A position from Slawa, CTO of Next Levels. Commentary, not a neutral comparison.
Imagine you are moving out of your rented flat. Your furniture comes with you. The paint on the walls does not. Nor does the built-in wardrobe you had fitted yourself. The smart-home wiring you installed stays in the plaster. After three years of renting, what you are left with is a list of your belongings. Not the flat.
That is your Shopify shop.
A disclaimer up front: Next Levels’ focus is Shopware. That is exactly why this is a commentary and not a ‘Shopware vs. Shopify’ comparison with a feature matrix. Read what follows as the position of an interested party who knows the market from the other side.
My point in one sentence: In 2026 there is no longer a good economic argument for launching a new shop on Shopify. The last argument that really carried weight (‘Shopify is cheaper and faster to go live than anything else’) collapsed in the year of agentic development. What remains: a rented shop window with limited differentiation, hard platform lock-in and a cost structure that gets worse as you grow.
Three observations on this. From the most obvious to the most structural.
Hall of Fame: five shops, one look
Try the test. Open five of the best-known D2C shops in five tabs: Allbirds, Gymshark, Brooklinen, Olipop, MVMT. All of them run on Shopify Plus. Cover up the logos. Which shop belongs to which brand?
Promo bar at the top. Full-bleed hero with a lifestyle photo and a three-word promise. Below it, three columns with an icon, a headline and one sentence. Product grid. Sticky cart drawer. If you recognised a brand, it was down to the logo. Not the shop.
This is no coincidence. Shopify lists more than twenty free themes. Under the bonnet they are presets of exactly two cores: Dawn (2021) and Horizon, the new default theme for every newly opened shop since May 2025. Both optimise for a visual default that demonstrably converts. Converts the same way for everyone. That is the business model: you are not supposed to change very much. The closer you stay to the standard, the less the next theme update breaks your checkout and the more smoothly the app ecosystem on top of it runs.
Shopify’s mission statement reads, word for word: ‘Make commerce better for everyone.’ It is meant seriously. And that is exactly the problem. For everyone means: for your competitors on the same preset, with the same hero layout and the same cart drawer. A platform that makes commerce equally good for everyone makes it distinctive for no one.
Starting with the standard is fine. Staying on the standard means your brand looks like 50,000 others, including your competitors. Leaving the standard means custom Liquid, fragile section files and, after the next Online Store update, a few days of rework because the schema has changed. On Shopify, differentiation costs disproportionately. The more expensive it gets, the less often it happens. That is why the shops look like sisters.
Vendor lock-in: the shop is not yours
Vendor lock-in is a term that gets waved away in sales meetings (‘everyone has lock-in’). With Shopify it is structural, not a matter of degree. And it is not where most people think it is.
Liquid, Shopify’s templating language, really is open source. You can look at it on GitHub and use it in Jekyll, which means in any GitHub Pages site. What is not open source is everything that operationally turns your Shopify theme into your Shopify theme: the section schema, the theme editor bindings, the settings and app block system, the Online Store 2.0 conventions, the entire CDN and rendering path. Your theme is not ‘just Liquid’. It is Liquid plus a stack that exists only in Shopify’s runtime. When you leave, the theme stays behind.
You have just been able to watch live what it feels like when the landlord renovates: 26 August 2026 was the deadline for all non-Plus shops to upgrade their thank-you and order status pages to the new version. Additional Scripts are not supported on the new pages, and since that date script tags no longer run on the order status page. Every conversion tag, every affiliate pixel and every piece of custom logic that had been running there for years had to be migrated to the new, sandboxed extensibility framework by then. Shops that missed the date still see the upgrade notice in their checkout settings, and visual customisations there have no replacement until the store upgrades. This is not malice; there are good technical reasons for it. But remember the mechanism: a platform decision in Ottawa puts a deadline in your calendar, and you rebuild. Not because your shop needs it. Because your landlord decided so.
The checkout itself has been the toughest sticking point for years. Shopify has gradually opened up access through Checkout UI Extensions; post-purchase extensions and limited customisation via approved apps now also work on Standard and Advanced. Truly in-depth customisation of the information, shipping and payment steps remains exclusive to Plus, and even there it stays within narrow guard rails. If you need B2B tax logic or an alternative confirmation chain, you will get stuck. Two delivery addresses per order are simply not possible. Full stop.
Apps are your second lock-in. Every function outside the core is a separate subscription. Every data model lives in someone else’s database. Any app vendor can double its prices tomorrow or be sold. If you subscribe to 25 apps, you have 25 mini SaaS contracts. To put it concretely: what do you do if Klaviyo puts its prices up by 50 per cent next quarter and your 60,000 contacts, including segments, flows and purchase history, are stored there? Cancelling is not an option. That is precisely the contract you sign with every app installation.
The data may be exported. Products, customers and orders via CSV or the Admin API. What is not exported: everything that operationally makes your shop your shop. Theme logic. Discount scripts. Flow automations. App configurations. Metafield structures. Checkout customisation. That is the real lock-in. Not the data. The machinery around it.
That is why replatforming away from Shopify is rarely a 6-week project. It is a full replatforming.
