The man building Google's brain just put a date on when machines match humans—and he's more worried about the journey than the destination.
The Summary
- Demis Hassabis, Google DeepMind CEO, says AGI will arrive by 2030, plus or minus a year—a timeline far more aggressive than peers like Andrew Ng, who thinks it's decades away
- At Google I/O 2026, the company packed three hours with AI product announcements spanning Search, Gemini, Docs, Gmail, YouTube, and Android
- Hassabis spent as much time discussing AI's problems and mitigation strategies as celebrating breakthroughs—a rare posture for a Big Tech exec at a developer conference
The Signal
Demis Hassabis doesn't do hedging. While most AI leaders wave their hands about "eventually" or "when the technology matures," the Google DeepMind CEO just marked his calendar: AGI by 2030. That's machine intelligence on par with human thinking across multiple domains, not just narrow tasks. Four years out. Maybe three, maybe five.
The timeline matters because Hassabis has the credibility and resources to make it real. He co-founded DeepMind in 2010, sold it to Google in 2014, merged it with Google Brain in 2023, and now runs the unified lab behind Gemini and AlphaFold. When he says 2030, he's not selling vaporware. He's reporting from the frontier.
"2030 is when I expect it to arrive, either plus or minus a year."
But here's what separates this from typical conference hype: Hassabis spent equal time at I/O talking about what could go wrong. Most executives treat risk like a mandatory compliance slide. Hassabis treated it like the actual engineering problem it is. The risks aren't theoretical anymore because the AI isn't theoretical. It's in Gmail, YouTube, Search, Docs. Billions of people are using it daily, which means billions of edge cases, misuses, and unintended consequences.
Google's I/O announcements spanned every major product line, but the subtext was integration at scale. AI isn't a feature toggle anymore. It's infrastructure. And infrastructure failures cascade.
Key tensions in play:
- Timeline compression: AGI in four years vs. decades (Andrew Ng's camp)
- Scale vs. safety: Shipping AI to billions while still figuring out guardrails
- Research vs. product: Hassabis still frames this as advancing the science, even as Google monetizes every breakthrough
The Implication
If Hassabis is right about 2030, we're already in the last sprint before AGI. That means the agents, tools, and systems being built today aren't just products. They're the scaffolding for what comes next. The companies moving fastest on agent orchestration, context windows, and multimodal reasoning aren't building for Web4. They're building Web4's foundation while the concrete is still wet.
For builders: the window to define how agents interact with humans, own assets, and operate autonomously is narrowing fast. The defaults set in the next 24 months will be hard to change once AGI lands. For everyone else: the AI you're using today is the dumbest it will ever be. Plan accordingly.