Reddit, Inc. RDDT

Key data

The business

Reddit runs a network of user-created communities, one for nearly every interest, hobby, city, illness, and video game on earth. People post questions and answers, argue, and vote each other's replies up or down, so the best answers rise. Nobody is paid to write any of it. Reddit sells two things off that pile: advertising slots inside the feeds, which is most of the money, and licensed access to the text itself, sold to companies training AI models, which is smaller but higher-margin and growing. The moat is the archive: twenty years of human conversation a competitor cannot buy or rebuild, and that both advertisers and AI firms now need. Advance Publications, the Newhouse family's media company, controls the votes through a second class of shares, so public holders own the economics but not the steering wheel. The thing a user actually touches is free, which is why the whole model rests on selling attention and selling the words.

The numbers

Five quarters show a business that bent sharply upward, dipped on seasonality, then bent again.

The Q1 dip is normal, ad budgets fall after the holiday quarter. Q2 is the tell: revenue up 61% against a year ago and up 21% sequentially, the fastest since listing, with earnings landing about a third above what analysts modeled. The only recent stumble was Q3 2025, when profit came in below expectations after a Google algorithm change thinned logged-out traffic. That scare was the exact thing to watch coming into this quarter, whether the borrowed, search-fed audience would crack the growth, and it has not bitten again since, the audience recovered and monetization reaccelerated.

Revenue compounded about 49% a year from 2022 to 2025. The deep 2024 loss was a one-time stock charge taken on the market listing, not the operating business. 2025 was the first profitable year, and the first half of 2026 has already nearly matched all of 2025's profit.

The engine underneath is that Reddit barely spends to grow. It rents its computing rather than building data centers, so capital spending was $6.7M last year against $2.2B of sales, and free cash flow ran $684M, about 31 cents on every revenue dollar. That is the rare software shape, growth that funds itself. The catch is the share count, up 24% in three years, because stock pay, roughly 12 cents of every revenue dollar, funds the staff, so per-share gains lag the company's. What this memo believes and the multiple does not: the market prices Reddit as an ad-cycle stock, but licensing the archive to AI firms is a second, less-correlated stream it barely credits. The print that settles it is the "other revenue" line breaking out of its low-teens share of the total.

Management

Insiders sold and never bought: 70 sales, about $44M, over the past year, roughly 0.15% of the company, led by COO Jennifer Wong's near $8.7M across a single August day. The filings here do not separate scheduled 10b5-1 sales from discretionary ones, so read the cluster as routine post-lockup diversification, not a signal, until the Form 4 detail says otherwise. CEO Steven Huffman took a $193M stock-and-option award on the 2024 listing, the number that defines his pay for years, while his cash since is ordinary, $2.6M in fiscal 2024 per the most recent proxy data, well under 1% of profit. No buyback and no dividend, the right restraint while the shares are dear and the cash is young. Founder-controlled through Advance's supervoting stock, aligned on the long horizon, unaccountable on the short one.

How it fails or surprises you

The search-box landlord. Google sends a large slice of Reddit's traffic and pays to license its data. Alphabet's AI answer boxes are already cutting clicks across the web, and the licensing deals renew on someone else's calendar. If either the referral flow or a renewal turns, reported users and revenue fall together, fast, the way they wobbled in the third quarter of last year. Watch logged-out user growth each quarter.

The multiple against the users (the number the read explains least). Sales grew 61%, yet the stock sits 45% below last year's high, which says the market is already discounting a user slowdown the income statement does not yet show. If US daily users, the ones who carry the profit, flatten for two quarters while the shares hold near 36 times earnings, the stock re-rates down even as revenue climbs. That gap is this memo's weakest spot.

The monetization gap (right tail). Reddit earns a fraction per international user of what it earns per US user, and its ad load is light beside Meta or Pinterest. If it closes even part of that gap while AI licensing grows into a real second line, earnings run well past today's estimates and the 34-times forward figure looks cheap in hindsight. First sign: international revenue per user rising two quarters running.

Closing thoughts

Quarterly user counts and the licensing revenue line tell you which way this breaks. The fatter tail points up, because the monetization gap is real, documented, and largely unpriced, while the downside, a Google-driven traffic shock, shows early in the logged-out user line and has so far recovered each time it appeared. What is genuinely at risk is the multiple, not the business. At 36 times earnings, one stalled user quarter can cost you 30% before profit itself ever falls. Call that judgment, not a probability.

The bet is still that Reddit stays where people go for a human answer, and that it earns more from each of them over time. It breaks if Google throttles the traffic it feeds Reddit and the archive-licensing money does not grow fast enough to replace it. The one pair of numbers that tells you first: logged-out user growth against the "other revenue" line, quarter by quarter. If owned users keep growing while licensing scales, the borrowed-audience worry stays a footnote. If not, this was an ad-cycle stock priced like a compounder.