanford ranking, cited repeatedly at the AI Impact Summit, deserves both acknowledgement and proportion. India scores 21.59 on the Global AI Vibrancy Tool. The US scores 78.60. China scores 36.95. India is indeed third. It is also closer to the fourth-ranked countries than to China, and dramatically distant from the US.
The distinction matters. Third place can suggest podium status. In this race, it indicates membership in a secondary tier. India’s relative strength derives largely from AI deployment, talent and policy signalling rather than from model-building or compute dominance. It is more adept at using AI than at shaping its direction.
That gap is not shameful. It is developmental. But it complicates the narrative of arrival.
Deployment vs production
India’s AI story is strongest where the technology is applied. UPI processes billions of digital payments monthly. Aadhaar has enrolled over a billion residents. Bhashini seeks to translate India’s linguistic diversity into machine-readable form. These systems provide a fertile substrate for AI-driven services.
Yet most Indian AI firms rely on fine-tuning existing open models rather than training foundational ones. The difference is subtle but consequential. Fine-tuning adjusts behaviour. Training defines architecture. One improves a tenant’s flat; the other builds the apartment block.
China, despite export controls and hardware restrictions, has produced competitive frontier models such as DeepSeek-V3, which performed strongly on global benchmarks while operating under computational constraints. India’s BharatGen, launched in June 2025 as a government-funded multimodal model, is a welcome beginning. It is not yet a challenger.
The AI Impact Summit’s rhetoric implied parity. The evidence suggests distance.
The compute question
AI is capital-intensive. It depends not merely on talent but on computational power. The IndiaAI Mission, approved with an outlay of roughly $1.2bn over five years, aims to subsidise compute access and stimulate start-ups. That figure sounds large in the Indian context. In the global AI economy, it is modest.
Leading US firms spend sums of comparable magnitude in a matter of months. Chinese technology giants deploy capital at a similar scale. India’s programme is therefore better understood as seed funding than as a transformative investment.
The government has highlighted the availability of 38,000 high-end GPUs to Indian start-ups at subsidised rates of Rs 65 per hour. This is progress. But perspective matters. Baidu alone announced a 30,000-chip AI computing cluster in 2025. China’s national AI compute capacity stood at roughly 246 EFLOP/s in mid-2024, and it had ambitions to reach 300 EFLOP/s by 2025. In comparison, India’s programme is a starting line, not a leap.
If India intends to compete with China—let alone the US—its computing base will need to expand by an order of magnitude. That expansion requires sustained capital, reliable power and regulatory clarity. None is yet assured, despite repeated assurances during the Union Budgets.
Capital, and whose capital it is
Foreign investment in India’s AI ecosystem is significant. Google, Microsoft and Amazon announced combined AI-related commitments in India totalling tens of billions of dollars in 2025. These figures make for impressive slides.
They are also foreign capital, guided by a foreign corporate strategy. Such investment can accelerate deployment. It does not necessarily deepen indigenous model-building. Nor does it guarantee strategic autonomy. If geopolitical conditions shift, capital can be redirected. India needs to learn that quickly.
When Peter Navarro, a senior US trade adviser, publicly asked why Americans were funding AI expansion in India, it was less a rhetorical flourish than a strategic reminder. India’s AI rise remains intertwined with decisions taken in Washington and Silicon Valley.
China’s model is different. State support, domestic hardware firms and local venture capital combine to create a more internally anchored ecosystem. It is not necessarily freer. It is more self-contained. For India to reach there, it needs a similar vision of utilising the state as a leading force, which Mr Modi appears to be reluctant to do.
Energy and infrastructure
AI infrastructure is energy-intensive. India generates roughly 78% of its electricity from fossil fuels. Data-centre water consumption is projected to rise sharply this decade, with dozens of centres located in water-stressed regions. India accounts for 18% of the world’s population but only 4% of its freshwater resources.
Building hyperscale AI infrastructure on such foundations presents logistical and environmental constraints. These are solvable challenges. They are also expensive ones. Expanding data-centre capacity tenfold by 2030, as government white papers suggest, is necessary and will require capital mobilisation beyond what has so far been pledged.
The AI Impact Summit speeches were less explicit about these constraints. Infrastructure, unlike conferences, cannot be improvised.
Governance in draft form
Ahead of the AI Impact Summit, Mr Modi’s government released AI Governance Guidelines proposing new oversight bodies and ethical frameworks. The documents are thoughtful. They are also largely aspirational.
India lacks a dedicated AI regulator. Copyright law provides no clear text-and-data-mining exemption, leaving developers uncertain about training data legality. Liability regimes for AI-induced harm remain undefined. Proposed governance institutions exist on paper, not yet in statute.
This ambiguity may offer flexibility. It also deters high-risk frontier research, as investors and researchers prefer clarity.
China’s governance model is more prescriptive, if less liberal. The US model is more market-driven. While India’s AI model remains in formation.
Robodog: Tragic symbolism at AI Impact Summit
The imported robot that caused such embarrassment at the summit does not determine India’s AI trajectory. It does illustrate a tendency to conflate branding with capacity. The university involved issued multiple statements, oscillating between denial and clarification. The IT secretary responded that the government did not wish “the controversy” to continue.
The controversy was instructive. It revealed how easily symbolic claims can outrun institutional discipline. If a flagship summit cannot vet its exhibitors thoroughly, it raises questions about the broader oversight architecture required for frontier AI.
China’s ascent in AI did not hinge on summit optics. It rested on long-term industrial policy, hardware investment and disciplined state backing. While the US rests on capital markets, deep research universities and corporate risk appetite, India’s formula remains less defined.
Long journey, not a sprint
India’s advantages are real—a vast domestic market, English-speaking engineering talent, geopolitical flexibility in sourcing chips and a vibrant start-up culture. Its weaknesses are equally clear—limited frontier compute, modest public investment relative to ambition, regulatory uncertainty and dependence on foreign capital for advanced hardware.
The gap with the US is vast. The gap with China is narrower but still significant. Closing it will require sustained spending far beyond current allocations, coordinated energy and water planning, clearer governance and a willingness to prioritise infrastructure over imagery.
India has demonstrated that it can deploy digital systems at an extraordinary scale. Whether it can design, train and own frontier AI models remains an open question. The difference between deploying and building is not rhetorical. It is structural.
Events like the AI Impact Summit are useful signals. They are not substitutes for capacity. A country that aspires to lead the AI race must invest at the scale of its ambition. India has begun. It has not yet arrived.
Join our channels on Telegram and WhatsApp to receive geopolitical updates, videos and more.
