After merging with Udemy, Coursera's revenue surged 94%, but core subscription revenue is shrinking

After merging with Udemy, global online education giant Coursera has posted a quarter in which revenue nearly doubled.

Big companies are all cutting budgets—so how else can online education make money? On October 29, Coursera delivered a contradictory report card: thanks to the Udemy merger, revenue ostensibly surged 94%; but if you strip away the surface of the data, its core subscription revenue is actually shrinking. This giant is using "puffiness" to mask "exhaustion."

Looking at the numbers alone, revenue reached $377 million, up 94% year over year—a very impressive result. But a deeper breakdown of the financial statement structure reveals that this growth stems mainly from the consolidation effect. On a comparable basis, its actual revenue declined about 3% year over year.

On one side is the sharp lift in reported revenue; on the other is pressure on the core base. After completing the initial integration of the two major platforms, Coursera urgently needs to prove to the capital markets its ability to sustain profitability and growth.

The "two sides" of the data: revenue doubled, so why is the core base still under pressure?

Coursera's sharp revenue growth this time comes mainly from Udemy's consolidation contribution.

In the third quarter, the company's revenue reached $377 million, up 94% year over year. But calculated on a comparable basis including both companies' results for the same period, revenue actually fell about 3% year over year, with the decline widening further from the previous quarter. This shows that although the merger expanded Coursera's revenue scale, it has not yet reversed the downward trend of the original business.

In fact, this report card from Coursera reflects the deep-water contest across the entire global online education industry. After the traffic era of "burning money for growth" came to an end, even education giants must "huddle together for warmth" to cope with shrinking B-end budgets and rising customer acquisition costs. Merging with Udemy is precisely the card Coursera has played to build its moat.

Looking at specific businesses, enterprise training has become Coursera's largest source of revenue. In the third quarter, this business generated $201 million in revenue, accounting for 53% of total revenue, but comparable revenue grew only 2%. At the same time, the number of enterprise customers fell 4% year over year to 11,897, and the net revenue retention rate was 93%. 11,897 enterprise customers—this churn figure is more dangerous than it looks on the surface. The hardest-hit area is small and medium-sized enterprises. This exposes a fatal weakness after the merger: Coursera focuses on systematized, high-ticket offerings, while SMEs that were used to Udemy's cheap, fragmented courses will, once faced with economic cyclical fluctuations, be the first to cut "heavy-investment" training like Coursera's. The blind merger did not bring customer retention of 1+1>2. The B-end base is churning, and the C-end is faring no better.

The pressure on the consumer business is even more obvious. In the third quarter, comparable revenue for this business fell 9% year over year. Although paid subscribers after the merger reached 1.687 million, up 30% year over year, the personal certificate subscription and one-time course purchase businesses continued to come under pressure, offsetting part of the subscription growth. At present, enterprise and consumer subscription revenue combined account for nearly 90% of Coursera's total revenue. As the revenue structure concentrates toward the subscription model, customer renewals, user retention and the ability to keep paying have also become important factors affecting the company's growth. In addition to revenue pressure on the B-end and C-end, the cost improvements brought by the merger have already emerged.

In the third quarter, Coursera's adjusted EBITDA—compared with the weakness on the revenue side, the cost improvements brought by the merger have already emerged. In the third quarter, Coursera's adjusted EBITDA reached $61.5 million, with the margin rising to 16.3%, up 8.3 percentage points from the same period last year. Since completing the merger in May, the company has successively streamlined its organizational structure, integrated sales teams and cut duplicate functions, and expects to achieve more than $150 million in annualized total cost savings by the end of the year. The original target of $115 million in annualized net synergies to be achieved in 24 months is expected to be reached about eight months after closing. However, affected by factors such as merger and restructuring-related expenses, the company still recorded a GAAP net loss of $25.6 million for the quarter.

The progress of cost integration has not changed management's cautious expectations for short-term revenue. Coursera expects fourth-quarter revenue of $358 million to $366 million, below the third-quarter level. Among this, comparable revenue for the enterprise business is expected to be roughly flat, while the consumer business may slow further.

For post-merger Coursera, cost reduction has achieved phased results, but revenue growth still faces pressures such as enterprise customer churn and weak consumer payment demand. How to use the courses, instructors and customer resources accumulated by the two platforms to develop more competitive products has become the problem the company needs to solve next.

Integration pains: the two major platforms' "genetic recombination"—can 1+1 be greater than 2?

Cost integration is progressing rapidly, but the product-level integration of Coursera and Udemy is still at an early stage. The essence of this "cost reduction and efficiency improvement" is to dress up the income statement by cutting labor costs. But cutting expenses cannot hide the inability to generate revenue; saving money alone cannot support a tech giant's growth ambitions.

The real trouble for Coursera is that its underlying business logic is completely different from Udemy's. The former is more like a "degree and certification system," emphasizing rigor and long cycles; the latter is more like a "fast-food-style skills supermarket," quick to add new offerings but lacking depth. How to fit these two kinds of instructors and two systems into one framework and generate genuine synergies is the hardest nut to crack after the merger.

This choice is related to the content supply models of the two companies. Coursera has long relied on universities, tech companies and professional institutions to develop courses, and currently has more than 425 partners, with strengths concentrated in systematic teaching, professional certification and degree education. Udemy, by contrast, has nearly 100,000 instructors, mainly industry practitioners and professionals providing courses, and can respond more quickly to technological updates and changes in job skills. Over the past year, Udemy added more than 5,000 instructors, further expanding the supply of vocational skills courses.

Each model has its advantages and its limitations. Universities and professional institutions can provide relatively systematic knowledge systems and qualification certification, but course development usually takes a certain amount of time; individual instructors can quickly launch new content, but find it difficult to independently cover the full needs of vocational competency development. After the merger, Coursera hopes to combine the two types of content, both increasing the speed of course updates and retaining the advantages of professional teaching and certification.

