Over the past year, OpenAI has launched multiple products focused on the consumer market, including social applications and AI hardware devices planned for release later this year. However, just last week, the company's CEO, Sam Altman, hosted a dinner in San Francisco for corporate executives, including Disney CEO Bob Iger, sending a clear signal: OpenAI is seriously pursuing the enterprise customer market.
According to an informed source, during the lavish multi-course dinner, Altman told attendees that OpenAI could become a "one-stop shop" for corporate AI needs, with its product portfolio covering the ChatGPT chatbot, the Codex code generation tool, and various workflow automation models.
Another source indicated that the core purpose of the dinner was to preview a new service OpenAI is building specifically for large enterprises. While specific details of the service have not yet been disclosed, its primary goal is to help enterprise customers complete large-scale AI transformation. Typically, such transformation requires companies to completely overhaul their existing technology architecture, fully integrating AI into various business scenarios such as customer service, legacy application code refactoring, and enterprise data management.
This new service also aims to consolidate a company's AI application work. A possible implementation method involves bundling OpenAI's existing diverse products, thereby helping enterprise customers more easily track related expenditures. Currently, services OpenAI provides for enterprise customers include different versions of ChatGPT, paid premium features (such as access to advanced models, unlimited messaging), and model access via Application Programming Interfaces (APIs).
Some dinner attendees keenly perceived the subtext behind these moves: Altman intends to win customers away from main competitor Anthropic. Over the past month, Anthropic's code generation tool Claude Code and office automation tool Cowork have garnered widespread attention in the business world.
Public financial data from the two companies shows that Anthropic disclosed last year that its revenue from selling AI models to enterprises via APIs was projected to surpass OpenAI's. This achievement is quite remarkable, considering OpenAI had a multi-year head start in this market.
Nevertheless, OpenAI's enterprise business scale is still likely larger than Anthropic's. The reason is that many enterprise customers not only purchase its model services via API but also pay for ChatGPT subscriptions. However, many large clients report that within the circle of corporate executives, Anthropic is already viewed as the preferred supplier for enterprise-grade AI services. This reputation has been largely built on Anthropic's decision not to invest heavily in consumer-facing features, instead maintaining a consistent focus on the enterprise market.
In contrast, some major clients point out that OpenAI continues to release various consumer-grade products at a staggering pace, and its enterprise strategy still seems to be in an exploratory and adjustment phase. Over the past few years, OpenAI's primary growth logic has relied on word-of-mouth from ChatGPT's individual users attracting the attention of their employers, leading to large subscription contracts with those enterprises. However, the process of selling large contracts to enterprise customers is often time-consuming, typically requiring multiple rounds of negotiations with executives from different departments over several months, executives who are already accustomed to various business dinners and courtesies.
Informed sources reveal that OpenAI hopes to attract more customers like data platform company Databricks and financial software company Intuit – customers who proactively seek out and directly purchase its products.
Last week, during the World Economic Forum in Davos, Switzerland, two OpenAI executives – Chief Operating Officer Brad Lightcap and Chief Revenue Officer Dennis Woodside (former CEO of Slack) – were also actively engaging with potential enterprise clients there, a source confirmed. Meanwhile, on Wednesday, OpenAI announced to employees at its San Francisco headquarters that Barrett Zoph, a top AI researcher recently hired from Thinking Machines Lab, will lead the development of enterprise-grade products.
Anthropic's Competitive Edge
Anthropic's Claude Code, launched last year, is a core driver of its recent success in the enterprise market. This tool possesses code writing and optimization capabilities. The CEO of a startup spending millions of dollars monthly on both OpenAI and Anthropic platforms stated that for some clients, Claude Code's productivity boost far exceeds that of chatbots used by employees for general office tasks and information retrieval. Furthermore, non-technical staff have also begun using Claude Code and its derivative tool Cowork to handle various non-programming tasks related to office software.
