Silicon Valley has spent years arguing about when artificial general intelligence will arrive. Databricks CEO Ali Ghodsi thinks the argument is over.
In a conversation published by a16z, Ghodsi said AI already has enough intelligence. The harder problem, in his view, is getting companies to put that intelligence to productive use in ordinary work.
The AGI finish line, by an older measuring stick
Ghodsi’s claim rests on a specific yardstick. He argues AGI has effectively arrived when judged by the definitions people used before 2022: AI that performs at or above human intellectual ability across a range of fields.
His bigger point is about what is missing. According to Ghodsi, the bottleneck is no longer intelligence. It is context.
Put simply, a brilliant new hire is useless if nobody gives them access to the files, the systems, or the history of the business. Ghodsi wants enterprises to connect AI models to their proprietary data and operational systems so the technology can do real tasks, rather than sitting inside a chat window answering questions.
“So many aspects of our lives are going to change over the next twenty years.”
Ghodsi also framed the shift around people, not just software. In the a16z piece, he pointed to employee empowerment to take initiative as one of two major breakthroughs he sees on the horizon.
The numbers behind the confidence
Databricks closed a $5 billion funding round in August 2026 at a $190 billion valuation. That places it among the most valuable private technology companies tied to the AI buildout.
The company’s revenue run rate now exceeds $7 billion, with year-over-year growth of more than 80%.
More than 17 million databases are launched every day by AI agents on the Databricks platform.
Over 90% of the software code Databricks produces is written by AI.
Building for agents, not chatbots
Ghodsi’s comments line up with a clear product strategy. Databricks has been building infrastructure meant specifically for AI agents, the semi-autonomous programs that can take actions rather than just generate text.
Its recent launches include Genie, Unity AI Gateway, Lakebase, and Agent Bricks. Each fits the broader thesis: give AI systems secure, governed access to company data and the tools to act on it.
Not quite the singularity
Ghodsi’s AGI declaration comes with an important caveat. He is not claiming AI can improve itself without limit.
In discussions in October 2026, Ghodsi laid out four conditions he believes would need to hold at once for AI self-improvement: lower resource use, faster training, better accuracy, and repeatability. He said those four have not been achieved simultaneously.
In other words, he is calling the intelligence question settled while leaving the runaway-AI scenario firmly in the not-yet column.
What this means
For enterprise buyers, Ghodsi’s framing changes the shopping list. If the models are already good enough, the competitive edge comes from how well a company organizes, secures, and exposes its own data to AI systems.
A company where more than 90% of code is written by AI, and where agents launch more than 17 million databases daily, is a preview of how technical work may be reorganized.
Declaring AGI achieved is partly a definitional choice, and others in the industry will dispute which definition counts. And by Ghodsi’s own account, the conditions for self-improving AI remain unmet, which means the most dramatic predictions about the technology are still unproven.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.