A year of Shopify: what mattered and what did not

A year of platform news is only worth reading twice if it changed something you were going to do anyway. Five items did.

Theme blocks and the Horizon foundation. Theme blocks landed in the Winter ’25 Edition — reusable blocks, nestable to eight levels, usable across sections — and in the summer arrived as the foundation under Horizon and the new theme family. This is the year’s most consequential change and the least glamorous, because what it altered is not a feature list but the boundary of who can build a page. Layouts that were a developer ticket in 2024 are a merchandiser dragging blocks in 2026. If you scoped a theme rebuild before this and have not re-read the scope, do that before signing anything.

GraphQL became the only door. The REST Admin API has been legacy since 1 October 2024, and Shopify’s documentation is explicit that from 1 April 2025 all new public apps must be built exclusively with the GraphQL Admin API. If you are commissioning an integration, that is now a requirement rather than a preference, and anyone quoting you a REST build for something new should be asked why.

checkout.liquid finally ran out of runway. The information, shipping and payment steps went first; the thank you and order status pages were sunset for Plus stores on 28 August 2025, with script tags on those pages sunset for non-Plus stores on 26 August 2026. Plenty of stores discovered in August which of their tracking pixels and post-purchase upsells lived in a template that no longer exists.

Duties collection in checkout, on every plan. Previously Plus-shaped, now broadly available, along with the option to include duties in displayed prices. For anyone shipping cross-border whose customers currently meet a courier invoice at the door, this was the cheapest genuine improvement of the year, and it fixes a repeat-purchase problem rather than an aesthetic one.

The automatic discount limit went from 5 to 25. It is one line in the Winter ’25 Edition and it removes a whole category of workaround. Every merchant who was rotating discounts on a spreadsheet, or paying an app to sequence them, can stop. Shopify Functions also gained support for carts of up to 200 items in the same release, which quietly raises the ceiling for anyone doing bundle or B2B logic.

A fair amount of the year was real engineering with no business in a mid-market roadmap.

Checkout in Roblox and stablecoin payments. Both genuinely interesting, both answers to questions almost nobody in the $2M–$50M range is asking. If you sell to an audience inside Roblox you did not need an Edition to tell you.

AI theme and block generation. Describe a block, get a block. It is a decent prototyping tool and we would not put its output into a production theme without a developer reading every line of it. Generated markup nobody designed is the same maintenance debt as inherited agency code, only newer and with less excuse.

Storefront web components and the next-generation developer platform. Promising, early, and the sort of thing that changes what a brochure site can do in a year or two. Watch it. Do not stake a launch on it.

Sidekick. It got broader and better all year. Try it, because trying it is free, and do not restructure a process around it yet.

The pattern in that list is worth naming: features that make a great keynote demo are usually the ones with the fewest existing stores waiting for them. The changes that actually move money — a discount ceiling, a duties toggle, a block model — are boring enough to miss on a first read.

Both Editions carried performance claims: cart request speeds improved by up to 50%, accelerated checkout buttons faster by a stated percentage, admin page loads a quarter faster, and in the summer a checkout and admin both described as meaningfully quicker. These are Shopify’s own measurements of Shopify’s own products, published in a marketing document. That does not make them false. It makes them unverifiable by you.

The useful way to treat a vendor figure is as a direction to check against data you own. If checkout got faster, it shows up in your funnel and you have the numbers. If the admin got faster, your team noticed without being told. A project defended by a percentage nobody on your side can reproduce is a project with a hole in the middle of it.

The genuinely checkable claims are the counts and the lists — payment methods added, countries added, carriers added. Those matter if and only if one of them is yours, and the Edition names them. Read the list, not the total.

Read a few years of these in sequence and the platform’s two habits are obvious. It absorbs things merchants used to buy apps for — duties, store credit, discount sequencing, performance reporting, returns — and it moves customisation out of arbitrary code and into surfaces with declared boundaries: theme blocks instead of edited Liquid, Functions instead of scripts, checkout and customer account extensions instead of checkout.liquid.

Both directions are good news if your store sits close to platform defaults, and bad news if your competitive advantage was a clever hack in a template. The August sunset was the year’s clearest demonstration: stores that had moved to extensions noticed nothing, and stores that had not spent a fortnight in a panic.

So the posture that pays is not "adopt everything". It is to keep as much of the store as possible expressed in the parts of the platform Shopify is actively improving, so that when the platform ships the thing you were about to commission, you can simply switch it on. Every custom workaround you carry is a migration you have already agreed to do, on a date somebody else will choose. We scope [Shopify builds](/services/build) that way on purpose: fewer apps, fewer bespoke mechanisms, more of the store sitting on surfaces with a documented future.

• Shopify store development
• Reading the Summer Edition: what to act on and what to ignore
• The Admin GraphQL API: pagination, cost and bulk operations