# oura takes subscription public while the rented channel gets pricier > oura files for ipo with $240.5m in subscription revenue over nine months; meanwhile agent tokens, creator usage rights and search that sends no clicks push the bill up. - edition: Friday, September 4, 2026 (2026-09-04) - notebook: product & markets - topics: market · media · brand · ai - items: 10 from 8 sources - original: https://tonho.wtf/en/daily/2026-09-04-produto/ - portuguese edition: https://tonho.wtf/diario/2026-09-04-produto/ - authorship: written by an llm pipeline, reviewed and translated by antonio leandro (tonho.wtf) --- Oura filed its S-1 with the SEC today, and the number that explains the filing isn't the hardware one. It sold 3.1 million rings in nine months, up 75%, but what takes it to the Nasdaq at a valuation above the $11 billion of its last round is the subscription line: $240.5 million, 121% growth, paid subscriber base doubling from 2.5 to 5 million. That's a fifth of revenue growing at twice the speed of the rest. In the whole of today's dossier, it's the only item where the bill for whoever is building is predictable. The rest is variable cost going up, and going up in units nobody budgeted for. The unit cost of a token fell and companies' AI bills went up anyway, because agents multiply calls. The creator's price didn't go up on the post, it went up on the usage rights attached to it. USA TODAY's search traffic didn't get expensive, it simply stopped arriving, and the bill came in the form of layoffs and a new org chart. Ads on X cost less because they're worth less: $710 million in the half, about 70% below 2022. Yesterday the thread was cost cut by an owner's decision (Nadir at Globo, 3,300 at Uber); today it's cost going up without anyone having decided anything. And Niederhoffer, in the reading section, is the oldest counterpoint to this: leverage is a variable cost too, and what it charges isn't money, it's the right to decide when to get out. ## brand and ip **[Spencer's buys Hot Topic](https://licensinginternational.org/news/spencers-purchase-of-hot-topic-expected-to-spur-ip-growth/)** — The licensing industry's read is that the merger doesn't add up stores, it creates a hub for emerging IP to compete with PopMart and Miniso. The asset being bought is the entry shelf that disappeared with Toys 'R' Us in 2017: in the toy aisle at Walmart and Target, if you're not Barbie, Disney or a movie studio you don't get in. That's 1,502 stores across the US and Canada, with separate buying organizations, the way Hot Topic already operates with Box Lunch. The test is now: buyers start working the supplier showrooms in mid-September for the 2027 lines, and the overlap in Halloween costumes (Spirit has strong DTR programs, Hot Topic buys from Disguise) is the first visible conflict. **[Runblock](https://thedieline.com/run-dont-walk-to-check-out-the-steamy-branding-for-the-sunscreen-designed-for-runners-runblock/)** — Sunscreen cut by behavior, not by benefit: the audience is runners, and Tess Lockhart and George Fletcher's brand system evokes heat and sweat instead of the category's pharmaceutical codes. The Dieline piece is short and works as a pointer to the case study, but the positioning decision is already in the first paragraph. The same note from Tuesday about the shape of the sandwich applies: when the category is a regulated commodity, what's left to differentiate on is the visual code and the occasion of use. ## media and attention **[USA TODAY reorganizes its audience team](https://searchengineland.com/usa-today-search-pressure-487003)** — Monica Richardson, SVP, tied the restructuring directly to search in an internal memo: "search traffic is under pressure, and platforms are increasingly keeping user experiences to themselves instead of sending them back to us". The new structure has three desks, and the one that matters is the Strategic Platforms Desk, dedicated to audience outside the company's own properties. Consultant Glenn Gabe's counterpoint is the detail that changes the decision: it could be visibility lost cumulatively to algorithm updates, and not just AI Overviews eating clicks. One diagnosis is recoverable with SEO work; the other is structural, and you reorganize the team around the wrong problem if you confuse the two. **[Usage rights are driving up the creator contract](https://digiday.com/media/marketers-say-usage-rights-are-driving-up-the-price-to-work-with-creators/)** — The Digiday piece (only partly open in the feed) argues that the pricing bottleneck isn't the upfront price of the post, it's the usage rights stuck to it: where the asset runs, for how long, with what exclusivity. Danielle Wiley, of Sway, describes the standard agency ask as a single price covering the post, 30 days of paid usage, two months of organic and 30 days of exclusivity. Anyone budgeting creators under the production cost line is getting the nature of the expense wrong: this is a media buy with a licensing contract on top. **[X shrinks in Europe while Reddit, TikTok and Pinterest make their pitch](https://www.stackedmarketer.com/news/x-shrinks-in-the-eu-as-reddit-tiktok-and-pinterest-make-their-pitch/)** — X's