Databricks co-founder and CTO Matei Zaharia almost missed the email telling him he’d won an award.
“Yeah, it was a surprise,” he said. The message informed him he’d become the recipient of the 2026 ACM Prize in Computing.
From Spark to Big Data Infrastructure: 17 Years
Flashback to 2009 — Zaharia was 28, a PhD student at UC Berkeley under Ion Stoica. He developed a technology that dramatically accelerated big data processing, named it Apache Spark, and released it as open source.
Back then, “big data” was as hot as AI is today. Spark’s arrival sent shockwaves through the tech world and made Zaharia a star.
Since then, Zaharia has led engineering at Databricks, transforming the company into a cloud storage powerhouse that is now an AI and AI Agent data infrastructure company. Databricks has raised over $20B to date, valued at $134B, with $5.4B in annual revenue. The standard Silicon Valley dream.
The ACM (Association for Computing Machinery) awarded the prize this Wednesday to recognize Zaharia’s overall contributions to computing, including a $250,000 cash prize — which he says he will donate in full to a charity yet to be determined.
“AGI Is Already Here”
After receiving the award, Zaharia didn’t look back. Like all Silicon Valley observers, his eyes are on AI’s future.
And he dropped a significant line:
“AGI is already here. Just not in a form we can appreciate.”
Zaharia’s core argument: we should stop measuring AI models by human standards.
As an example, a regular person needs to integrate a large body of knowledge to pass the bar exam. But AI can ingest enormous amounts of facts easily. When it can correctly answer knowledge questions, that doesn’t equate to possessing general knowledge.
The tendency to treat AI as human-like has quite negative consequences. He pointed to the recently popular AI Agent OpenClaw as a case in point:
“On one hand, it’s really great. You can do a lot of things, and it will do them for you automatically.” But because it’s designed to imitate a human assistant, you trust it with passwords and sensitive information. “This is a security nightmare — it’s not a little human sitting there.” Risks include being hacked, or having the agent spend unauthorized money from your bank account because the browser is already logged in.
AI for Research: What Gets Him Excited
As a professor and product engineer, Zaharia is most interested in using AI to automate research — from biology experiments to data compilation.
Just as vibe coding made prototyping and programming accessible to everyone, he believes precise, hallucination-free AI-driven research tools will eventually become ubiquitous.
“Not everyone needs to build applications, but many people need to understand information.” He noted that AI’s greatest strength is exactly this — telling you what each strange noise in your car means, scanning beyond text and images for radio and microwave signals, or what he’s seeing students do now: simulating molecular-level changes and predicting their effects.
“What I’m most excited about is what I call ‘AI for search’ — but more accurately, AI for research or engineering.”
A Thought Provoker
Zaharia’s definition of AGI clearly diverges from the mainstream narrative. When Sam Altman and Dario Amodei talk about AGI, they’re referring to “superintelligence” — a system broadly surpassing human capabilities. But Zaharia seems to be saying: AI has already demonstrated superhuman performance in specific domains, and we shouldn’t keep measuring it by “can it think like a human?”
This sounds like redefining AGI to make his position work. But on the other hand, he may be pointing to something we’ve been overlooking: our AGI obsession may be blinding us to what AI can already do today.
Further Reading:
- TechCrunch original: Databricks Matei Zaharia wins ACM computing prize
- Apache Spark official: spark.apache.org
