OpenAI announced on June 22, 2026 that Samsung Electronics is rolling out ChatGPT Enterprise and Codex to every Samsung Electronics employee in Korea and to all employees worldwide in its Device eXperience (DX) division. The deal covers far more than engineering seats: marketing, manufacturing, and product teams are in scope, and it lands as OpenAI converts customer logos into standardized seats at a footprint large enough to shift enterprise procurement defaults.
What did Samsung actually buy?
The commitment is enterprise-wide access to ChatGPT Enterprise and Codex, not a developer-only Copilot replacement. Under the agreement, per OpenAI’s announcement, ChatGPT and Codex go to all Samsung Electronics employees in Korea plus all employees worldwide in the Device eXperience (DX) division. The stated use spans software development, marketing, product development, and manufacturing. That scope is the important nuance: this is a workforce productivity contract in which coding seats are one component, not the whole deal.
Codex, in OpenAI’s framing, has outgrown its original software-development role. Employees can use it to “turn ideas into working software, internal tools, websites, and automated workflows,” according to the announcement. Harrison Kim, General Manager of OpenAI Korea, called it a deployment where Samsung is “embracing AI not as a tool limited to certain teams or functions, but as a core platform for improving how employees around the world work and innovate.”
What the announcement does not contain matters as much as what it does. There is no statement that Samsung is dropping GitHub Copilot, no confirmation that Samsung was ever a Copilot customer, no seat count, and no deployment timeline. The relationship is also not new: Samsung was already working with OpenAI to supply advanced memory semiconductors for future AI infrastructure, and this deal extends the relationship from components into workforce tooling. The bundling of ChatGPT Enterprise with Codex, covering non-technical work, means the commitment cannot be read as a straight IDE displacement.
How big is the footprint Codex just got sanctioned across?
The sanctioned surface sits inside a 262,647-employee manufacturer that posted roughly US$220.7 billion in 2024 revenue, though only two groups are confirmed in scope: Korea-based staff and the global DX division. DX is a subset of total headcount, so it would be inaccurate to imply all 262,647 employees receive Codex. Even narrowed, the confirmed groups represent the kind of named anchor that procurement teams cite when standardizing a vendor.
The deal is a cluster signal, not an isolated win. OpenAI lists a dense roster of Korean enterprises already on ChatGPT Enterprise, OpenAI APIs, or Codex: LG Electronics, Samsung SDS, Krafton, Toss, MUSINSA, Nexen Tire, Korea Zinc, plus LG Uplus, LG CNS, GS E&C, TVING, and HanaTour. When a national market’s flagship manufacturers, a top game studio (Krafton), and a major fintech (Toss) all land on the same vendor’s enterprise tier, the cluster effect compounds: local procurement norms, reseller relationships, and peer benchmarking all tilt toward the incumbent. Separately, Seoul National University provides ChatGPT Edu at no cost to all 47,000 community members, and a Kakao integration puts ChatGPT responses inside KakaoTalk group chats. OpenAI is not winning one account in Korea; it is building a national default.
Where is Codex’s enterprise growth coming from?
OpenAI reports more than 5 million weekly Codex users, with Korea weekly active users up nearly 800% since February 1, 2026. A separate OpenAI figure puts overall Codex weekly active usage up 400% earlier in 2026, according to the company’s Ona acquisition announcement. Every one of these numbers is self-reported by the vendor. There is no independent audit, and the baselines (400% “from earlier in 2026,” 800% “since February 1”) are vendor-defined. Read them as directional, not precise.
The growth numbers accompany a structural push that goes beyond direct seat sales. OpenAI is acquiring Ona to give Codex persistent cloud environments where agents keep working inside a customer’s cloud even when laptops are closed, a signal that Codex is being repositioned for long-running, multi-session agentic work rather than inline IDE completion. On the procurement side, OpenAI and Oracle partnered so OCI customers can apply eligible Oracle Universal Credits toward OpenAI models and Codex, broadening the buying surface beyond direct seats into existing cloud commitments. The OpenAI Partner Network, launched June 4, 2026 with a $150 million investment and a target of 300,000 certified consultants by year-end, names Codex as one of its first specialization areas.
The trajectory is a vendor building the procurement rails that turn individual pilots into standardized enterprise seats. Distribution, partnerships, certification, and cloud credits are the rails; the Samsung announcement is a milestone along them.
Does this actually pressure GitHub Copilot?
