Shopify Hydrogen's developer preview picked up six capabilities on August 18, 2026, and read as a group they are not really new features. They are a bill being settled. Cart and customer session continuity, analytics on every navigation, a Shop Pay button that paints before JavaScript runs: a Liquid theme has done all three for years. Which makes this the clearest published list yet of what headless has actually been costing premium merchants, and it shifts the honest answer on when a brand should leave Liquid.
What the August 18 Hydrogen update shipped
Six capabilities, and every one of them is commerce fidelity rather than developer ergonomics. Cart session attribution keeps the cart aligned with the Customer Account session, so a shopper who is already signed in lands on checkout with their stored details rather than a guest flow. Cart attributes let you attach order-wide data such as a gift message, plus per-line data such as engraving text bound to the line it belongs to. Page view events now fire on every navigation, which is how Liquid themes have always reported; a Hydrogen storefront previously reported the first load only. The Shop Pay button arrives painted and clickable before Shopify's hosted script loads, where it used to render blank until that script arrived. Local HTTPS lets you exercise Customer Account login on a trusted local hostname with no tunnel and no hand-rolled certificate. And logging is pluggable, with configurable levels and a route for warnings and errors into an external observability platform. All six are listed in Shopify's August 18 changelog entry.
Why every item on that list is a headless tax
Because a Liquid theme gets all six for free, and a headless storefront has to re-earn them one at a time. That is the part missing from most headless pitches, which argue on speed. Speed was never the expensive bit. Commerce fidelity is, and three of these six were quietly costing real revenue.
Session continuity is the worst of them. A signed-in shopper dropped into a guest flow at checkout has to retype an address they already gave you, on the single screen where friction converts directly into abandonment. Analytics is the second. A storefront that reports one page view per visit instead of one per navigation does not just undercount traffic, it feeds a broken attribution model, which means paid budget gets allocated from numbers that are wrong in a direction nobody audits. Third, a Shop Pay button that renders blank until a hosted script lands is a layout shift plus a trust gap on the highest-intent control on the page. None of these show up in a Lighthouse score. All of them show up in revenue.
The pattern generalises past Shopify. When you decouple a storefront from its commerce platform, you inherit every integration the platform used to handle silently, and you find them one at a time in production. We wrote down where that line sits for checkout specifically in our Shopify headless checkout guide, and the short version has not changed: keep checkout on Shopify, and budget for the fidelity work nobody quotes.
Hydrogen preview vs Liquid themes in 2026
Liquid still wins for most premium merchants this quarter, but the lock-in objection to Hydrogen is gone. The preview is a framework-agnostic SDK rather than a framework. It drops the React Router requirement, installs as @shopify/hydrogen@preview, and runs on Next.js, SvelteKit, Nuxt, SolidStart, Astro or React Router, per Shopify's preview documentation. The July 30, 2026 update added Vue through a @shopify/hydrogen/vue entrypoint, along with Suspense cart reads so the cart streams in after the shell renders. Runtime choice opened up too: the current generation of Hydrogen expects Oxygen or Node, the preview does not. Shopify also bundles 16 skills for coding agents building storefronts, which tells you where they think this gets used.
That is a genuinely better deal than Hydrogen 2022, when adopting it meant adopting Shopify's framework opinions wholesale. But it is still labelled a developer preview with no published general availability date, and the August 18 list is itself the argument for waiting: a stack still adding session continuity is a stack whose commerce surface is incomplete.
So the criteria we actually use. Stay on Liquid when your storefront is catalog, cart and content, when your team is two people, or when your differentiator is merchandising rather than interface. Go headless when the storefront needs interaction a theme cannot express, a configurator, a realtime scene, a stateful flow that survives navigation, which is the column our motion and 3D work sits in. Go headless when the storefront serves several brands or markets from one codebase. The test is never speed. A well-built Liquid theme and a well-built Hydrogen storefront both serve a category page in under a second, exactly as we found comparing Next.js 16.3 against Hyvä on the Magento side of the same question.
What we would change this quarter
If you run Hydrogen in production today, audit the analytics gap before anything else. Compare reported page views against sessions, and if the ratio sits near one, you have been running on first-load reporting and every trend line you compared to your old Liquid numbers is wrong. Re-baseline before you touch spend. Then instrument the signed-in path to checkout and count how often an authenticated shopper hits a guest form, because that number is a conversion bug wearing an engineering ticket. Move the Shop Pay button to the new rendering path in the same sprint. It is a small change on the highest-value pixel you own.
If you are on Liquid and this update makes headless tempting, do not migrate a live storefront onto a preview. Prototype instead. Take one route that Liquid genuinely cannot express, build it against @shopify/hydrogen@preview in the framework your team already knows, and measure the maintenance rather than the demo. The framework-agnostic SDK makes that prototype cheap in a way it has never been, and cheap prototypes are how you make this decision with evidence instead of a vendor deck.
Either way, the decision has not moved to the framework layer. It is still about which team owns the storefront for the next three years, which is how we scope every headless commerce engagement and why we keep it in the same e-commerce engineering practice as our Liquid and Hyvä work. The stack is an outcome of that answer, not an input to it.