ad revenue in the first half of 2026 was $710 million, down 18.4% year over year and about 70% below 2022, with only 1.9% of users paying for Premium. The other three are selling positioning to the advertiser, each on the axis it has: Reddit created a 15-second goal that lifted view-through rate by 51% and completion by 71% in alpha, TikTok moved interaction into the comments (audio, polls, nine-photo carousel), and Pinterest sells itself as anti-doomscroll days after the fines on Meta and TikTok. None of them announced a feature; all three announced what they're for. **[RedVest Media, Ace Hardware's network, ties programmatic to the weather](https://digiday.com/marketing/ace-hardware-adds-weather-triggered-programmatic-ads-to-retail-media-strategy/)** — A year after launch, the hardware chain's retail media network opened up incrementality with a control group for off-site media, a Pacvue integration and campaigns that swap the message on their own according to rain, snow or air quality index. The most useful number is the one from the DoorDash partnership: 91% of people who buy Ace through the app have never set foot in an Ace store. Molly Hjelm says the network feels no urgency to announce big, and the result is a catalog that sells verifiable mechanism instead of reach promised at the upfront. ## market and capital **[Oura files for a Nasdaq IPO](https://neofeed.com.br/negocios/o-anel-de-us-11-bilhoes-oura-entra-com-pedido-de-ipo-nos-estados-unidos/)** — $1.21 billion in revenue for the nine fiscal months ended in June, up 74%; net income of $60.8 million against $1.6 million a year earlier; gross margin from 51% to 55%. The read that matters to people who build: the subscription isn't an upsell on the hardware, it's what reprices the hardware, because it changes the margin and the predictability of the flow before it changes the volume. In the prospectus the company lists Apple, Google, Samsung, Garmin and Whoop as competitors with more resources and more channel, and writes that it expects the growth rate to slow. It's the kind of S-1 where the stated risk is more informative than the stated growth. **[Priscyla Laham and the jump in the token bill](https://neofeed.com.br/videos/neosummit-ianareal-videos/priscyla-laham-da-microsoft-alerta-para-o-salto-na-conta-dos-tokens/)** — "The cost of tokens fell a lot, but consumption went up a lot," the president of Microsoft Brazil told NeoFeed, and the next sentence is that companies are surprised by how much they're spending. The previous phase she calls a corporate science fair, with $10,000 PowerPoints made on the most expensive model available. Microsoft's commercial answer is routing: Foundry with more than 11,000 models exists so the customer saves the expensive model for the complex job. On the other side of the counter the same squeeze shows up, with Microsoft Cloud gross margin at 65% and $41 billion of capex in a quarter, two thirds of it in CPUs and GPUs, short-lived assets. My read: if the number of calls per task grows with agents, per-seat pricing is a margin promise you don't control, and choosing the model per route becomes a product decision, not an infra one. ## reading **[The lessons of Victor Niederhoffer's life](https://ritholtz.com/2026/09/lessons-victor-niederhoffer/)** — Barry Ritholtz reposts Attain's roundup on the trader who blew up three times and died this year, with the rule that opens everything: "Being right is a luxury good. Solvency is a necessity." The mechanism is what matters here: leverage doesn't just amplify the loss, it hands the exit decision to the clearing broker, who owes nothing to your thesis. And Niederhoffer's own autopsy, in a 2010 interview, is that nobody sets a stop-gain: every risk framework is built around loss, and it was the run of wins that blinded him. The detail Ritholtz highlights at the end isn't about risk: after suing the CME and winning a settlement, he distributed the entire amount to clients without deducting the fees he had paid. **[What daily returns say about the real economy](https://alphaarchitect.com/daily-stock-returns-2/)** — Larry Swedroe reviews the paper "Main Street in Wall Street," by Fabrizio Ghezzi (June 2026), which decomposes the S&P 500's daily return into high-frequency noise and low-frequency signal, using a Kalman filter anchored in four macro series (payrolls, unemployment, industrial production, capacity utilization). The resulting factor has a 0.80 correlation with the monthly CFNAI and 0.74 with the Philadelphia Fed's daily index. What interests me is the negative test: the factor doesn't predict stock returns or inflation, and that's exactly what validates it as a measure of real activity. It works as discipline for product metrics — if your user health index seems to predict everything, it probably isn't measuring anything. ## stalled sources First Round Review has gone 312 days without publishing, Gurwinder 250, Calculated Risk 236 and Elad Gil 137. In smaller volume, but alongside them: SVPG (Marty Cagan) and Adjacent Possible at 26 days, Elena's Growth Scoop and Kyla Scanlon at 22, The Generalist, Sherwood News and Collab Fund at 18.