AI & EdtechSeptember 19, 2026

The Great EdTech Compression

How 90% of EdTech Disappears, Part 5 People have not stopped learning. The capabilities EdTech built are not disappearing either. What is disappearing is the amount of independent industry required to deliver them.

Sahra-Josephine Hjorth lying in an open coffin wearing sunglasses and a blue gown, holding a bouquet of white lilies

In 2021, my team and I tried to reverse-engineer an EdTech unicorn. We mapped 22 of the world's most valuable EdTech companies across eight clusters and asked a fairly ambitious question: if you stripped away the founders, the brands, and the timing, was there a formula underneath? There was.

The winners combined content and software. They sold to both businesses and consumers. They made content available on demand. They used AI and data to personalize what learners saw next. And perhaps most interestingly, most of our findings contradicted what all the edtech investors told us. At the time, it was popular to peddle that you either had to make content or software, never both, and the roadmap definitively showed that to be not true.

After months of handheld research in a pre-Generative AI environment, my team and I turned the findings into six articles and a 33-page roadmap for CanopyLAB. CanopyLAB was nominated as a future possible edtech unicorn in Europe, but it never did become one.

The other day, I reopened that roadmap while writing this series. I expected parts of it to look naïve five years later. Instead, I had the opposite problem. We were right. Almost everything we identified as an advantage in 2021 has become ordinary. Content and software belong together. Personalization is expected. On-demand access barely qualifies as a product decision anymore. AI is not something you sprinkle on top of an EdTech company to make it special.

The formula survived. Its scarcity did not. And that creates a much more interesting question than whether individual EdTech companies are succeeding or failing:

What happens to an industry when the things that made its best companies valuable become available almost everywhere?

The answer is starting to become visible. The capability survives. The industry around delivering it gets compressed.

First, some things really do die

2U is the obvious place to start. In 2021, it agreed to acquire substantially all of edX's assets for $800 million in cash. Three years later, 2U entered a restructuring that would cut its debt by more than half to approximately $459 million and inject roughly $110 million in new capital. Its educational programs continued without interruption. [1] The courses did not disappear. The students did not suddenly stop wanting degrees. A very large amount of financial structure disappeared.

Anthology offers an even more physical illustration. After Chapter 11, Blackboard's core Teaching and Learning business emerged in March 2026 as a debt-free standalone company. Anthology's SIS and ERP business moved to Ellucian. Encoura acquired its engagement and student-success portfolio. [2] The products survived. The corporate structure holding them together did not.

BYJU'S is messier. Debt, governance, legal disputes, aggressive expansion, and financing all played a role, so turning its collapse into an AI morality play would be ridiculous. But look at what it accumulated on the way up. BYJU'S paid $500 million for the children's reading platform Epic in 2021 and reportedly around $200 million for the coding platform Tynker. [3] Children still read. Children still learn to code. What turned out to be less durable were the ownership structures and prices we had attached to those needs. That distinction matters. Because bankruptcy is actually the easiest version of this story to see.

The coffin is the least interesting part

Bankruptcy makes good coffin photography. But compression gets much more interesting when nobody dies.

Coursera and Udemy completed their merger in May 2026. The combined company says it encompasses 290 million learners, 18,000 enterprise customers, and more than $1.5 billion in 2025 annual revenue. It expects $115 million in annual run-rate cost synergies within 24 months. [4]

I love the word synergy because it makes subtraction sound like addition. Nobody is saying people need fewer skills. Nobody is saying companies have stopped training employees. The proposition is that all of that learning no longer requires the two separate cost structures that existed before.

Now zoom out. PowerSchool completed 19 strategic acquisitions while assembling student information, enrollment, LMS, assessment, special education, finance, HR, analytics, and communications in one platform. Bain acquired it in 2024 for approximately $5.6 billion in enterprise value. [5] Instructure added Parchment's credentialing network to its learning ecosystem. [6] Learning Pool acquired WorkRamp, Elucidat, WorkStep, and Confirm between October 2025 and May 2026. [7]

These are not stories about educational functions disappearing. They are stories about fewer commercial relationships being required to buy them. Assessment can still exist. Credentials can still exist. Authoring can still exist. Training can still exist. They just do not necessarily need to exist as four different companies.

Figure 1

Four functions, four companies, or one

How to read it. The four functions are the ones the essay names: assessment, credentials, authoring and training. On the left each one is its own company, so the buyer needs four commercial relationships. On the right the same four functions sit inside one company, so the buyer needs one. Every pink function is still there on both sides; only the orange lines, the relationships, are compressed. This is a diagram of the mechanism, not a picture of any one company.

Sometimes the company stays, and the fortune disappears

Unacademy was valued at $3.44 billion in 2021. On September 1, 2026, upGrad acquired it in an all-stock transaction valuing the company at approximately $206 million. That is about 94 percent below its peak. [8] And yet this is not a story about an empty shell. Unacademy reportedly still had around $95 million in cash. Most of its businesses were profitable or close to it. The brand remained. There were no transaction-related layoffs planned. The company did not disappear. Most of the valuation did.

Pluralsight gives us another version. Vista Equity Partners acquired it in a transaction valued at roughly $3.9 billion. In 2024, existing lenders took ownership through a recapitalization that reduced its debt and provided more than $200 million in new capital. Reuters reported that Vista had marked nearly all of its approximately $1.6 billion equity investment down to zero. [9] The login page did not have to vanish for compression to happen underneath it.

