Sarah Friar: ChatGPT hits 700M weekly users and 5M enterprise seats as OpenAI scales toward AGI

Summary

OpenAI CFO Sarah Friar used the GPT-5 launch on August 7, 2025 to publicly anchor several significant business metrics for the first time, signaling the company's growing comfort disclosing operational scale ahead of any eventual public market transaction.

ChatGPT now reaches 700 million weekly active users, with Friar projecting a further lift from the GPT-5 launch. Geographically, 85% of users are outside the United States, with usage tracking closely to global population density across India, Indonesia, Brazil, Vietnam, and the Philippines. That international skew is both a mission statement and a future revenue unlock.

Enterprise and Developer Traction

On the business side, OpenAI has accumulated 5 million enterprise seats across industries from banking to biotech, a figure Friar describes as seats at companies rather than individual company counts. That customer base was built in roughly two and a half years, a timeline she characterizes as extraordinary given the compliance and security table stakes enterprises demand, including SSO, HIPAA compliance, and reference case studies.

The developer ecosystem sits at 4 million builders on the API platform, spanning large companies like Grab through to early-stage Y Combinator startups. Friar frames this distribution as mission-critical, arguing OpenAI cannot deliver AGI to humanity without a functioning developer ecosystem.

Metrics Philosophy and Investor Framing

Friar pushes back on revenue as a primary operating metric, describing it as too lagging for an operator to rely on. Her preferred input metrics run from monthly to weekly to daily actives, with secondary signals including token throughput, API latency, and demand elasticity across price, performance, and uptime for the developer segment.

For investors, she is explicit that OpenAI should not be optimizing for free cash flow in the current period. Compute commitments must be made two to three years ahead of when capacity is needed, creating an intentional mismatch between current costs and future revenue that is likely to sustain losses for some period. She draws a direct analogy to the Mag Seven's earlier growth phases and argues the company's private status is a structural advantage in absorbing that investment cycle.

Capital Structure and Compute Strategy

The $40 billion SoftBank-led round closed recently and was described as massively oversubscribed. Friar acknowledges that pure equity fundraising is an expensive long-term strategy and signals OpenAI is moving toward a more diversified funding approach, including deeper commercial alignment with supply chain partners Microsoft, CoreWeave, Oracle, and Nvidia.

On compute infrastructure, Friar is explicit that small-scale data centers are not useful to OpenAI. She cites Oracle OCI's Abilene, Texas deployment as a model for what good partner behavior looks like, specifically the ability to scale a single footprint by 5x and connect multiple sites. The original Microsoft pre-training fabric, which she recalls as roughly 20 megawatts, now looks trivially small against current gigawatt-scale ambitions.

OpenAI is also increasingly transacting directly with sovereign governments. Recent announcements in Norway and the UK, plus Estonia's deployment of ChatGPT across its high schools, reflect what Friar calls a new category of customer she has not encountered in prior enterprise sales careers.

Federal Access and Open Source Positioning

The decision to offer ChatGPT to every US federal agency at $1 per year is framed as a strategic national asset play rather than a commercial one, with Friar declining to elaborate further on the economics.

On open source, Friar draws on her own history covering Red Hat as a Goldman Sachs research analyst, noting that the Linux ecosystem eventually required enterprise-grade support contracts and predictable maintenance cycles to scale commercially. She stops short of committing OpenAI to that model directly but leaves the door open, pointing to the company's second open source model release two days prior as the starting point for watching how the community develops.