Sundar Pichai at India AI Summit: $15 Billion Bet and the Case for Bold, Responsible AI

Google’s CEO stands on stage in New Delhi and makes one of the largest single-country AI infrastructure commitments to date: $15 billion for India, including a full-stack AI hub in Visakhapatnam with gigawatt-scale compute and a new international submarine cable gateway. But the speech is not just an investment announcement. Pichai uses India as the canvas to paint his vision of how AI should be built, deployed, and governed.

The speech in context

Sundar Pichai speaks at India’s AI Summit in February 2026, addressing Prime Minister Modi and global leaders. He opens with a personal story: as a student, he took the Coromandel Express through Visakhapatnam, a quiet coastal city. Now Google is building a global AI hub there. The contrast anchors his central argument: progress happens when ambition meets infrastructure, and India is uniquely positioned for both.

The speech spans three pillars: be bold, be responsible, work together. Within this framework, Pichai moves from scientific breakthroughs to agricultural applications, from deepfake detection to government adoption, building the case that AI’s value depends not just on capability but on how widely and wisely it’s deployed.

The $15 billion infrastructure play

The headline number is $15 billion in Indian infrastructure investment. The Visakhapatnam hub will house gigawatt-scale compute and a new international submarine cable gateway, bringing jobs and AI capabilities across India. This is Google’s largest single-country infrastructure commitment outside the US, signaling that India is central to its global AI strategy, not peripheral.

The investment is not just hardware. Pichai frames it as end-to-end: compute, connectivity, and capability. The submarine cable gateway connects India’s AI infrastructure to global networks, making Visakhapatnam a node in Google’s worldwide compute fabric rather than an isolated data center. On top of this, the “America India Connect” initiative will build four new submarine fiber optic cable systems between the US and India, announced just a day before the speech.

Beyond India, Google is investing in compute infrastructure in Thailand, Malaysia, and other regions. The underlying logic: AI inclusion requires compute infrastructure and connectivity to come first.

Science as the boldest argument for AI

Pichai leads with AlphaFold. For 50 years, predicting protein structures was a grand challenge that stalled drug discovery. Demis Hassabis and the Google DeepMind team solved it, and made the database freely available. Over 2.8 million researchers across 190+ countries are using it, from malaria vaccine development to antibiotic resistance research. Two-thirds of approved drugs now involved a protein whose structure was predicted by AlphaFold in their development.

“For 50 years, predicting protein structures was a grand challenge and a blind spot that stalled drug discovery.”

The materials science example is equally striking. Google’s GNoME model discovered over 380,000 new stable materials, the equivalent of 800 years of accumulated human knowledge in materials science. Some of these materials are already being synthesized in labs, with applications ranging from better solar cells to more efficient batteries.

Google is also asking bold questions across the scientific stack, from cataloging DNA disease markers to building AI agents that serve as partners in the scientific method. These examples serve a rhetorical purpose: they demonstrate that AI’s most transformative applications are not chatbots or content generation, but accelerating scientific discovery in domains where human progress was fundamentally bottlenecked.

AI for India’s billion-person problems

The speech’s most compelling section maps AI onto India’s specific challenges at scale:

Healthcare: AI systems detecting diabetic retinopathy and diabetic macular edema from retinal scans, deployed in Indian hospitals. Google’s AI-powered TB screening tool can detect tuberculosis from chest X-rays and has screened over 3 million people across 31 countries. In India, where TB remains a major public health challenge, AI enables screening at a scale that human radiologists simply cannot match.

Agriculture: Over 30 million Indian farmers use Google’s AI tools for crop disease detection. A farmer takes a photo of a diseased plant and gets an instant diagnosis. This is not a tech demo; it is economic infrastructure. Agriculture employs hundreds of millions of Indians, and crop disease is one of the largest causes of yield loss. In summer 2025, the Indian government sent AI-powered weather forecasts (via Google’s Neural GCM model) to millions of farmers for the first time, helping them protect their livelihoods during monsoon season.

Flood forecasting: Google’s AI-powered flood forecasting system now operates across 24 countries, providing early warnings that help communities evacuate and prepare. India, with its vast river systems and monsoon-driven flooding, is a primary beneficiary.