In the third quarter of this year, Coursera has already launched two cross-platform experiments.

One of them comes from Udemy instructor Ryan Ahmed. He previously taught technical courses for developers on Udemy, serving hundreds of thousands of learners in total. After the merger, he adapted a series of Claude-related courses into a specialization on Coursera, which became one of the platform's most outstanding new courses in 2026. This is also an early attempt to bring Udemy instructor content into Coursera's course system, indicating that the two platforms have begun exploring shared instructor resources and content distribution channels.

Another experiment directly changed the way courses are organized. Coursera launched the AI Series learning paths, combining courses from different partners and having practitioners in the relevant fields review the content and learning sequence. Currently, the product covers areas such as data analysis, business leadership, software development and cybersecurity, and is planned to expand to more areas in the fourth quarter.

Unlike searching for, purchasing and studying courses separately by course topic, AI Series organizes content around specific career directions. For example, software developers can study courses from different institutions in sequence according to the skills required for the job, without having to judge for themselves the connections between courses and the learning order. At present, these paths mainly rely on professional staff to select and arrange them, and automated combination for different learners has not yet been achieved.

At the same time, Coursera is also expanding the scope of content partnerships. At the company's FWD conference in September this year, it announced the addition of 16 university and industry partners and the launch of 25 professional certificates. The new partners include Harvard Business School Online, which offers leadership and business strategy courses, and GitHub, which develops courses on agentic AI and modern software development. In October, Anthropic added another 10 Claude-related courses, with plans to update them monthly to keep up with changes in AI technology and tools.

These new courses have expanded the platform's content coverage, while the cross-platform experiments further explore how to use existing resources. For the merged Coursera, the number of courses and instructors has already reached scale, and how to organize scattered content into complete learning paths according to the skill requirements of different jobs has become the focus of product integration.

At present, Coursera still mainly relies on manual work to complete course selection, adaptation and combination. Through these dazzling partnerships, it is not hard to see Coursera's anxiety: it is trying to manually stitch together the differences between the two systems at extremely high labor costs. In the AI era, this kind of "human-wave tactic" simply cannot sustain a huge revenue target in the long run. The real remedy may be exactly the AI-native platform Project Helix that they plan to deploy in 2027—enabling the system to automatically arrange learning and assessment according to enterprise needs.

Betting on the future: from "selling courses" to "verifying skills"—is the new story easy to tell?

While course resources are being integrated at an accelerating pace, Coursera is also readjusting its product direction.

In September this year, the company unveiled for the first time the AI-native skills platform Project Helix, which is planned to be broadly opened to enterprise customers in the first half of 2027. Compared with existing online course platforms, Project Helix will focus more on enterprises' actual hiring needs, integrating skills identification, personalized learning, practical training and competency assessment into a single platform.

This adjustment is closely related to the changing skill demands of the AI era. Coursera's 2026 Global Skills Report shows that the enrollment rate for generative AI courses has grown from 25 per minute in 2025 to more than 40 per minute in 2026, while enrollment in critical thinking courses increased by more than 100% year over year. At the same time, a Deloitte survey cited by Coursera shows that 43% of corporate managers believe their organizations are fully prepared in terms of AI technology infrastructure, but only 20% believe their talent is fully prepared. As enterprises increase AI investment, they also need employees to master the ability to apply technology to actual business operations.

In response to this demand, Project Helix translates enterprises' business goals into specific skill requirements, then combines employees' roles, proficiency levels and learning progress to generate corresponding learning paths. Unlike traditional course platforms, which mainly record learning time and course completion, the new platform will also use practical tasks, skills assessments and professional certifications to help enterprises determine whether employees are competent for the relevant work.

Unlike traditional course platforms that only record "learning time," Coursera is betting on Project Helix, which will not be launched until 2027, in an attempt to do "talent assessment." But this is tantamount to snatching food from the tiger's mouth. When it comes to assessing employee skills, a company's own HR and business supervisors are the ones with the most say. If Coursera wants to set standards as a third-party platform, why should enterprises pay for it? What's more, there are now countless free AI tools and assessment systems on the market. Can the capital market still have enough patience for a vision that will only take shape three years from now to support the valuation?

Some related functions have already been put into application. For example, Graded Role Play uses AI to simulate work scenarios such as business negotiations and assess learners' communication and decision-making performance; Graded Labs tests technical capabilities through real operational tasks, including deploying web applications. In addition, the company is working to connect its learning products with tools such as ChatGPT, Claude and Copilot, so that employees can obtain courses and training content based on actual problems at work, further integrating learning into daily work.

However, Project Helix is still in phased development. It is currently being refined together with some enterprise customers, and the consumer version will be launched after the enterprise version. Management clearly stated on the earnings call that a significant revenue contribution from the new platform has not yet been included in the preliminary performance assumptions for 2027.

Coursera's adjustment also provides a reference for domestic vocational education and corporate training institutions.

With the spread of AI tools, the cost of course development and content updates continues to decline, and it will become more difficult to attract customers simply by relying on the number of courses and price. Rather than increasing course supply, developing practical training projects around specific roles, establishing skills assessment standards, and helping enterprises measure employees' actual performance after training will become important directions for vocational training institutions to expand enterprise customers and improve renewal rates.

Coursera's earnings report is actually a microcosm of the entire online education industry entering deep waters. When "selling courses" hits a growth ceiling, how to deeply bind education to real industry needs (such as AI skills) is a hurdle it must overcome. The merger is only the beginning, but the time left for it to trial and error is actually already limited.

This article comes from the WeChat public account "Duojing" (ID: DJEDUINNO), author: Siluo, published by 36Kr with authorization.