The India AI Impact Summit ends today. The Leaders' Declaration — the Delhi Declaration — is being adopted this morning, the GPAI Council is convening to align multilateral priorities, and the final sessions are wrapping up after six days that redefined the global AI landscape.
This was not a typical technology conference. Twenty-two heads of state. Sixty ministers. The CEOs of Google, OpenAI, Anthropic, Meta, Microsoft, and Reliance Industries — all in one room with the Prime Minister of India. One hundred and eighteen countries represented. Three hundred thousand registrations. Five hundred sessions. And more than $200 billion in AI investment commitments made from a single stage.
The India AI Impact Summit was the largest AI gathering in history, the first hosted in the Global South, and the event where the world's most powerful governments and corporations collectively declared that AI infrastructure — not AI models — is now the strategic priority.
For the past three weeks, this blog series has tracked the forces converging on this moment. Today, as the summit closes, I want to do something different: not report what happened, but distil what survives the summit. What actually changes for enterprises after this week? What commitments will materialise? What should you act on — and what can you safely ignore?
The Investment Ledger: What Was Actually Committed
Let's start with the numbers, because the scale deserves a clear accounting.
Adani Group: $100 billion by 2035 for renewable-energy-powered AI data centres. Five gigawatts of capacity across multiple campuses, partnering with Google (Visakhapatnam) and Microsoft (Hyderabad, Pune). Expected to catalyse an additional $150 billion in server manufacturing, sovereign cloud, and supporting infrastructure — a projected $250 billion AI ecosystem.
Microsoft: $50 billion by the end of the decade for AI infrastructure across the Global South. This follows $17.5 billion already committed to India last year. Five-part programme: infrastructure, skills for schools and nonprofits, multilingual AI, local innovation, and AI diffusion measurement.
Google: $15 billion infrastructure investment in India. Full-stack AI hub in Visakhapatnam with gigawatt-scale compute, international subsea cable gateway, four new US-India subsea fibre optic cables. Partnership with Karmayogi Bharat for 20 million civil servants, $30 million AI for Science Challenge.
Blackstone: $600 million in Indian AI infrastructure startup Neysa.
India Government: $1.1 billion AI venture capital fund. Adding 20,000 GPUs to the existing 38,000 under the IndiaAI Compute Portal. $18 billion in approved semiconductor projects.
Anthropic: Bengaluru office (second in Asia), Infosys partnership, partnerships with EkStep Foundation, Pratham, Central Square Foundation for AI in education, agriculture, health.
Other announced investments: Amazon and others in India AI infrastructure; Nvidia partnering with Indian venture capital firms; Sarvam AI launching indigenous foundation models (30B and 105B parameters).
Total direct commitments announced at or around the summit: over $200 billion. Projected ecosystem impact: over $400 billion by 2035.
What the Delhi Declaration Means
The Leaders' Declaration being adopted today is the summit's formal output — a shared roadmap for global AI governance adopted by the participating nations. It builds on the progression from Bletchley Park (2023, focused on existential AI risk), through Seoul (2024, expanded to safety frameworks), and Paris (2025, shifted toward innovation and action).
The Delhi Declaration represents the first major AI governance framework to emerge from the Global South, and its emphasis is different from its predecessors. Where previous declarations centred on safety and regulation, Delhi centres on three pillars: People (human-centric AI), Planet (environmental sustainability), and Progress (inclusive economic growth). Seven thematic working groups produced deliverables across these pillars — covering shared compute, AI commons, inclusion, safety, human capital, science, and economic growth.
The practical implication for enterprises: the regulatory and governance environment for AI is crystallising around principles that are now consistent across multiple major frameworks — MANAV (India), EU AI Act (Europe), the International AI Safety Report (global), and the Delhi Declaration (multilateral). These principles converge on data sovereignty, transparency, human oversight, accountability, and inclusive access.
Enterprises building AI infrastructure aligned with these converging principles will operate across jurisdictions. Those that don't will face increasingly complex compliance requirements.
Five Things That Survive the Summit
Summits generate announcements. Most don't survive contact with quarterly earnings reports. Here are the five developments from this week that will actually shape enterprise AI for the next two to five years.
