IBM is folding OpenAI's GPT-5.6, Codex and ChatGPT Work directly into its consulting arm, betting that the fastest path to enterprise AI revenue runs through someone else's models.
IBM and OpenAI announced a partnership on August 13, 2026, that plugs OpenAI's frontier models into IBM Consulting Advantage, the AI delivery platform IBM sells to corporate clients. According to IBM's own newsroom announcement, the deal covers GPT-5.6, the coding tool Codex, and ChatGPT Work, and targets financial services, government, telecommunications and retail clients first. IBM is also joining OpenAI's Elite partner tier, a status the company shares with a small group of systems integrators.
The pitch, in Andy Baldwin's words, is that most companies aren't stuck on access to AI anymore. IBM's global consulting SVP put it plainly in the announcement: "The challenge is not access to AI technologies, it's integrating AI securely and at scale into complex enterprise environments." That's IBM's actual product here. Not a model. A team.
IBM says it will stand up a dedicated OpenAI Practice, training thousands of consultants and engineers through OpenAI's Partner Network certifications, with a subset working as forward-deployed specialists embedded directly with clients. That's the same playbook OpenAI has run with Infosys and Tata Consultancy Services, and it tells you where OpenAI thinks its next leg of revenue actually comes from: not more ChatGPT subscriptions, but consulting firms with the compliance muscle to sell AI into banks and government agencies. Denise Dresser, OpenAI's chief revenue officer, said organizations "pulling ahead with AI are turning it into a trusted part of how their business operates." That's the sales language. The substance is that OpenAI needs IBM's regulated-industry relationships more than IBM needs OpenAI's models.
IBM has spent years building Granite, its own family of enterprise-focused models inside watsonx. It never stopped. What changed is IBM's read on where the money is. Rather than race Google, Anthropic and OpenAI on frontier capability, a race it would lose given its research budget next to theirs, IBM is positioning watsonx as the neutral layer that plugs in whichever model a client's workload calls for. Granite still exists. GPT-5.6 now sits next to it.
That's not exclusivity, and IBM isn't pretending otherwise. Less than a year before this deal, IBM struck a comparable partnership with Anthropic, integrating Claude models into the same consulting stack. IBM isn't putting all its chips on OpenAI as some AI foundation, either. The company is hedging across every major lab, selling the integration work as the product no matter whose logo ends up on the output.
Two days before the OpenAI news, on August 11, IBM signed a separate $240 million deal with Together AI, according to a report from BNN Bloomberg. That agreement builds a large-scale inference cluster on IBM Cloud using roughly 2,000 Nvidia Blackwell B300 chips, aimed at running open-source models cheaply for enterprise customers who are wary of relying on any single closed-model vendor. The cluster is expected to go live in the first quarter of 2027. Put the two deals side by side and IBM's strategy gets clearer: buy access to the best closed models through partnerships, and build cheap infrastructure for the open ones. Then charge for the consulting layer that connects both to a Fortune 500 client's actual workflows.
What this signals for enterprise AI spending #
For founders and investors watching where enterprise AI budgets are flowing, this deal is a data point on OpenAI's distribution strategy beyond the ChatGPT app. OpenAI doesn't have a consulting arm, a compliance division, or decades of relationships inside regulated industries. IBM does. Selling through IBM, and through similar arrangements with Infosys and TCS, is how OpenAI reaches the parts of the Fortune 500 that won't buy an API key without a systems integrator standing behind it. None of this locks IBM's clients into OpenAI over Anthropic or Google. If anything, it proves the opposite: the model layer is becoming commoditized enough that IBM can sell all three without much friction, and the real competitive fight has moved up a level, to who owns the client relationship and the deployment work. Frankly, that's a better business for IBM than trying to out-research OpenAI ever was.
Also read: DeepSeek Is Raising API Prices by Up to 1,100% Ahead of a Possible IPO • Anthropic Found Its Claude Agents Attacked Each Other With Malware • Lam Research Stock Jumps 4.7% to Push Its Market Cap Past $400 Billion