Artificial intelligence is now being treated as core economic infrastructure, and OpenAI CFO Sarah Friar says most organizations are barely scratching the surface of what it can do.
Friar wrote a LinkedIn post on Monday reflecting on her experience at the annual meeting of the World Economic Forum (WEF) in Davos last week. “This year felt different,” she writes. AI is no longer just a side conversation or a future bet, she explained. AI is currently being evaluated as core economic and strategic infrastructure, similar to geopolitics, energy, and security, she noted.
OpenAI’s finance chief since June 2024, Friar highlighted “capability overhang” as a concept that kept resurfacing at Davos—the gap between what AI can already do and the value organizations actually capture. According to Friar, there is a mismatch between today’s powerful AI capabilities and the relatively shallow way most people and companies use them, with advanced tools still only lightly integrated into real workflows and decision-making.
“Experience and execution are closing that gap faster than any amount of rhetoric,” she writes. “At OpenAI, we see that frontier users use seven times the amount of intelligence than the average user—they’re going deep on coding, deep research, and pushing the models to really be thought partners.”
Along those lines, OpenAI recently released “Ending the Capability Overhang,” new research documenting the phenomenon. The researchers observed a clear country-level gap that is not driven by income alone. Across more than 70 countries where ChatGPT is widely used, some countries use advanced AI features three times more than others, per person.
Interestingly, while large economies like the U.S. and India have the most total users, and small wealthy nations like Singapore and the Netherlands use AI the most per capita, advanced AI adoption is spreading everywhere, according to the researchers. Meanwhile, countries like Pakistan and Vietnam are among the world’s biggest users of agentic tools, using them more than twice as often as average.
In essence, some countries are already using AI to tackle harder problems and move faster, regardless of their resources. These early adopters are seeing real productivity gains—their workers can focus on more complex work, new products and services, and accelerate innovation in ways that drive economic growth and improve living standards, OpenAI finds.
Another moment that resonated with Friar in Davos was the WEF CFO gathering, which “reinforced how pragmatic finance leaders are,” she writes, adding that there’s “broad conviction that AI is inevitable, but deployment hinges on ROI, clean data, and simpler systems; this is a change-management challenge, not a belief gap.”
That focus on tangible results is reflected in OpenAI’s own recent performance. In an interview with Fox’s Maria Bartiromo last week, Friar said, “An IPO isn’t off the table; it’s a question of when.”
OpenAI was valued at around $500 billion in its most recent completed share sale. In 2023, revenue reached $2 billion in annual recurring revenue; it rose to $6 billion in 2024 and jumped to more than $20 billion in 2025, according to Friar’s Jan. 18 blog post.
This revenue growth closely tracked an expansion in computing capacity. OpenAI’s computing capacity rose from 0.2 gigawatts (GW) in 2023 to 0.6 GW in 2024 and about 1.9 GW in 2025.
Beyond infrastructure investments, OpenAI is also expanding into new consumer-facing domains. The company announced earlier this month the debut of ChatGPT Health—a dedicated experience inside ChatGPT where it says users can securely connect medical records and wellness apps such as Apple Health, Function, and MyFitnessPal to further personalize conversations. The company stated it would not train its models on personal medical data, Fortune reported.
The company’s strategy of combining infrastructure expansion with practical, domain-specific applications reflects the pragmatic approach to AI deployment that Friar observed among finance leaders in Davos.