1. The Infrastructure Build-Out Is Real — and It's Already Under Construction
This is not speculative investment. Google's Vizag hub is an extension of infrastructure they need — Gemini processes 10 billion tokens per minute, and they're supply-constrained. Adani's $100 billion is backed by existing partnerships with Google and Microsoft, an operational 10-gigawatt renewable portfolio, and a joint venture (AdaniConnex) already running 2 gigawatts of data centre capacity. Microsoft's $50 billion follows $17.5 billion already deployed in India.
What makes this different from previous summit announcement cycles: the infrastructure announcements are tied to existing commercial relationships, operational assets, and demonstrable demand. The compute is needed today. The energy partnerships are in place. The construction is already underway.
For enterprises: this means AI compute capacity in emerging markets — including the India-Gulf-Africa corridor — will be substantially more available within 18 to 24 months than it is today. Planning enterprise AI deployments around current compute constraints will leave you under-deployed relative to what's coming.
2. Multilingual AI Has Crossed a Viability Threshold
Anthropic's Sonnet 4.6 supporting 10 Indic languages. Google's platform operating in 18 Indian languages for 20 million civil servants. India's BharatGen Param2 model supporting 22 Indian languages with multimodal capabilities. Sarvam AI launching indigenous models optimised for Indian contexts.
This is no longer a roadmap. Multilingual AI at enterprise scale is being deployed today — not perfectly, but at sufficient quality for production use cases. Amodei acknowledged the technology isn't perfect yet, but described it as actively bridging the "long tail" of regional languages.
For Gulf enterprises: the Arabic, Hindi, Urdu, and multilingual capabilities that have been perpetual "coming soon" features are now being prioritised by every major AI provider simultaneously. Evaluate providers on current multilingual production performance, not promises. The competitive advantage goes to enterprises that deploy multilingual AI agents first — not those that wait for perfect translations.
3. The Platform Competition Has Clarified — and Model-Agnostic Architecture Is Now Table Stakes
This summit brought together every major AI platform company on the same stage. Pichai, Altman, Amodei, and Wang (Meta) all delivered keynotes or major addresses. Each is investing billions to become the platform through which enterprises operate.
But the investment patterns reveal something important: every platform company is also investing in open infrastructure. Google's subsea cables. Microsoft's Global South programme. Anthropic's partnerships with open Indian foundations. The Model Context Protocol under Linux Foundation governance. India's own indigenous foundation models.
The platform layer is competitive. The infrastructure and connectivity layers are increasingly open and interoperable. Enterprises that lock into any single platform are making a bet that their chosen provider will win — against competitors with hundreds of billions in capital. Enterprises that build model-agnostic architectures can leverage all of them.
This was the thesis of our very first blog in this series three weeks ago. The summit validated it at the highest levels of the industry.
4. Sovereign AI Isn't a Buzzword Anymore — It's Built Into the Infrastructure
India's "whole-of-nation" AI strategy described by Minister Vaishnaw — "frugal, sovereign, and scalable." Adani's explicit framing of "sovereign energy and computing backbone." Modi's MANAV framework with national sovereignty over data as a core pillar. Macron's praise for India's "open, interoperable, sovereign" digital infrastructure. The Delhi Declaration's emphasis on equitable access and national self-determination.
The pattern is clear: every major economy is building AI infrastructure designed to maintain sovereign control over data and compute. This isn't protectionism — it's pragmatism. AI systems that process enterprise data, make business decisions, and interact with customers create dependencies that governments are no longer willing to leave entirely to foreign providers.
For Gulf enterprises: the sovereign AI frameworks being built across the GCC — Saudi Arabia's national AI strategy, UAE's AI investment programme, Qatar's technology vision — are now part of a global movement validated at the highest diplomatic levels. Building enterprise AI that can operate within sovereign frameworks isn't an optional compliance exercise. It's a strategic requirement.
5. The AI Divide Is the New Digital Divide — and It's Creating Urgency
Microsoft's Brad Smith framed it explicitly: AI usage in the Global North is roughly twice that of the Global South, and the gap is widening. The UN Secretary-General's message was equally direct: AI with "dignity as the default setting." Brazilian President Lula da Silva is addressing the summit today on "AI for the good of all."
