OpenAI Unifies ChatGPT and Codex

Greg Brockman is now officially OpenAI’s product chief, with a single mandate: merge ChatGPT, Codex, and the Atlas browser into one desktop agentic environment for all users — not just developers.
Enterprise already accounts for more than 40% of OpenAI’s revenue, and Codex has crossed 3 million weekly active users; the unified platform is designed to deepen that dependency, not merely simplify the interface.
The underreported risk: this consolidation is a vendor lock-in play — enterprises signing agreements today should understand what it means for their switching costs in 2028.

The internal memo arrived with the usual corporate framing — “leadership evolution,” “product alignment.” What it actually described was a strategic reset. OpenAI is no longer running three products that happen to share a brand. It is building one product, with ChatGPT, Codex, and its Atlas browser as tributaries feeding a single agentic platform.

Greg Brockman, OpenAI’s co-founder and president, is the person who will now own that vision. As reported by Wired, his move from interim product oversight to official product lead formalizes a direction the company has been moving toward since March 2026, when it first confirmed plans to merge its core applications into a desktop superapp. The announcement was not a surprise. But its timing — amid OpenAI’s ongoing restructuring from non-profit to for-profit — reveals something deliberate about the company’s priorities: get the product house in order first, let the legal drama resolve on its own schedule.

From Co-Founder to Product Chief — and Why the Split Matters

What’s striking about this move is how deliberately OpenAI has separated two functions that most companies tangle together. Fidji Simo, who leads the commercial side, focuses on taking the unified product to market — enterprise sales, partnerships, pricing. Brockman decides what the product actually does. That is a clean, uncommon division, and it typically signals a company confident enough in its product direction to stop debating it internally.

Brockman’s memo used direct language: “invest in a single agentic platform and to merge ChatGPT and Codex into one unified agentic experience for all.” Three words in that sentence deserve attention. For all. Not for developers. Not for enterprise customers specifically. For all — meaning the same agent environment will serve a startup founder writing code and a global COO managing operations through a natural language interface.

That is simultaneously a product ambition and a distribution strategy. One platform means one contract, one integration surface, one enterprise relationship — and one expanding area of organizational dependency that grows harder to unwind with every passing quarter.

3 Million Users and 15 Billion Tokens Per Minute

Codex — OpenAI’s AI coding agent — crossed 3 million weekly active users this year, with underlying APIs processing more than 15 billion tokens per minute. Those are not experiment metrics. They are production numbers at enterprise scale, with companies like Cisco and NVIDIA already embedded in the stack. Enterprise revenue has now crossed 40% of OpenAI’s total. The consumer side built the brand. The enterprise side is building the business.

A unified product makes that enterprise story cleaner to pitch: one integration surface, one escalation path, one platform your teams already use expanding into every workflow it can reach. That consolidation narrative is not wrong. It is just incomplete.

Is your enterprise AI stack built for agents — or still organized around isolated tools? Epinium’s Transform practice maps your current setup and identifies where consolidation creates leverage, not just convenience →

The Lock-In Question Enterprises Are Not Asking Out Loud

The official narrative positions this as simplification. One app instead of three. Less context-switching. More coherent agent workflows. For an operations leader managing multiple teams across multiple tools, that pitch is genuinely compelling.

What we’re seeing at Epinium is that the brands making the fastest progress with AI aren’t necessarily running on the most powerful individual models — they’re the ones that have established a clear orchestration layer connecting AI capabilities to actual business workflows. OpenAI’s consolidation move speaks directly to that problem. A unified environment reduces the integration friction that slows deployment.

Epinium data

Across the brand teams entering Epinium’s AI transformation programs, the median company arrives managing 4 to 6 separate AI subscriptions with no shared orchestration layer. That fragmentation — not budget — is what consistently delays measurable ROI by three to six months.

The harder question sits one layer deeper. Once your developers write code in Codex, your analysts run queries in ChatGPT, and your researchers browse with Atlas — all within one environment — the switching cost becomes structural. Not insurmountable. Just expensive: in retraining time, in workflow rebuilding, in the institutional knowledge that calcifies around a platform over 18 months.

Anthropic, Google, and Microsoft are each running their own agent orchestration plays. OpenAI’s move accelerates the competition. For a CTO in 2026, the question is no longer which AI tool to try — it’s which platform you are willing to anchor your core workflows on, because that is increasingly a multi-year decision with structural consequences. For broader context on OpenAI’s enterprise momentum, see our earlier piece on what $4B from 19 investors signals for enterprise AI.

OpenAI has never understated its ambitions. What Brockman’s appointment confirms is that those ambitions are now organized into a single product roadmap, with one person accountable for the whole. The companies that will navigate this well are not those that panic about lock-in or dismiss it — they’re the ones that audit their AI dependencies clearly, understand where they would switch and why, and make any new platform commitment with an explicit plan for the exit they hope never to need.

Frequently Asked Questions

What exactly merges when ChatGPT and Codex become one product?

The unified desktop environment brings together ChatGPT’s conversational interface, Codex’s coding and agent execution capabilities, and the Atlas browser’s research features into a single application. Teams will be able to initiate a task in natural language, have an agent execute code and browse the web to complete it, and review the result — without switching tools. Underlying APIs remain accessible separately, so integrations built against individual products are not immediately broken. What changes is where new feature development is concentrated over time.

Will current OpenAI enterprise contracts change as a result of this consolidation?

OpenAI has not announced changes to existing enterprise agreements tied to individual products. Platform consolidations of this kind typically phase in over 12 to 18 months, giving customers time to renegotiate at renewal. Commercial operations are led by Fidji Simo separately from Brockman’s product mandate — a deliberate signal of continuity through the transition. That said, enterprises should clarify at their next renewal what unified pricing and support look like under the new platform structure.

What happens to workflows built specifically on the Codex API?

OpenAI has confirmed that API access for Codex will remain available independently from the consumer superapp. Workflows and integrations built against the Codex API — including CI/CD pipelines and internal developer tools — should not be disrupted by the product consolidation. Over time, however, new capabilities may appear in the unified environment first, creating a feature lag for API-only users. Teams with business-critical Codex integrations should monitor OpenAI’s developer roadmap closely for signals about where investment is shifting.

Should a mid-market brand with limited IT resources pay attention to this?

Yes — but for a different reason than large enterprises. For smaller brands, the consolidation means that entering OpenAI’s ecosystem is soon a single-door decision rather than a multi-tool selection. If ChatGPT is already embedded in daily workflows and Codex is being adopted for task automation, switching costs accumulate faster than teams realize. Mid-market organizations with lean IT teams have less capacity to manage migrations, making the initial platform choice more consequential than it appears in the early stages of adoption.

When does it make more sense to choose a competing AI platform instead of consolidating with OpenAI?

If your primary use case is heavily specialized — real-time financial data processing, domain-specific compliance, or deep cloud-native integration with AWS or Google infrastructure — alternatives like Anthropic’s Claude or Google’s Gemini for Workspace may offer tighter native integrations without requiring a new tool migration. OpenAI’s unified platform is optimized for generalist knowledge work and coding automation; it is less differentiated in verticals like healthcare record processing or regulatory reporting, where domain-specific governance controls matter most. The right moment to evaluate a competitor seriously is when OpenAI’s roadmap gaps are actively costing you implementation time — not simply when you’re curious about what else exists.

Ready to map your enterprise AI stack before the platform wars force the decision? Epinium’s Transform practice works with brand teams and operations leaders to build an AI roadmap that preserves optionality while accelerating deployment. Discover how Epinium structures enterprise AI transformation →