Published: April 29, 2026
Tuesday was a weird day for Google.
On one hand, the company broke ground on what might be its most ambitious project outside the US — a $15 billion AI hub in Visakhapatnam, with enough computing power to light up a small city. The stage had Union ministers, Andhra’s CM, Adani’s folks, Airtel’s top brass. Big Indian energy all around.
On the other hand, literally the same day, news broke that Google had quietly signed a classified AI deal with the US Pentagon. For context: this is the same Google that walked away from a military AI contract in 2018 because employees revolted over “Don’t Be Evil” and drone targeting.
Two very different Google stories. One day. Let’s talk about both.
So What’s Google Actually Building in Vizag?
It’s not just a data centre. That would be underselling it.
Google is building three data centre campuses in partnership with AdaniConneX and Nxtra by Airtel, with nearly 1 GW of power capacity. That’s the kind of number you usually associate with aluminium smelters, not servers. This is gigawatt-scale AI infrastructure, plain and simple — the kind of compute that lets you train the next generation of models without looking over your shoulder at power constraints.
The broader plan:
– Three new subsea cables landing in Vizag under the “America-India Connect” initiative
– Long-term clean energy agreements to power this thing without torching the planet
– Workforce training — Google’s STAR program, local hiring, digital literacy for fishing communities
– Community initiatives like watershed management (because building a giant AI hub in a coastal city means asking the locals if they’re cool with it, and actually helping them)
Ashwini Vaishnaw was there. Chandrababu Naidu was there. Jeet Adani said something worth remembering: “When energy becomes more affordable and increasingly powered by clean sources, intelligence becomes more accessible — and that is how India will lead the next phase of digital growth.”
That’s the pitch. And honestly? It’s a good one.
But Here’s the Other Thing That Happened Tuesday
The Information reported — and Reuters and The Guardian confirmed — that Google signed a classified AI deal with the Pentagon.
This is part of the Pentagon’s AI Rapid Capabilities Cell, the same program that signed agreements worth up to $200 million each with OpenAI, Anthropic, and Google in 2025. The work is on classified networks, so details are murky, but the gist is: Google is now building AI systems for US military intelligence analysis on classified systems.
Remember 2018? Google walked away from Project Maven — a drone AI project — because 4,000 employees signed a protest letter and a bunch of them quit. The company’s AI Principles explicitly ruled out “weapons or other technologies whose principal purpose is to cause or directly facilitate injury to people.”
Well, 2026 Google has different math. The cloud business is massive. Competition from Microsoft and Amazon for government contracts is fierce. And the Pentagon is the biggest customer on the planet.
Anthropic, by the way, is currently locked in a dispute with the Pentagon over refusing to do military work. OpenAI and xAI have been more accommodating. Google seems to be threading a needle — doing the work, but keeping it quiet enough that employees don’t riot.
These Two Stories Are Actually About the Same Thing
Here’s why I find this week interesting:
Google is simultaneously building India’s AI future and the Pentagon’s AI infrastructure. Same company. Same technology stack. Same week.
That tells you something about where we are with AI in 2026: AI infrastructure is national infrastructure now. Whether it’s powering a Vizag startup building the next Indian unicorn, or helping the Pentagon analyse intelligence data, the underlying compute, the cloud platforms, and the engineering talent are all the same.
Countries that don’t build their own AI capacity will rent it from whoever does. India is making sure it’s on the building side of that equation.
The timing works. With the Vizag hub, the subsea cables, and a government pushing for “Viksit Bharat 2047” (fancy government-speak for “we want to be a developed country”), India is positioning itself as an AI powerhouse — not just a market for other people’s AI products.
The Bharat AI Shakti Conclave, held alongside the groundbreaking, made this explicit: build local supply chains, build a domestic SME ecosystem around the hub, don’t let this be another case of “foreign company builds, India just consumes.”
A Few Thoughts on What This Means
– India just became Google’s most important AI infrastructure bet outside the US. Fifteen billion dollars at one location over five years isn’t casual. It’s a signal.
– The Adani-Airtel-Google triangle is India’s AI backbone. AdaniConneX brings the real estate and power. Airtel brings the connectivity and cable landings. Google brings the compute stack. Three different strengths, one very powerful combination.
– India needs to think about sovereign AI compute. The Pentagon deal shows how critical AI infrastructure is for national security. India’s government needs to ask: do we have our own version of this?
– The ethics question isn’t going away. Google’s pivot on military AI, Anthropic’s resistance, the broader debate — this stuff matters. India has a history of non-alignment and principled tech policy. There’s room for an interesting voice here.
What do you think — is Google’s dual-track strategy smart business or a slippery slope? And does India get enough credit for building its own AI infrastructure?
Tags: Google, AI, India, Vizag, Adani, Airtel, Pentagon, Data Centres, AI Infrastructure, Viksit Bharat