WATCHconviction 3published 2026-09-05

COUR — Coursera, Inc. · 2026-09-05 · Verdict: WATCH · Conviction 3

Price $6.01 (screen close 2026-09-04; no live prices) · Mkt cap $1.59B (264.4M sh) · EV ~$767M · 4.4x FY26 guided Adj EBITDA, ~7.7x normalized owner earnings · FCF yield negative LTM · Net cash ~$822M, no debt · ADV $25.1M Sources read: 10-K 2026-02-23 (Items 1, 1A, 7), 10-Q 2026-08-05, DEF 14A 2026-05-11, 8-Ks 2026-05-18, 06-11, 06-23 (modeling deck with CFO remarks), 07-06, 07-29 (Q2 release and shareholder letter), Form 4s (12m). No call transcript in the bundle.

1. What the business actually does

Two segments. Consumer sells subscriptions (Coursera Plus), single courses, and a small degree business taking a share of university tuition. Enterprise licences the platform to employers, universities and governments (10-K 2026, Item 1). Content comes from 375-plus partners paid a revenue share, so content cost is variable and gross margin moves with mix (Item 7). Coursera closed an all-stock combination with Udemy on May 11, 2026 at 0.800 COUR per Udemy share, which is why share count is up 61.5% and all trailing consolidated numbers are stale (10-Q Q2 2026).

2. Why it is mispriced — the edge case

Real, but narrower than it looks. Q2 mixes seven weeks of Udemy into the quarter, so revenue "grew 60%" while the combined business shrank, and any screen reading consolidated tags gets nonsense. There is also a mechanical seller: Udemy holders received COUR paper they did not choose, and Insight Partners sold 1.93M shares at $5.24 to $5.41 on May 19 and 20, days after close (Form 4, 2026-05-20). Against that, this is no ignored microcap: ADV is $25.1M, Vanguard owns 9.5%, Baillie Gifford 8.6%, BlackRock 6.3% (DEF 14A 2026). The screen is wrong in both directions, counting $871.7M of cash while missing $110.5M of securities, $60.1M of July buybacks and the $100M LearnVector cheque. Corrected, net cash is about $822M and EV about $767M, not $717M.

3. Unit economics and growth

FY25 standalone: revenue $757.5M (+9%), gross margin 54.6%, GAAP operating loss $(77.4)M, Adjusted EBITDA $63.5M, company-defined Free Cash Flow $78.5M (10-K 2026, Item 7). The screen's 6.7% FCF yield used $107.2M, subtracting only $1.5M of hardware capex and missing $18.1M of capitalised software and $10.6M of content assets.

Standalone profits were falling into the merger: Adjusted EBITDA ran $18.7M, $18.0M, $15.6M, $11.2M, $13.5M from Q1'25 to Q1'26, so Q1'26 was down 28% on revenue up 9% (Q2 letter, p.10). Combined FY25 pro forma revenue was $1,547.3M and, before purchase accounting, the two together lost about $82M at the operating line (8-K 2026-06-23, slide 25). FY26 guidance of $1.220B to $1.245B as reported is $1.49B to $1.52B normalized, minus 4% to minus 2% (slide 15). Q1'26 combined was roughly flat, so the guide arithmetically requires H2 declines of five to six percent; the CFO said the comparison is "expected to face further compression in subsequent quarters."

Working: Q2 Enterprise segment gross margin 79.3% (up 960bp) and Consumer 65.1% (up 380bp) on lower content revenue shares and a platform fee introduced in January 2026, with Consumer subscription revenue up 19% and Paid Subscribers up 44% to 1.655M. Not working: Enterprise Customers down 2% to 12,107, Net Retention 91% versus 95%, combined Consumer non-subscription revenue down 24% in Q1, degrees declining again (10-Q Q2 2026; slide 13).

4. Balance sheet and capital allocation

$871.7M cash plus $110.5M securities at June 30, no debt, and the $780.8M of liabilities are all operating, mostly $501.4M deferred revenue (8-K 2026-07-29). Less July's $60.1M of buybacks and $100M for LearnVector, net cash is about $822M, 52% of the cap. The $500M authorization had $409.9M left at June 30 and roughly $350M after July; $140M was spent by July 28 at about $5.45 in Q2 and $5.67 in July, and the remainder would retire about 22% of shares. Offsets: 2025 gross equity burn was 10.5%, with 28.8M awards outstanding, 20.8M shares available and a 5% evergreen, partly checked by a new $10 share price floor for equity awards. Zero insider open-market buys in twelve months against $10.2M of sales (Form 4 summary). Say-on-pay drew 60.5% in 2025; bonuses still paid at 106.3%.

