Artificial intelligence has moved from experimental technology to core economic infrastructure, but most organizations are failing to extract its full value. That was the central message from Sarah Friar, chief financial officer of OpenAI, following last week’s annual meeting of the World Economic Forum in Davos.
In a LinkedIn post reflecting on discussions with global leaders, Friar said AI is no longer treated as a side bet or future possibility. Instead, it is now evaluated alongside geopolitics, energy, and security as essential economic infrastructure. Yet despite that recognition, she warned of an AI capability value gap, a mismatch between what AI systems can already do and how lightly most companies are actually using them.
Why AI’s Promise Is Outpacing Its Real-World Use
Friar described the concept of “capability overhang” as a recurring theme in Davos. Powerful AI tools exist today, but they are often only superficially embedded into workflows, decision-making, and strategy. In many organizations, AI is still treated as an add-on rather than a core operating system.
According to Friar, the companies closing this gap fastest are not those making the loudest claims, but those focused on experience and execution. At OpenAI, she noted that advanced users consume roughly seven times more AI capacity than average users. These frontier users rely on AI for deep coding tasks, advanced research, and ongoing analytical support, effectively using models as thought partners rather than simple tools.
The implication for executives is clear. Competitive advantage will not come from access to AI alone, but from how deeply it is integrated into daily work.
Global Data Shows Adoption Is Uneven but Accelerating
OpenAI recently published research titled Ending the Capability Overhang, which documents how uneven AI usage remains across countries. The study examined usage patterns in more than 70 countries where ChatGPT is widely available.
While large economies such as the United States and India account for the highest total number of users, smaller wealthy nations including Singapore and the Netherlands lead on a per-capita basis. More striking, however, is that advanced AI usage is not strictly tied to income levels. Countries such as Pakistan and Vietnam rank among the most active users of agentic tools, using them more than twice the global average.
The research suggests that some economies are already applying AI to complex problems at scale, regardless of their overall resources. These early adopters are seeing tangible productivity gains, allowing workers to focus on higher-value tasks, develop new products, and accelerate innovation. Over time, OpenAI argues, these gains could translate into stronger economic growth and improved living standards.
CFOs See AI as an ROI and Change Management Challenge
Another key takeaway for Friar came from conversations with finance leaders at the WEF CFO gathering. She said the discussions reinforced how pragmatic CFOs remain about AI deployment. There is broad agreement that AI adoption is inevitable, but investment decisions hinge on measurable returns, clean data, and simplified systems.
For many organizations, the barrier is not skepticism about AI’s potential, but the difficulty of integrating it into legacy processes. Friar framed this as a change management challenge rather than a belief gap, one that requires leadership, training, and operational redesign.
OpenAI’s Growth Mirrors Its Infrastructure Bet
That focus on execution is reflected in OpenAI’s own financial trajectory. In a recent television interview, Friar said an initial public offering is a matter of timing rather than possibility. The company was valued at roughly $500 billion in its most recent completed share sale.
Revenue growth has been rapid. Annual recurring revenue reached $2 billion in 2023, climbed to $6 billion in 2024, and exceeded $20 billion in 2025, according to Friar. This expansion closely tracked OpenAI’s investment in computing capacity, which rose from 0.2 gigawatts in 2023 to approximately 1.9 gigawatts in 2025.
Beyond infrastructure, OpenAI is also expanding into consumer-facing applications, including a new health-focused experience within ChatGPT that allows users to connect wellness data while maintaining privacy safeguards.
Taken together, Friar’s message from Davos is consistent and measured. AI’s economic impact will depend less on ambition and more on disciplined execution. The organizations that treat AI as true infrastructure, rather than a peripheral experiment, are likely to define the next phase of growth.
