The State of Headless Commerce in 2026: Shopify Hydrogen vs. Medusa vs. Saleor vs. BigCommerce

What 'headless' means in 2026

Headless commerce decouples the storefront from the commerce backend. The storefront is a custom Next.js, Astro or Remix application; the backend exposes products, carts, checkout and orders through APIs. The benefit is design and performance freedom plus the ability to compose your stack — pick a CMS, a search engine, a payment processor — without being trapped in a monolithic platform's templating system.

The four backends we actually ship

Shopify Hydrogen

Hydrogen is Shopify's official Remix-based React framework for headless storefronts, hosted on Oxygen (Shopify's edge runtime). The pitch is operational simplicity: cart, checkout, payment, fraud, tax and PCI scope all stay inside Shopify's existing infrastructure, while you get full design control on the storefront. In 2026 it is the most-recommended starting point for DTC brands shipping a new flagship in under three months.

• Strengths: fastest time-to-launch, zero PCI burden, native checkout extensibility, generous Storefront and Admin APIs, Oxygen edge hosting included.
• Weaknesses: locked to Shopify checkout (you can extend it but not replace it), Shopify Plus required for B2B parity, ecosystem of paid apps adds up.
• Real cost: Shopify plan $39 to $2,300/mo + transaction fees (3rd-party processors) + apps. The hidden cost is the apps.

Medusa

Medusa is the leading open-source commerce backend, written in TypeScript on Node.js with a strong modular architecture. Version 2.0 (2024) shifted to a fully composable module system, meaning you replace cart, pricing, inventory or fulfilment without forking. We ship it with a managed Postgres, the Stripe module and the Stripe Tax module, and a custom Next.js storefront. It is the platform we reach for when checkout has bespoke logic.

• Strengths: full ownership, no per-order fees, modular architecture, EU-friendly self-hosting, strong B2B and marketplace patterns, active commercial backing (Medusa Cloud).
• Weaknesses: you operate the database, the workers and the queues; ecosystem of paid extensions is smaller; requires real engineering to run at scale.
• Real cost: hosting (Postgres, app, workers) typically $200 to $1,500/mo at moderate scale; engineering hours dominate the TCO.

Saleor

Saleor is a GraphQL-first commerce platform written in Python and Django, with a pure GraphQL API and a strong multi-channel and multi-warehouse model. It is the platform we recommend when the deal is enterprise B2B — multiple price lists, complex tax rules, regional channels, multi-warehouse fulfilment — and the customer is not ready for the price tag of commercetools.

• Strengths: GraphQL throughout, robust multi-channel, B2B-grade product modelling, App Bridge architecture for extensions, clean event webhooks.
• Weaknesses: smaller ecosystem than Shopify or BigCommerce; the Cloud product has a meaningful price step from open-source self-host.
• Real cost: Cloud from $1,000/mo to $10,000+ at enterprise tiers; self-hosted scales with infra cost and a Python-comfortable engineering team.

BigCommerce

BigCommerce is the under-appreciated SaaS choice for mid-market US brands. It exposes both REST and GraphQL Storefront APIs with no transaction fees on top of the platform plan, supports multi-storefront natively, and has stronger built-in B2B than Shopify's standard plans. In 2026 it is the rational choice for a US brand doing $5M to $200M annual GMV that wants managed PCI without Shopify's app sprawl.

• Strengths: no platform transaction fees, robust B2B without an upcharge, multi-storefront, strong native search and catalogue.
• Weaknesses: weaker brand and ecosystem outside the US, fewer headless reference architectures than Shopify, customisation limits compared with Medusa or Saleor.
• Real cost: Pro tier from ~$400/mo, Enterprise quoted at $20k+/yr; payment processing is your contract, not theirs.

Decision tree we actually use

1. Are you launching a DTC brand in under three months and doing under $20M GMV? Hydrogen.
2. Do you need full control of checkout logic, custom marketplaces, or zero per-order fees? Medusa.
3. Are you B2B with multi-channel, multi-warehouse, multi-price-list complexity but not yet at commercetools scale? Saleor.
4. Are you US mid-market with predictable scale and a preference for managed PCI without app sprawl? BigCommerce.
5. Are you over $200M GMV with global complexity and an internal commerce engineering team of ten or more? Look at commercetools.

What we ship the storefront on

For all four backends our default storefront stack in 2026 is Next.js 15 on the App Router with React Server Components, Tailwind CSS, GSAP for premium motion, and a thin product-grade design system. We do not use Hydrogen's React framework on Medusa or Saleor; we use plain Next.js and call the backend's API directly. The framework lock-in is not worth the small velocity gain.

Performance: a real benchmark

Numbers are from a same-design Next.js storefront we built across the four backends in late 2025 with identical product catalogues (~3,000 SKUs). Differences in TTFB are dominated by API gateway latency and edge proximity rather than framework speed.

What changed in 2025-2026

• Shopify made Hydrogen the default for new headless customers and shipped Oxygen pricing parity with non-headless storefronts.
• Medusa 2.x stabilised the modular architecture and Medusa Cloud went GA, removing the operational burden for teams who did not want to self-host.
• Saleor pushed App Bridge as the canonical extension model, deprecating the older webhooks-only pattern.
• BigCommerce shipped a new GraphQL Storefront API and improved multi-storefront support — narrowing the gap with Shopify on the storefront API surface.

Migration realities

Re-platforming is the largest commerce engineering project a brand will run. Budget six to nine months for any of these migrations on a non-trivial catalogue. The dominant cost is data migration (products, customers, orders, subscriptions) and parity testing on tax and shipping. The platforms differ at launch; the pain afterwards is mostly operational, not technical.