Databricks Co-Founder Wins Top Computing Prize — and Says AGI Is 'Already Here'

The Association for Computing Machinery has awarded its 2025 ACM Prize in Computing to Matei Zaharia, the Databricks co-founder whose work on Apache Spark fundamentally changed how large-scale data processing is done and who is now focused on making AI systems more capable and more reliable for scientific research. The prize, which carries a $250,000 award and is given to researchers who have made extraordinary early-career contributions, places Zaharia among a cohort that includes recent winners working on cryptography, programming languages, and computer architecture.

The AGI Claim

The announcement generated more attention for Zaharia's comments about AGI than for the award itself. In interviews with TechCrunch and other publications, he argued that the ongoing debate about whether AGI has been achieved is semantically confused in a way that prevents useful analysis. His position: if AGI is defined as AI systems that can perform tasks at or above human level across a wide range of cognitive domains, then that threshold has already been crossed for significant portions of knowledge work — and the interesting question is not whether AGI is here but what the societal and economic implications are of deploying it at scale. "AGI is here already," he said. "The question is what we're going to do about it."

Why This Framing Matters

Zaharia's AGI argument is a variation on a position that has been gaining traction among AI practitioners who are tired of a discourse structured around a definitional finish line that keeps moving. The objection to "AGI is here" claims has historically been that current AI systems are narrow, brittle outside their training distribution, and lack the general reasoning capability that the term implies. The counterclaim — which Zaharia is articulating — is that this objection keeps moving the goalposts: as each capability frontier is crossed, the definition of AGI expands to exclude it. Whether or not you accept the specific claim, the underlying point has practical force: the economic and social disruption from AI systems that are very good at coding, writing, research, and analysis is already happening, and waiting for a definitive AGI declaration before taking it seriously is a policy failure regardless of how the philosophical debate resolves.

Databricks' Research Direction

Zaharia's current work at Databricks focuses on AI reliability and compound AI systems — architectures that combine multiple specialized models, retrieval systems, and verification components to achieve more robust performance than single large models. This is the direction that Databricks' DSPy framework and its DBRX model work have been pointing toward. The research agenda is implicitly skeptical of the scaling-laws-solve-everything view that has dominated the frontier model conversation: Zaharia's approach treats reliability and composability as research problems worthy of the same attention as raw capability improvement.