The deal pressures Copilot on procurement inertia, not on confirmed displacement. No source says Samsung is leaving Copilot, and the announcement provides no evidence that Samsung was ever a Copilot customer. The honest reading is that a flagship manufacturer just sanctioned Codex across its workforce, and sanctioned seats are how enterprise procurement defaults get reset.
The mechanism is straightforward. GitHub’s incumbent advantage in developer tooling has rested on IDE integration and the muscle memory of individual engineers. But enterprise purchasing is decided at a different layer, where security review, data governance, single-vendor consolidation, and existing cloud credits carry more weight than which autocomplete plugin a developer prefers. ChatGPT Enterprise bundles data protection, user and access management, and security controls, per OpenAI’s description. When that bundle wins a 260,000-employee manufacturer, the IDE default stops being automatic for the next procurement cycle, even if no seat has formally switched yet.
GitHub is not standing still. GitHub’s own site continues to market Copilot for writing, testing, and fixing code and lists enterprise customers including American Airlines, Duolingo, Mercedes-Benz, Shopify, and Spotify. Both vendors are signing large enterprise logos. The open question is whether the decision gets framed as a developer-tooling choice or a broader AI-platform choice. A deal like Samsung’s, where coding is one of several sanctioned use cases inside a single enterprise contract, favors the platform framing, and that framing advantages OpenAI’s bundled approach over Copilot’s narrower IDE positioning.
Claude, from Anthropic, sits in the same pressure zone. Enterprise buyers evaluating Claude Code against Codex now weigh OpenAI’s distribution momentum, the Oracle credit path, and the partner network against Anthropic’s offering, none of which this announcement resolves. The point is structural rather than conclusive: standardized enterprise seats at flagship manufacturers are the scarce resource, and OpenAI is accumulating them faster than its competitors are publishing counterexamples.
What should dev-tool buyers watch next?
Three signals would convert procurement momentum into confirmed displacement. First, actual deployment scope: how many of the in-scope DX and Korea employees are active Codex users versus simply licensed, and whether Samsung publishes usage figures rather than the access figures in the announcement. Second, any confirmed displacement, an explicit statement that Copilot or another tool is being retired, which no source currently provides. Third, whether the Ona-driven persistent-cloud model changes which workflows move off local IDEs entirely, since that would shift the competitive frame from “which autocomplete” to “which agent runtime.”
Until those signals appear, the defensible takeaway is narrower than “Copilot is losing Samsung.” OpenAI has won a large, named, workforce-wide commitment at a flagship manufacturer, bundled it into a broader AI platform deal, and stacked it on top of an acquisition, a cloud distribution partnership, and a certification network. That is a procurement story with real switching-cost implications for teams still defaulting to Copilot. It is not yet a displacement story, and the difference between the two is where the honest analysis lives.
Frequently Asked Questions
Does the deal cover Samsung’s semiconductor and foundry operations?
The confirmed scope is all Korea-based Samsung Electronics staff plus the global Device eXperience division, which covers phones, TVs, and home appliances. Samsung’s Device Solutions arm, the memory and foundry business that supplies HBM to AI accelerators, is confirmed only for its Korea-based workers, not its global fabs and design centers, so a semiconductor engineer outside Korea falls outside the deployment.
How is this different from a per-seat Copilot rollout?
Copilot has been sold chiefly as a per-developer subscription anchored in the code editor, so procurement tracks named engineering seats. The Samsung contract is a ChatGPT Enterprise bundle that folds marketing, manufacturing, and product staff into the same agreement, which makes the sanctioned surface the employee base rather than developer headcount. The closest structural analogue in OpenAI’s portfolio is the BBVA banking deployment, which also spanned non-technical roles rather than coding seats alone.
What does the pre-existing memory-chip relationship give OpenAI?
Samsung was already supplying OpenAI with advanced memory semiconductors for AI infrastructure, so the workforce deal sits on top of an existing supplier relationship rather than a cold vendor review. That history gives OpenAI an account-level foothold, with security and procurement teams already engaged on a separate contract, which compresses the evaluation cycle a standalone software vendor would face. Memory supply and software seats can also be negotiated as a package in future renewals.
Why might a Copilot displacement never be announced?
Enterprise tool retirements are rarely publicized because confidentiality clauses get written into renewal contracts, and a conglomerate has no incentive to signal a vendor switch to GitHub or Microsoft. The displacement signal to watch is therefore indirect, such as procurement filings or engineering-blog tooling mentions, rather than a press release. A Samsung usage figure for Codex, if OpenAI ever publishes one, would carry more weight than any statement about leaving a competitor.
