Databricks co-founder earns ACM honor, claims ‘AGI already exists’

Databricks co-founder and CTO Matei Zaharia nearly overlooked the email informing him that he had been named the 2026 recipient of the ACM Prize in Computing. Reflecting on the moment, he said it came as a surprise.

Zaharia’s journey traces back to 2009, when the technology he developed during his PhD at the University of California, Berkeley, under the guidance of Professor Ion Stoica, became the foundation for Databricks.

At the time, Zaharia created a system that significantly accelerated large-scale data processing tasks. He released it as an open source project known as Spark. During that period, big data held importance similar to that of artificial intelligence today, and Spark quickly reshaped the technology landscape. Zaharia, then just 28 years old, gained widespread recognition in the tech industry.

Since then, he has led engineering efforts at Databricks, helping grow the company into a major cloud data platform that now plays a central role in powering AI systems and agents. Over the years, Databricks has raised more than $20 billion in funding, reached a valuation of $134 billion, and achieved a revenue run rate of $5.4 billion.

On Wednesday, the Association for Computing Machinery awarded Zaharia the ACM Prize in Computing in recognition of his broad contributions to the field. The award includes a $250,000 cash prize, which Zaharia has said he plans to donate to a charitable cause that has yet to be decided.

While the recognition highlights his past work, Zaharia remains focused on the future, particularly the role of artificial intelligence. He expressed a perspective that differs from many in the field, stating that artificial general intelligence, or AGI, already exists, though not in a form that aligns with traditional expectations.

According to Zaharia, one of the key misunderstandings around AI is the tendency to evaluate it using human standards. He explained that while humans must accumulate deep, structured knowledge to pass exams or perform specialised tasks, AI systems can process and recall vast amounts of information in fundamentally different ways. Correctly answering knowledge-based questions, he argued, does not necessarily equate to human-like understanding.

This comparison between AI and human behaviour, he said, can lead to practical risks. He pointed to tools like OpenClaw, which can automate complex tasks but also introduce security concerns. Because such systems are designed to function like human assistants, users may trust them with sensitive information such as passwords or financial access, increasing the risk of misuse or unintended actions.

Zaharia emphasised that AI systems should not be treated as human equivalents, noting that they operate differently and require different safeguards.

In both his academic role and his work at Databricks, Zaharia is particularly interested in how AI can accelerate research and engineering. He sees potential for AI to automate processes such as biological experimentation, large-scale data analysis, and information synthesis.

He also compared this shift to the rise of “vibe coding,” which made software development more accessible. Similarly, he believes AI-driven research tools — especially those that can operate with high accuracy and minimal hallucinations — could become widely used across many fields.

Zaharia noted that while not everyone needs to build software applications, many people need to understand and interpret complex information. He expects AI to evolve in ways that leverage its strengths, such as analysing mechanical issues, processing signals beyond human perception, or simulating molecular-level changes to predict outcomes.

Looking ahead, he highlighted what he sees as one of the most promising areas: AI applied to search and research, particularly in scientific and engineering contexts, where it could significantly expand what individuals and organisations can achieve.