This framing matters for enterprises because it's driving policy. Governments across emerging markets — India, Brazil, the Gulf states, Southeast Asia, Africa — are creating incentive structures to accelerate AI adoption. Subsidised compute access (India's GPU programme at $0.72/hour). Tax incentives for AI infrastructure. Regulatory frameworks designed to facilitate adoption rather than constrain it.
Enterprises operating in these markets have a window — probably 18 to 24 months — where government incentives, infrastructure build-out, and competitive dynamics all align to make enterprise AI deployment faster and cheaper than it will be later. That window closes as markets mature and first-mover advantages consolidate.
What You Can Safely Ignore
Not everything announced this week will materialise as described. Here's what to weight appropriately.
Headline investment figures: $100 billion over nine years (Adani) and $50 billion by end of decade (Microsoft) are commitments spread across many years and regions. They represent direction and scale, not immediate deployments. Track actual infrastructure coming online, not announcement totals.
Model-specific claims: Amodei's suggestion that India could see 25% economic growth from AI is aspirational framing, not a forecast. Treat capabilities demonstrations and growth projections from AI providers as marketing, and evaluate actual production performance independently.
The "Magna Carta of AI" framing: The Delhi Declaration matters as a governance signal, but previous AI summit declarations (Bletchley Park, Seoul, Paris) have had limited enforcement impact. Track which specific provisions get implemented in national regulations — that's where it becomes real.
Summit organisation critiques: Bloomberg reported logistical issues. TechPolicy.Press critiqued the structure. The Galgotias University robot dog incident went viral. These are genuine concerns about execution but don't change the strategic significance of the investments and governance frameworks announced.
The Enterprise Playbook After This Week
If you've followed this blog series from February 2, here's how the fourteen posts map to an enterprise action plan:
Architecture layer (Blogs 1, 10, 13): Build model-agnostic. The platform competition confirmed at this summit — Google, OpenAI, Anthropic, Meta, Salesforce all competing for enterprise adoption — means locking into any single provider creates risk. Build orchestration layers that allow you to leverage multiple providers.
Infrastructure layer (Blogs 3, 5, 6, 8, 14): The $200+ billion in infrastructure investment announced this week means compute capacity, connectivity, and deployment infrastructure will be substantially more available in 18–24 months. Plan deployments to the infrastructure that's coming, not just what's available today.
Governance layer (Blogs 2, 4, 12, 15): MANAV, Delhi Declaration, EU AI Act, International AI Safety Report — governance frameworks are converging on consistent principles. Build AI systems aligned with the common principles across frameworks now, not after regulations are finalised.
Execution layer (Blogs 7, 9, 11): Production deployment is what matters. The summit validated that every major AI company and government is past the exploration phase. Enterprises that are still evaluating are falling behind those deploying.
Security layer (Blog 12): Production safeguards — runtime monitoring, circuit breakers, audit logging, agent governance — are engineering requirements, not optional additions. The AI Safety Report findings presented at this summit confirm that pre-deployment testing alone is insufficient.
Restructuring layer (Blog 13): Your software stack is being restructured whether you act or not. The $285 billion SaaSpocalypse and Mistral CEO's replatforming statement at this summit confirm the structural repricing of enterprise software. Navigate the transition deliberately or be navigated.
The Closing Thought
Three weeks ago, I started this series by arguing that the model-agnostic enterprise would define the next era of AI adoption. Fourteen blog posts later, the India AI Impact Summit — the largest AI gathering in history — has confirmed every major thread we've tracked.
The models are commoditising. The platforms are competing. The infrastructure is being built at unprecedented scale. The governance frameworks are converging. And the enterprises that will lead are the ones acting on all of this now — not waiting for the next summit, the next announcement, or the next model release.
Two hundred billion dollars of AI infrastructure investment was committed from a single stage this week. The Delhi Declaration was adopted by more than a hundred nations. The CEOs of every major AI company stood shoulder to shoulder with heads of state, announcing partnerships and investments that will reshape enterprise technology for the next decade.
The summit is over. The build-out has begun. The question for every enterprise leader is the same one it's been since February 2: are you building on this foundation, or watching others build first?
Next week, we shift from coverage of the India AI Summit to a new phase of the series. The global context has been established. Now we turn to execution — the specific, practical work of turning these macro developments into enterprise outcomes.
Stay tuned.