Figure 2

The company did not disappear. Most of the valuation did

How to read it. Both rows share one scale in US dollars, so length is money. The dashed orange outline is what was there before; the solid pink part is what is left. Unacademy: $3.44 billion in 2021 against about $206 million in the 2026 sale, which is the essay’s “about 94 percent below its peak”. Pluralsight: Vista’s approximately $1.6 billion equity investment, which Reuters reported was marked nearly all the way down to zero; the thin pink sliver is drawn only to show that, not to a measured value. The two rows measure different things (a company valuation and one investor’s equity), so compare the drop within each row, not the rows with each other. On the right is what stayed.

Source: TechCrunch and Reuters, as cited in the essay.

That may be one of the reasons this shift is easy to underestimate. We keep looking for dead companies. Sometimes what disappeared was the valuation. Sometimes it was the margin. Sometimes it was the number of vendors. Sometimes it was the reason the category needed to exist independently in the first place.

The middle layer can disappear too

Online Program Management companies built an entire service industry around helping universities market, recruit, enroll, and support online students. Universities did, and still do, need those things. But look at what happened to the economics surrounding them.

Wiley paid $220 million for Deltak in 2012 and later added Learning House and XYZ Media. When it sold University Services in 2024, the fair-value selling price was $122.6 million, and Wiley recorded a $107 million pretax loss. [10]

Pearson paid $650 million in cash for EmbanetCompass in 2012. When it sold Pearson Online Learning Services in 2023, the consideration was deferred and tied to future EBITDA and any later sale. [11]

Figure 3

Bought in 2012, sold for less

How to read it. Each receipt is drawn on the same dollar scale, so its length is its price. Ink receipts are what Wiley and Pearson paid in 2012; orange receipts are what they got back when they sold. Pearson’s sale receipt is blank because the essay gives no price: the consideration was deferred and tied to future EBITDA and any later sale. One caveat: Wiley also added Learning House and XYZ Media after buying Deltak, and their cost is not on the 2012 receipt, so the gap between Wiley’s two receipts is not the loss. The $107 million pretax loss is the figure Wiley recorded.

Source: the Wiley and Pearson filings cited in the essay.

Online education did not disappear. The middle layer around delivering it stopped supporting the same economic weight. And once you see that, the boundaries of EdTech itself start looking strangely temporary.

EdTech can leave EdTech

Accenture acquired Udacity in 2024. More than 230 Udacity employees joined Accenture LearnVantage, bringing its content, services, and learning technology into a much broader consulting and workforce-transformation business. Accenture also committed $1 billion over three years to LearnVantage. That was not Udacity's purchase price. It was an investment in the larger capability that absorbed it. [12]

Dayforce acquired eloomi to strengthen learning inside its all-in-one HCM platform. [13] Go1 bought Blinkist to bring consumer content into workplace learning. [14] Google acquired Photomath. [15] None of those examples requires the educational function to become less valuable. Quite the opposite. It can be valuable enough to get absorbed into something bigger. Learning becomes part of HR. Learning becomes part of consulting. Learning becomes part of a general technology platform. The capability survives. The category boundary does not.

And then there is one final version of compression, where nobody even needs to acquire you.

The shortest route wins

I have written about this case before. Chegg's total revenue fell from $617.6 million in 2024 to $376.9 million in 2025, a decline of 39 percent. Academic Services revenue fell 43 percent. [16]

Chegg itself describes the problem in its annual report: students increasingly see general-purpose generative AI products such as ChatGPT as alternatives to specialized educational services. Google AI Overviews can also answer questions without sending users onward to Chegg.

Students have not stopped getting stuck on homework. They have not stopped wanting explanations. The need is still there. The route changed. Nobody has to buy Chegg, split Chegg up, or put Chegg through Chapter 11 for compression to happen. The learner can simply go around it. That may be the most consequential version of all.

Figure 4

The learner can simply go around it

How to read it. Top: the same need at both ends, a student stuck on homework and an explanation. The grey road is the old route through Chegg; the pink road is the shorter one through general-purpose generative AI such as ChatGPT, or Google AI Overviews answering without sending users onward. The moving dot is where the learner now goes. Chegg is still standing on the old road: nobody bought it or split it up. The road map is a metaphor. Bottom: Chegg’s total revenue, both years on one scale; the dashed outline is the part that fell away, 39 percent.

Source: Chegg 2025 Form 10-K, as cited in the essay.

So what is actually being compressed?

Not learning. Not uniformly, anyway. And not even necessarily the products. What gets compressed changes depending on where you look.

Debt and equity disappear. Two companies become one. Several product categories collapse into one platform. A specialist EdTech company becomes a feature inside HR software, consulting or general technology. Or the company remains exactly where it was while the learner discovers a shorter route to the same outcome.

Figure 5

The capability survives. The industry around it gets compressed

How to read it. One row for each kind of compression the essay lists, in its order, with the essay’s own examples underneath. The pink dot is the need or capability, and it is there in every “after”. What changes is the orange structure around it: the money shrinks, two companies become one, separate products move into one platform, a specialist ends up inside something bigger, or the learner takes a shorter route while the company stays put. The shapes are diagrams of each mechanism, not to scale.

This is also why the story is not simply that EdTech is shrinking. Compression creates winners. Duolingo ended 2025 with 52.7 million daily active users, 12.2 million paid subscribers, $1.038 billion in revenue and $1.158 billion in bookings. [17] When users' attention and spending concentrate into fewer places, the companies occupying those places can become enormous.

That is what compression does. It does not distribute pain evenly. It removes space. Which brings me back to that roadmap from 2021. We thought we had identified the characteristics of the companies that would win the next generation of EdTech. Content plus software. Personalization. AI. On-demand access. We were right about almost all of it. What we got wrong was assuming those capabilities would remain scarce enough to support the same amount of independent industry around them.

Five years later, the question is no longer whether an EdTech company has the winning formula. The question is whether that formula still needs a company. The future of learning may look remarkably similar to what we predicted in 2021. It may just contain far less EdTech.