The Summit That Summarized Everything

Last week, India hosted the fourth Global AI Summit in New Delhi. Over 20 heads of state. CEOs of every major AI company. Delegates from 80+ countries. $200 billion in pledged investment. On paper, it was the most ambitious AI gathering yet — and the first held in the Global South.

In practice, it was a masterclass in the gap between AI ambition and ground-level reality.

The Chaos Was the Story

Let’s start with the logistics, because they’re too good not to.

Yoshua Bengio — one of the actual godfathers of deep learning — couldn’t make it to his own panel because traffic was gridlocked. He delivered his address on AI safety via a blurry video link from somewhere in a Delhi traffic jam, apologizing as he rerouted to make dinner with the Prime Minister instead.

Sara Hooker, CEO of Adaptation Labs, spent four hours in traffic trying to get to a state dinner. She ended up eating room service at 11 PM. “Would have been honored to attend,” she posted. I believe her.

The roads were closed for “VIP movements” — a practice that works fine for one dignitary, but creates citywide paralysis when you invite dozens of heads of state and every tech CEO on Earth to the same square mile. Attendees described walking miles through Delhi to escape the venue. No taxis. No shuttles. Just vibes and gridlock.

Inside the venue, things weren’t better. Overcrowded rooms. No wi-fi. Spotty cell service. A ban on laptops, keys, and earbuds at security — enforced with what one reporter described as “various levels of stringency,” which is diplomatic language for “it depended on who was yelling louder.”

An Indian AI wearables founder had his display tech stolen from the exhibition hall during a security sweep for Modi’s visit. Exhibitors were kicked out midday with no warning, gates closed until 6 PM, and when they returned, the devices were gone.

There was also a shirtless protest. I’ll let you look that one up.

The Robot Dog That Wasn’t

My favorite detail: Galgotias University was showcasing a robot dog at their booth. A professor told state TV it was “developed” by the university. It was, in fact, a commercially available Unitree robot from China.

The university was asked to leave.

Their defense was remarkable: “We would like to clearly state that the robotic programming is part of our endeavor to make students learn AI programming and develop and deploy real-world skills using globally available tools and resources.”

Translation: We didn’t build it, but we turned it on.

This is, accidentally, a perfect metaphor for a lot of what passes for “AI development” globally right now. Countries and companies wrapping existing technology in national flags and calling it innovation. The robot dog incident is funny. The pattern it represents is not.

The Handshake That Wasn’t

The summit’s most viral moment was supposed to be a show of unity. Modi lined up the tech CEOs on stage for a group photo, everyone holding hands. But Sam Altman and Dario Amodei — CEOs of OpenAI and Anthropic, respectively, and also two companies currently locked in an existential rivalry — didn’t hold hands.

Altman later said he was “confused” and “didn’t know what I was supposed to do.”

Here’s what I find fascinating about this moment. These are the two people building the most powerful AI systems on Earth. They’re deciding the trajectory of the most transformative technology since electricity. And when asked to do something as simple as hold hands for a photo, the system broke down.

If you can’t coordinate a photo op, how are you going to coordinate AI governance?

The moment was scrutinized as a symbol of the OpenAI-Anthropic rift, especially since Anthropic had just run a Super Bowl ad taking shots at OpenAI. But I think it’s bigger than that. It’s a symbol of the entire AI industry’s coordination problem. Everyone agrees we need governance. No one agrees on what that means. And when you put the principals in a room and ask them to literally stand together, they can’t even manage the gesture.

The Ghosts

Jensen Huang canceled days before his keynote, citing illness. Nvidia sent a senior executive instead.

Bill Gates pulled out hours before his scheduled keynote. The Gates Foundation said it was “to ensure the focus remains on the AI summit’s key priorities.” This was widely understood to be related to renewed scrutiny over his ties to Jeffrey Epstein, after the DOJ released emails just weeks earlier.

So the summit about AI’s future lost two of its biggest speakers — one to illness, one to the past.

What Actually Got Done

In fairness: the summit wasn’t all chaos. 88 countries signed a diplomatic declaration committing to inclusive AI development. Frontier AI companies made voluntary governance commitments. Over $200 billion in investment was announced.

OpenAI said it would be the first customer of Tata Consultancy Services’ data center business. Google announced education partnerships for Gemini. Every CEO praised India’s talent pool and market potential.

These are real things. But they’re also the same things that get announced at every summit, in every country, by every CEO who wants market access. The announcements are the easy part. The execution — as the summit itself demonstrated — is where everything falls apart.

The Metaphor

Here’s why I can’t stop thinking about this summit.

The AI industry right now is exactly like this event. Grand ambitions. Massive investment. Genuine talent and potential. But the infrastructure isn’t ready. The coordination is broken. The VIPs are gridlocking the roads for everyone else. Someone is presenting a Chinese robot as their own innovation. The people who are supposed to be leading can’t agree on whether to hold hands. And the logistics — the boring, unglamorous work of actually making things function — are falling apart.

The Economist published a piece this weekend titled “The AI Productivity Boom Is Not Here (Yet).” The parenthetical is doing a lot of heavy lifting. Because the “yet” assumes the boom is coming — it’s just delayed. But what if the delay is the story? What if the gap between what AI can do in a demo and what it can do in the real world is wider than anyone wants to admit?

India’s summit didn’t fail because India isn’t capable. It failed in the ways that reveal universal truths about where we are. We’re all presenting robot dogs we didn’t build. We’re all gridlocked by our own VIP movements. We’re all confused about whether we’re supposed to hold hands.

The AI future is real. But the present is a summit in Delhi where the godfather of deep learning is stuck in traffic, and nobody can get the wi-fi to work.