Renting is legitimate. It is not a moral question. It is an economic one. How much do you invest in a property that does not belong to you?
The cost argument collapses
For years, the rational defence of Shopify was simple: ‘Your own shop system is too expensive and too slow to go live.’ That is no longer true. Not slightly less true. Structurally no longer true.
What Shopify costs when you do the maths honestly. A mid-sized D2C brand on Shopify Advanced (list price $299/month on annual billing, $399 on monthly billing) plus the usual 15 to 25 apps ends up at $1,000 to $2,500/month before the first custom theme developer comes on board. Which apps? In a typical stack: Klaviyo for email (from ~$200/month depending on list size), Recharge or Bold for subscriptions ($99–$500/month plus transaction fees), Rebuy for personalisation ($200–$500), Loox, Yotpo or Judge.me for reviews ($50–200), Aftership for tracking, Gorgias for support, plus half a dozen smaller tools for translation, bundles, custom fields, wholesale locking and review sync. Every vendor has its own pricing model, and every one has its own cost curve pointing upwards.
Then there are the payment fees. And this is where the hidden penalty sits:
If you use Stripe, Adyen or another payment provider instead of Shopify Payments, every transaction costs you an additional 0.6 to 2.0 per cent, paid directly to Shopify on top of what you already pay the payment provider. On Basic it is 2.0%. On Advanced it is 0.6%.
That is not a service fee. It is a competition tax that Shopify uses to tie you to its own payment provider. On a GMV of €5 million a year, that is between €30,000 and €100,000 extra, depending on the plan, if you do not switch to Shopify Payments.
With a GMV in the single-digit millions, the running stack quickly adds up to a six-figure annual sum. Ongoing. Every year. Without you building up a single euro of ownership.
What is shifting on the other side. Self-owned implementations (Shopware 6, Sylius, Medusa) never had their economic problem in hosting or licences. Those sit in the low three-figure to low four-figure euro range per month, depending on load and the level of managed service. The problem was hourly rates. The industry classic: the custom feature quoted at €8,000 really costs €12,000. Half of the time goes into boilerplate, wiring, tests and code style discussions.
These are exactly the hours that agentic development clears away. Today an experienced developer writes less code and reviews more: they run a specification against an agent system (Claude Code, Cursor, Aider) and refine the result in a review loop. That is the whole change. And it is enough.
For a long time, the research was the best counter-argument. It has just turned, and that deserves to be said just as honestly. The much-cited METR study from July 2025, which measured experienced open-source developers using Cursor and Claude 3.5/3.7, found a slowdown of 19%, even though the developers themselves felt they had been sped up. In February 2026 METR measured again with a revised study design: the new cohort of 47 developers came in at just −4%, with a confidence interval reaching up to +9%. METR’s own assessment: it is now likely that AI tools speed developers up rather than slow them down, even if selection effects obscure the exact size of the effect. The sober picture remains domain-specific: in theme composition, CRUD UIs, storefront front ends, test generation and wiring boilerplate, double-digit percentage gains are realistic. In greenfield architecture decisions, hard domain modelling and production debugging, much less so.
In short: the question is no longer whether agentic development is faster. It is where. And the answer lines up almost exactly with what makes a shop project expensive.
Precisely the domains in which agentic development delivers (storefront, theme, form logic, state wiring, standard back-end plumbing) are the terrain on which Shopify sells its simplicity story. A feature for which a Shopware project in 2022 still needed €8,000 of implementation and €2,000/year of maintenance comes to €4,000–5,000 and €800/year today in a clean model calculation. Set that against a Shopify world in which each of these features is an app subscription whose costs rise every year, and the break-even between ‘your own shop system’ and ‘Shopify’ shifts to the left by an order of magnitude.
Half of the hours a custom feature ate up in 2022 no longer go into boilerplate today. They disappear.
What you should do instead
Three sober recommendations, depending on where you stand today.
If you have nothing yet and need to go live quickly, Shopify is still a legitimate first stage. But plan it honestly as a stage, not as the destination. Keep the custom logic as thin as possible. Avoid any app whose data model you cannot export straight away. Note on day one when the TCO comparison will be due: as a rule of thumb at a six-figure annual GMV, or before the Plus upgrade at the latest.
If your shop is working and making money, recalculate the TCO over five years. Including app subscriptions, payment margins, the Plus upgrade path and custom theme maintenance. Compare that with a Shopware implementation costed today with agentic development workflows. The figures look different from 2022, and the gap grows every year.
If you need B2B, international tax logic, complex product ranges, a brand world of your own or serious headless, Shopify is not the best tool here. It never was. Today it is simply less often the cheapest one as well. Shopware 6 with composable front ends is the more honest anchor for everything above the classic D2C product range.
This commentary is not about which system comes out on top in which row of a feature matrix. It is about the fact that the economic foundation on which Shopify was the first choice for so long is being pulled away right now. By a shift in productivity that initially has nothing to do with e-commerce and changes it all the same.
Anyone who still reaches for Shopify automatically in 2026 is making a decision with the arguments of 2020.
Slawa, CTO of Next Levels.