The market popularity of Anthropic's products puts dual pressure on OpenAI: on one hand, it needs to match and optimize competitor performance, and on the other, develop new features favored by large enterprise clients. Concurrently, some of OpenAI's products targeting knowledge work, such as AI agent tools for spreadsheet creation and editing, have not met market expectations.
A consultant working with OpenAI disclosed that OpenAI's leadership, in meetings with consulting partners, previewed plans to upgrade the Codex tool. They indicated that the new version of Codex would soon surpass Claude Code in performance and functionality. Currently, OpenAI is also adding collaborative features and other office scenario functionalities to ChatGPT. This development has not gone unnoticed by Microsoft – as a key partner of OpenAI, Microsoft is accelerating the upgrade of AI features within its Office software suite.
The aforementioned consultant noted that OpenAI is considering launching a new tool to help clients more visually quantify the economic benefits gained from using its AI products. The implementation of such tools will be crucial for OpenAI in securing large enterprise deals.
Additionally, OpenAI is developing tailored features for specific industries like healthcare. Altman posted on social platform X this Friday that the company is developing specialized features for areas like cybersecurity within the Codex tool – and cybersecurity happens to be a key focus area for Anthropic as well.
Informed sources say that over the past year, OpenAI has reorganized its customer sales model: abandoning the previous "multiple people responsible for multiple products" approach in favor of dedicated sales representatives assigned to clients, providing them with a one-stop shop for multiple product sales.
The source also revealed that OpenAI is analyzing the high-value points where enterprise clients use its AI products and plans to introduce tools to help clients quantify these economic benefits. Partner consultants emphasize that the launch of such tools will be core to OpenAI's ability to consistently secure large enterprise contracts.
According to informed sources, OpenAI has signed multi-year cooperation agreements worth at least $100 million each with 7 clients; another 6 clients have signed contracts worth no less than $75 million, and these contracts have the potential to grow to $100 million or more upon renewal.
In contrast, Anthropic announced last December that it had at least 9 clients with annual spending exceeding $100 million, including Microsoft – under their partnership plan, Microsoft's annual investment in Anthropic's models is set to be as high as $500 million.
Both the startup CEO and the OpenAI consultant mentioned that in designing enterprise client contract terms, Anthropic also appears to be ahead of OpenAI. Specifically, Anthropic allows clients to pre-commit to a certain volume of API calls in exchange for more favorable pricing. Some OpenAI clients have privately complained that they were unable to obtain such flexible contract terms.
However, OpenAI optimized its contract terms last year, increasing flexibility. Sources indicate that the company now provides enterprise clients with usage allowances for advanced models and tools, where the more capacity a client purchases, the lower the per-seat unit price becomes.
It is worth noting that some of OpenAI's business initiatives have not progressed smoothly. For instance, the company halved its revenue forecast for AI agent tools last year, projecting related revenue of $1.4 billion for 2025. Sources revealed that in July 2024, OpenAI launched the ChatGPT Agent feature for subscribers, enabling users to create and edit spreadsheets and presentations, but its actual performance failed to meet some internal metrics, such as the goal of "reaching 10% of ChatGPT's weekly active users."
Despite such setbacks, OpenAI remains undeterred – the company's long-standing vision is to establish a strong foothold in both the consumer and enterprise markets. Altman has described ChatGPT as a "super-intelligent personal assistant for work." For example, a year ago, OpenAI demonstrated a prototype AI agent for sales that could automatically qualify sales leads and determine which ones were worth pursuing.
Currently, approximately 40% of OpenAI's revenue comes from enterprise clients. The company's Chief Financial Officer, Sarah Friar, stated at the Davos forum this Wednesday that she expects the contribution from enterprise clients to increase to 50% by the end of this year. Altman posted on social platform X this Thursday that in the past month alone, API sales added $1 billion in annualized revenue for OpenAI. He specifically noted that this figure challenges the external perception of the company: "Many people think we only have ChatGPT."
Comments