5. Management: said versus did

They beat their own six-week-old numbers. On June 23 the CFO guided Q2 Adjusted EBITDA margin to 11.0%, FY26 to 13.0%, and $80M of run-rate synergies by end-2026. On July 29 Q2 printed 14.3%, the FY target rose to about 14%, and synergies to "at least $85 million," with the full $115M now expected before end-2027. Capital allocation is weaker. On July 28 the board put $100M into LearnVector for 33.33% fully diluted, a newly formed AI company founded by Andrew Ng, who chairs that board. A special committee of disinterested directors approved it, the right process, but the substance is a tenth of the cash into a pre-product startup at a $300M post-money with no commercial agreement, only an undertaking to "negotiate in good faith" later (8-K 2026-07-29, Item 8.01). Ng already receives 50% of revenue on DeepLearning.AI content hosted on the platform (DEF 14A 2026), while the same balance sheet buys COUR at $5.45 as undervalued.

6. Valuation

Guided FY26 Adjusted EBITDA is about $173M, so headline EV/EBITDA is 4.4x. Two corrections destroy that. The FY26 figure contains a purchase-accounting benefit of roughly $10M per quarter from Q2 to Q4 that management says will moderate (slide 16 note), and it excludes stock compensation, which ran $23.8M in Q2 outside the merger line on only seven weeks of Udemy, implying $95M to $120M annualized. Normalizing to 2027 on a $1.48B base at 16% margin, less residual purchase accounting, about $110M of SBC and modest cash taxes (large NOLs, full valuation allowance), leaves roughly $100M of owner earnings, or 7.7x EV.

Probability-weighted $7.68, 28% above price. Reverse DCF: the current price implies the operating business declines about 3% a year in perpetuity at a 10% discount rate, which is close to what management already guides for 2026 and therefore not obviously wrong.

7. Catalysts and timeline

Q3 2026 results in late October are the first full combined quarter and test the $364M to $372M guide and the 16%-plus Q4 exit margin. Q4 and FY26 results in February 2027 bring the first clean FY2027 guide on a full combined base. Buyback execution against the ~$350M remaining is visible quarterly.

8. Risks and pre-registered kill criteria

  1. Normalized combined revenue down more than 6% year over year in any reported quarter, or FY2027 revenue guided below $1.42B.
  2. Net Retention Rate below 90% for two consecutive quarters, or Enterprise Customer count down more than 4% year over year.
  3. Quarterly repurchases below $40M while the stock is under $8, or the $500M authorization not substantially complete by end-2027.
  4. Any further related-party investment in LearnVector, or aggregate related-party commitments above $150M, without a disclosed commercial agreement.

9. Verdict and one-paragraph summary

WATCH, conviction 3. Coursera trades at $6.01 with about $822M of net cash, over half the market cap, no debt, and a $500M buyback that has already retired $140M of stock at $5.45, so the operating business costs roughly $767M against guided FY26 Adjusted EBITDA of about $173M. That 4.4x is not the real number: roughly $30M of the FY26 figure is a purchase-accounting benefit management admits will moderate, and the measure excludes $95M to $120M a year of stock compensation, which corrects the multiple to about 7.7x normalized owner earnings and leaves the price implying only a 3% perpetual decline. That is the problem, because normalized revenue is guided down 2% to 4% this year with the CFO saying the comparison compresses further in the back half, Enterprise customers are down 2%, net retention has fallen from 95% to 91%, and consumer non-subscription revenue is down 24%. Management has executed on cost, beating its own six-week-old margin guide by 330bp and lifting synergies from $80M to $85M. But no insider has bought in twelve months, and on the eve of the print the board put a tenth of the cash into a pre-product AI startup founded by its own chairman with no commercial agreement attached. The two quarters that show whether the subscription base stabilizes arrive by February, and they cost nothing to wait for.

Research for discussion, not investment advice. Positions and sizing are the reader's decision.

Source markdown: 2026-09-05_COUR.md · how these notes are built · every verdict tracked since publication.

Research and education only. Nothing here is investment advice or a recommendation to buy or sell any security. No price targets are recommendations; positions and sizing are the reader's decision. Past performance does not predict future results.