Silicon Valley has spent years arguing over when artificial general intelligence will arrive. Ali Ghodsi, co-founder and CEO of data and AI company Databricks, believes the argument is over. In a conversation published by a16z, the venture capital firm Andreessen Horowitz, Ghodsi said AI already has enough intelligence, and that the harder problem 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 that 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 concerns what is still 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 carry out 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 said.

He also framed the shift around people, not just software. In the a16z conversation, 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, a figure that places it among the most valuable private technology companies tied to the AI buildout. The company's revenue run rate — an annualized figure based on recent revenue — now exceeds $7 billion, with year-over-year growth of more than 80%.

Operational metrics point in the same direction. More than 17 million databases are launched every day by AI agents on the Databricks platform, and 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 — 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 product 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. Those four, he said, have not been achieved simultaneously. Together, they give observers a concrete yardstick for weighing future claims about self-improving AI.

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, offers 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. 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.