OpenAI folds Statsig into Apps, taps founder Vijaye Raji as incoming CTO

OpenAI is bringing its experimentation stack in-house while saying Statsig will keep operating independently. In an all-stock, $1.1 billion deal announced September 2, the company said founder Vijaye Raji will become CTO of Applications after close, per OpenAI’s Statsig acquisition announcement. The transaction is subject to regulatory approval.

What’s actually new

Raji is slated to report to Fidji Simo, OpenAI’s CEO of Applications. He will lead product engineering for ChatGPT and Codex across core systems, infrastructure, and Integrity. That gives Simo a seasoned operator who has shipped at consumer scale.

Statsig itself is not a research bet. It’s a build-faster system: A/B testing, feature flags, and real-time decisioning. OpenAI already uses the platform internally. Now it’s part of the Applications org’s backbone.

The Seattle-based team will join OpenAI once the deal closes. OpenAI says Statsig will continue serving existing customers from its Seattle office. Continuity is the promise. Speed is the goal.

The evidence and context

OpenAI describes the acquisition as a way to “accelerate experimentation” across Applications. That lines up with the immediate leadership move. Raji’s decade at Meta and his startup tenure at Statsig point to execution over novelty. It’s a scale hire.

The company is also reorganizing around that thesis. Kevin Weil, until now chief product officer, shifts to research as VP of AI for Science. He will work with chief research officer Mark Chen. Weil’s product team, including ChatGPT head Nick Turley, now reports to Simo.

Enterprise gets its own lead. Srinivas Narayanan becomes CTO of B2B Applications and will report to COO Brad Lightcap. That acknowledges different rhythms for business software: compliance, procurement, and longer sales cycles. Consumer features can sprint. Enterprise roadmaps can’t.

External reporting adds useful detail. Bloomberg puts the price at $1.1 billion in stock and notes Statsig raised $100 million in May at a $1.1 billion valuation. The Verge underscores the reporting lines and the “operate independently” promise. Both point to a classic pattern: reduce vendor friction, hire the builders, and wire them into the org that ships product.

Why this deal, and why now

ChatGPT runs at consumer-internet scale with research-lab volatility. Traditional product pipelines break under that load. Experimentation must be always-on, segmented, and risk-aware. Bringing the experimentation layer in-house reduces latency between idea, rollout, and rollback. It also concentrates data and governance in one place.

This is vertical integration for product, not just compute. OpenAI already pursues scale on the model and infrastructure side. Owning the experimentation fabric is the application-layer equivalent. It can privilege AI-specific metrics—hallucination rate, refusal behavior, latency under load—over generic clickthrough.

The appointment also clarifies Simo’s mandate. She is building an Applications machine with measured speed: ship, measure, iterate, guardrail. Raji is the systems counterpart who turns that philosophy into practice. It’s a pragmatic pairing.

Implications for OpenAI and rivals

If Statsig becomes OpenAI’s standard experimentation substrate, feature rollout can become safer and faster. Think graduated releases keyed to model updates, traffic cohorts, and safety thresholds. The practical effect: fewer jarring changes for users and fewer regressions in production.

Competitively, this is a moat move. An experimentation stack tuned for AI agents, multimodal prompts, and safety interventions can become a proprietary advantage. It is harder to copy than UI and easier to defend than marketing.

There’s also a recruiting angle. OpenAI keeps buying teams with scar tissue from shipping at scale. That shortens onboarding and lowers management risk. It also signals to senior operators across tech that OpenAI values operating expertise, not only research prestige.

What’s not said

Integration is always the hard part. OpenAI says Statsig will continue to serve external customers. That is reassuring for them, but it raises questions about roadmap priority and data governance. Clear walls will be needed.

Regulatory review remains a gating item. OpenAI has made several big moves this year. Another acquisition, even in product analytics, could invite fresh scrutiny of how much of the AI stack it controls. Expect questions.

Finally, “Applications” is still a moving target. Weil’s shift toward research and Narayanan’s move to B2B show an org finding its lines. The structure looks sound on paper. The test will be shipping without safety compromises while enterprise revenue grows. That’s the needle.

Bottom line

OpenAI is codifying a product doctrine: rapid, measured experimentation at scale. Buying the tool it already uses, and hiring its builder to run Applications engineering, is the cleanest way to make that doctrine real. It’s less glamorous than a new model. It’s more consequential for users.

Why this matters:

• OpenAI is hardwiring experimentation into its Applications org, a sign it’s shifting from sporadic launches to disciplined, measurable shipping at ChatGPT scale.
• The leadership split—Simo + Raji on consumer, Narayanan on B2B, Weil to research—clarifies how OpenAI plans to deploy models safely while building durable, enterprise-grade products.