Episode 7 - Run:ai Acquired for "$700 million" by Nvidia - Interview with co-founder Omri Geller

Run:ai was founded in 2018 to provide the software "operating system" for development of AI and large language models (LLMs). But the real story of run:ai is not the  acquisition price (unconfirmed) of $700 million nor its raise of $130 million in VC money, but in how run:ai scaled its customer penetration so rapidly after a year of invention and re-invention of its business model. (Disclosure: I was one of the first investors, happily so). The benefits for customers include significantly better use of scarce computing resources, since the development of AI requires such extensive computing resources, which also consume a lot of electricity.

It wasn't easy at first, because, hard to believe in 2025, AI was not yet a "thing" in 2018, and LLMs were at first developed mostly in academic and research institutions, not enterprises. Lucky or prescient? And run:ai was not born with a full set of teeth: early on they learned that enterprises wanted a completely different product than did research institutes... Corporations (think auto manufacturers or pharmas) conduct AI or LLM projects as teams, often cross-functional, with different members playing different roles and having different requirements for their part of the project. Run:ai's first product was NOT the product that customers needed!

Highlights:

• Learning to sell - Omri, as a young CEO-founder-engineer (30 at the time), had a steep learning curve as a leader, including learning how to sell. "I need to be out their selling" at least in phase 1.
• Participating in all sales calls - after a while, Omri actually began to like it!
• Each of the initial founders was able to reinvent himself as the needs of the market changed so rapidly.
• It is very hard to know who the right people are - and "I tend to wait to long" before making a tough decision. On the other hand, "sometimes waiting works and the person learns to grow new skills."
• After the first few $ million, the problem changes to "enabling others to sell the product."
• Process is essential to scale, but there are right processes and wrong processes! "You have to nail it!"