Language: Gemini now works in 10 Indian languages. India has 22 officially recognized languages and hundreds of dialects. Making AI accessible in local languages is the difference between AI serving India’s urban, English-speaking elite and AI serving all 1.4 billion people.

Developing world: In El Salvador, Google partnered with the government to bring affordable AI-powered diagnosis to thousands who could never afford to see a doctor. In Ghana, Google collaborates with universities and NGOs to expand research and open-source tools across more than 20 African languages.

“We cannot allow the digital divide to become an AI divide.”

India’s digital infrastructure advantage

Pichai makes an argument that gets less attention than it deserves: India’s existing digital public infrastructure, including UPI (Unified Payments Interface), Aadhaar (biometric identity), and ONDC (Open Network for Digital Commerce), gives it a structural advantage in AI adoption that few other countries possess.

The logic: AI needs data, identity, and distribution channels to work at scale. India has built all three as public infrastructure. UPI processes billions of transactions monthly. Aadhaar covers over a billion citizens. ONDC is creating an open commerce layer. These are not just government projects; they are adoption infrastructure that AI applications can build on top of.

This is a genuinely important observation. Most discussions about AI adoption focus on compute and models. Pichai is arguing that the bottleneck for real-world AI impact is distribution and integration, and India has solved more of this problem than any other developing country.

Responsible AI and the trust imperative

The responsible AI section is brief but substantive. Pichai highlights SynthID, Google’s tool for watermarking AI-generated content, used by journalists and citizen fact-checkers globally. In a country of 1.4 billion people where WhatsApp forwards can shape elections, content authenticity tools are not academic exercises.

AI will reshape the job market, automating some roles, evolving others, and creating entirely new careers. Pichai draws an analogy: the concept of a professional YouTube creator did not exist 20 years ago; today there are millions worldwide. The message is that workforce transformation is real but not unprecedented.

He frames responsibility not as a brake on innovation but as a prerequisite for trust, calling trust the “bedrock of adoption.”

The triangle of collective action

Pichai’s closing argument outlines three actors needed:

Governments must serve a dual role: as regulators (setting rules, addressing risks) and as innovators (bringing AI into public services). He cites Uganda using AI and satellite imagery to locate priority areas for electrification, and Memphis, Tennessee using AI scans from buses to identify and fix potholes. These are deliberately unglamorous examples, emphasizing that AI’s value extends beyond frontier research to making basic public services work better.

Tech companies must build products that boost knowledge, creativity, and productivity.

Businesses of all sizes need to adopt AI to innovate and transform their sectors and empower workers.

“We have the opportunity to improve lives at a once in a generation scale.”

A Few Observations

The speech is an investment pitch dressed as a policy address, but that does not diminish its substance. The $15 billion figure makes it concrete; the scientific examples make it credible; the India-specific applications make it relevant.

The most valuable thread is the infrastructure argument. Pichai is making the case that AI adoption is not primarily a technology problem but a systems problem. India has built digital rails, from payments to identity to commerce, that most developing countries lack. AI applications that work on top of these rails can reach a billion people. AI without these rails remains a demo.

One notable detail: Google teams are currently exploring how to deploy data centers into space, echoing Elon Musk’s earlier claim that in 36 months, the cheapest place to put AI will be space.

• The AlphaFold and GNoME examples are the strongest evidence for AI’s transformative potential: not generating text or images, but accelerating scientific discovery in domains where human progress was fundamentally stuck
• India’s digital public infrastructure (UPI, Aadhaar, ONDC) is a genuinely underappreciated advantage for AI deployment at scale
• 30 million farmers using AI crop disease detection is more impressive than most AI metrics reported by tech companies, because it represents actual adoption in a sector that directly affects livelihoods
• The tension in the speech is between Google’s commercial interest (selling cloud and AI services) and the genuine public goods argument; Pichai navigates this by focusing on open-access projects (AlphaFold database, flood forecasting) alongside commercial offerings
• The Waymo anecdote with his 83-year-old father is deft speechcraft: it humanizes him, connects his Indian roots to Silicon Valley innovation, and subtly signals that autonomous driving still has a long way to go in complex environments