Securities Times puts it in the plainest terms the Chinese financial press allows: Taobao's "dual engine" is stalling while JD.com poaches its merchants, Pinduoduo grinds it down on price, and Douyin siphons off the impulse spend. The three-front pressure is not new intelligence — but framing it as structural loss rather than cyclical dip is new, and it matters. The companies beating Alibaba on its home turf are not winning with larger models or more compute. JD wins on next-day fulfilment. Pinduoduo wins on rock-bottom price. Douyin wins on twenty-second videos.

We argued last week that Alibaba's 380-billion-yuan infrastructure bet is the wrong trade — that its rivals do not need the moat it is building. Securities Times just provided the market mechanics supporting that read. Bain sizes the prize at RMB 1.5 trillion. A market that large, with the leading platform losing share on every meaningful metric, is a fight, not a monopoly. Alibaba chose compute. Its rivals chose execution. The org chart that told you this story came out in April. The market data is confirming it now.

NewBird AI. The stock is up 400 percent. We said last week that the shell was worth more than the brand. Wall Street just confirmed it by bidding up a rebrand rather than a product. This is the same GPU-as-a-service pivot wearing a different logo. The company that couldn't defend market share in comfortable footwear is now valued as an AI infrastructure play. The shoes are still the shoes. The market, apparently, does not care.

Japanese Cherry Blossom turns twenty this year. Bath & Body Works has built 12,000 products around it and is using the anniversary as the spine of its retail recovery. This is not nostalgia as a crutch — it is a built-in cultural calendar that requires no trend-spotting, no influencer activation, and no algorithm. The fragrance already has emotional memory baked in. TikTok Shop proved you can sell scent without smell. Bath & Body Works is proving something different: you can anchor a retail recovery on a product the customer already loves. Every DTC fragrance brand spending CAC to introduce a new smell to a cold audience should be taking notes on what twenty years of repurchase looks like on a balance sheet.

Topshop no longer has stores. What it has is a brand name, an ASOS distribution deal, and now an AI-generated catwalk that pipes directly into a TikTok live session co-hosted by Shark Beauty and Lookfantastic. The architecture is minimal and clever: Topshop supplies aesthetic authority, the platforms supply audience and transaction, and nobody builds or staffs a retail floor. This is what a fashion brand looks like when it has been stripped to intellectual property and asked to perform with the infrastructure it can actually afford.

Pinduoduo has already replaced the human host with AI drama. Topshop's version is more modest — human presenters, AI imagery — but it signals something real: live commerce in the UK is maturing beyond the influencer-reads-SKUs phase. The format now has production value and brand architecture. M&S or Next will copy this within two quarters, probably without admitting they copied it.

Reale Actives sold out on launch day and collected mixed reviews before the orders shipped. The CEO saying publicly they "prepared for a mixed response" is smart crisis pre-emption: frame the criticism as expected, defuse the narrative, protect the founder's credibility. This is a commerce event, not a beauty launch. Creator-to-brand conversions live or die in the first hour, and this one lived. The reviews will settle into something closer to accurate over the next six weeks. The sell-through won't change.

Alana Pallister sold her house to fund 300,000 units of the Blare tracksuit. Less than five months later: one million sold. The brand is now staking its next moment on Coachella. The mechanics are the same ones driving Target's $6M first-hour collab numbers: the discovery event and the purchase window collapse into each other. At a festival, the conversion happens on the phone, not in a search bar. The difference is that I.AM.GIA cannot control the Coachella crowd the way Target can time a product drop. Festival virality is less architectured than feed virality. She was right once. The house is gone. This is the bet after the bet.

5.5% of all UK retail sales pass through Amazon marketplace sellers — not Amazon itself, but the hundreds of thousands of businesses that absorbed the inventory risk, logistics complexity, and customer service burden in exchange for access to the shelf. Amazon collects rent on all of it. The FBA surcharge we covered earlier this month is the landlord raising rents on tenants with nowhere to go. The connection worth making: DoorDash is now building an apparel marketplace on its delivery fleet — same structural logic, different infrastructure. The marketplace-as-shelf is the default entry point for retail distribution now. The only open question is who owns the shelf.

The opinion piece frames AI not as a shopping improvement but as a potential disintermediator — agents that complete purchases without the browse, the discovery, the impulse that generates margin. The anxiety is legitimate but under-segmented. Agentic commerce fails most visibly when the agent replaces discovery behaviour, not just the checkout click. But AI agents will dominate replenishment and utility purchases first — socks, supplements, household staples — and leave fashion's discovery-and-delight category mostly alone longer than the pessimistic read implies.

The structural threat is more specific than "AI eats shopping." When Honor moved AI inference onto the device, the discovery surface moved with it — before the retailer, before the marketplace, before the ad. That is what the piece should be worrying about: not the agent-buys-your-socks problem, but the phone-surfaces-the-answer-before-you-open-an-app problem.

Tokyo CPI up 1.4% year-on-year in March. Clothing and footwear: barely moved. Japan's fashion deflation has structural causes — a domestic apparel industry that spent a decade training consumers to expect Uniqlo-level value, and a fast fashion layer that absorbs demand before it reaches the margin-sensitive mid-market. Fast Retailing posted record revenue this quarter on speed, not AI. Pair it with Türkiye's clothing inflation cooling fastest of any consumer category: thin fashion margins absorb cost pressure that sticks everywhere else. The pattern is consistent across markets and it is not accidental.

Massimo Dutti: ten days on Rue Froissart in Paris. Zara Man: similar format elsewhere. Inditex uses pop-ups to collect foot traffic and sales data before committing to a lease — two weeks of test costs less than a bad ten-year contract. The pop-up as market-entry data collection has moved from luxury-indie experiment to mainline retail toolkit, and when Inditex adopts a format, it is mainstream. Every mid-tier European retailer will treat the two-week pop-up as standard market-testing practice by the end of next year.

Etailment — Germany's most-read retail trade publication — frames agentic commerce not as a future scenario but as an operational preparation deadline: retailers need to restructure catalogues, data architecture, and checkout flows now, before AI agents start purchasing on behalf of customers. The German retail press operationalizes things the English-language tech press theorizes, which is always useful. Shopify already moved semantic search and agentic storefronts to the platform layer — merchants on those rails got the upgrade automatically. The retailers etailment is worried about are the ones running proprietary stacks. They have less runway than they think, and the article does not soften that point.

Onton raised $7.5 million for AI-powered e-commerce product discovery. The threat is not other startups. Shopify made discovery free at the platform layer for millions of merchants. Algolia's own survey shows AI search is the top-priority budget line for B2C retailers, meaning the incumbent vendors are also investing heavily. Yupp.ai raised $33 million and ran out of market in nine months because the AI evaluation category it was built for pivoted to agentic systems faster than the company could pivot with it. The risk for Onton is the same: not competition from the right, but the platform layer absorbing the problem entirely before the Series B closes.