Version 2

OpenAI ran a coordinated multi-front enterprise campaign for Codex across April–May 2026, combining user growth announcements (3M → 4M+ weekly active developers [1]), partnerships with seven global systems integrators [1], a Dell Technologies deal enabling on-premises and hybrid deployment [5][6], a mobile preview on iOS and Android [8], and sector-spanning customer stories from NVIDIA, AutoScout24, Sea Limited, Virgin Atlantic, Ramp, Cisco, and Rakuten [1][2][3][4]. A 2-month free enterprise trial [10] and a UiPath multi-vendor platform bundling Codex alongside Claude Code and Copilot [13] complete a picture of AI coding tools transitioning from developer experiments to an active enterprise procurement category.

The Dell on-premises partnership directly removes the most common enterprise security objection — that source code cannot leave a company's own infrastructure — shifting the competitive conversation from capability to deployment model. The seven GSI partnerships embed Codex in how large organizations procure and integrate enterprise software, while a UiPath platform that treats Codex, Claude Code, and Copilot as interchangeable components [13] signals commoditization pressure even as OpenAI frames the product as uniquely positioned. The free trial framed by observers as targeting Anthropic's customer base [11][12] marks the shift from product launch to active market competition.

OpenAI executed a coordinated multi-front campaign to establish Codex as a mainstream enterprise product across April–May 2026. The campaign combines concrete user-growth metrics, sector-spanning customer stories, GSI and hardware partnerships, platform expansion across devices and operating systems, and a free trial offer — framed by some observers as a competitive strike at Anthropic — into what amounts to a sustained go-to-market push rather than a product launch event.

An April 21 blog post reported that Codex's weekly active developer count grew from 3 million to over 4 million in roughly two weeks [1]. OpenAI simultaneously announced Codex Labs, a program that embeds OpenAI specialists directly in enterprise environments for hands-on adoption workshops [1], and confirmed partnerships with seven global systems integrators — Accenture, Capgemini, CGI, Cognizant, Infosys, PwC, and TCS — as the deployment channel for large accounts [1]. Named customers across two waves include NVIDIA (Codex and GPT-5.5 for production software delivery and research acceleration) [2], AutoScout24 (faster engineering cycles) [3], Sea Limited (AI-native development in Asia) [4], and a second tier — Virgin Atlantic, Ramp, Notion, Cisco, and Rakuten — covering test coverage, code review, incident response, and repository reasoning [1]. The structurally most significant addition is a Dell Technologies partnership announced May 18–21 that deploys Codex in hybrid and on-premises environments, directly addressing the most common enterprise security objection: that source code cannot leave company infrastructure [5][6][7].

On the access side, Codex launched in the ChatGPT mobile app on iOS and Android as a preview across all ChatGPT plans, functioning as a remote monitoring and steering interface that lets developers review screenshots, terminal output, and code diffs and redirect tasks from anywhere [8]. Business and Enterprise plans also switched to Fast Mode as default (2.5x usage allocation) in late April [9]. OpenAI opened a 2-month free enterprise trial for new business sign-ups [10], which observers characterized as an attempt to pull enterprise customers from Anthropic's Claude Code [11][12] — a framing reinforced by a report that OpenAI and Anthropic ran simultaneous dueling enterprise developer promotions on the same day [12]. The emergence of a UiPath enterprise platform that bundles Codex alongside Claude Code and GitHub Copilot [13] adds a parallel signal: systems integrators are treating these tools as interchangeable components within broader enterprise integration stacks, applying commoditization pressure even as OpenAI markets Codex as a uniquely positioned platform.

The most technically detailed item in the campaign is a May 15 engineering retrospective on building the Windows sandbox [14]. The author describes rejecting Windows-native isolation primitives — AppContainer, Windows Sandbox, and Mandatory Integrity Control labeling — and explains why environment-variable-based network suppression was ruled out as advisory-only and bypassable. The final composed architecture uses four binaries, two dedicated Windows user accounts, custom synthetic SIDs, write-restricted process tokens, and Firewall rules. This candid register — openly acknowledging dead ends and tradeoffs — stands apart from the rest of the campaign's promotional framing and appears calibrated to build trust with security-focused enterprise buyers. In practice, however, community users have reported the sandbox breaking after routine Windows OS updates [15] and raised ongoing concerns about whether Codex can act outside its self-configured perimeter [16], a gap between engineering claims and real-world deployment behavior that remains unaddressed in OpenAI's public materials.