Good morning. OpenAI CFO Sarah Friar says AI tools like Codex are now becoming a baseline expectation in finance hiring, as the company works to secure enough computing power to meet rapidly rising demand.
Speaking at the Liquidity Summit 2026 this week, Friar said the way OpenAI evaluates talent is changing alongside the technology it builds.
“I would never hire a finance person who didn’t know how to use Excel, and I probably wouldn’t hire a finance person today that doesn’t know how to use a tool like Codex,” she said.
Codex is the company’s AI coding agent for automating software and technical tasks through natural-language prompts. Knowledge workers now represent about 20% of Codex users and are growing more than three times as fast as other user groups, according to OpenAI’s new report.
Friar’s stance on considering AI skills in finance when hiring reflects a growing trend. Deloitte’s Finance Trends 2026 survey of more than 1,300 global finance leaders found that AI and automation skills are now their top priority for developing and sourcing new finance talent—ranking ahead of traditional competencies like regulatory compliance and cost management.
This shift in hiring expectations reflects a broader transformation in how finance work is done inside AI companies. As AI tools become more capable, the ability to effectively use them will be a defining skill for future finance professionals.
That change is happening against the backdrop of a major constraint for OpenAI: computing power.
“Compute is a very scarce resource at the moment,” Friar said, describing demand as rising up a “vertical wall” that continues to outpace supply. She noted that even with aggressive infrastructure investment, OpenAI expects compute to remain tight into 2026.
Compute refers to the processing power required to run and train AI models, including the underlying chips and systems that support AI services.
Because of that shortage, Friar said OpenAI has had to invest heavily in infrastructure ahead of demand. The company previously faced criticism for its aggressive compute purchases, but she defended the strategy. “Thank God we did,” she said, “because in ’26 we still won’t have enough compute.”
She added that compute constraints extend far beyond chips alone. Energy supply, land availability, permitting speed, memory production, talent pipelines, and community relationships all play a role in whether AI infrastructure can scale quickly enough. Trust is part of the supply chain too, she said.
“It’s really important there on the trust side, that we don’t leave communities behind,” she said.
Have a good weekend.
Sheryl Estrada
sheryl.estrada@fortune.co
This story was originally featured on Fortune.com