
Ruby on Rails is our core stack. We have built products on it since 2016, and we believe it is one of the best choices for most web products in 2026 (we explain why in Is Ruby on Rails Still a Smart Bet in 2026?). That is exactly why we want to be clear about where it does not fit.
Choosing a technology is a business decision that will shape your hiring, costs and speed for years. A partner who recommends the same tool for every problem is selling, not advising. This guide shows the cases where we would steer you away from Rails, and the common objections that are not good reasons at all.
Key takeaways
- Rails is the wrong primary choice for ML-heavy products, extreme real-time systems, CPU-intensive processing, mobile-first apps with a thin backend, and teams locked into another ecosystem.
- Most "Rails is slow" and "Rails is dead" arguments are outdated. Shopify's Rails monolith peaked at 31 million API requests per minute on Black Friday 2025.
- The real trade-offs are talent and specialization: Ruby developers are fewer and well paid, and some workloads belong in Python, Elixir or Go.
- The best answer is often hybrid: Rails for the business application, a specialized service for the one part that needs it.
Six cases where Rails is the wrong tool
1. Your product is machine learning at its core
If the main value of your product is training models, data science or heavy numerical work, the core should be in Python. That is where the ML libraries, research code and specialists are. Rails can still run the customer-facing application around it, but forcing ML work into Ruby means fighting the ecosystem every day.
Note the difference: most "AI products" for businesses do not train models. They call AI models through APIs and wrap them in workflows, permissions and billing. For those, Rails works very well.
2. You need massive real-time concurrency
Chat platforms, multiplayer games, live trading or IoT systems with millions of open connections need a runtime built for concurrency. Elixir and Erlang were designed for this: Discord, for example, scaled its Elixir systems to millions of concurrent users (Discord). Rails handles real-time features such as notifications and live updates well, but it is not the right engine when real-time is the whole product.
3. The workload is CPU-heavy
Video encoding, large-scale data transformation, search indexing or high-frequency calculations need raw processing speed. Go, Rust or specialized services do this better and cheaper. Ruby 4.0 is much faster than older versions, but it is still optimized for developer productivity, not for squeezing every cycle out of a CPU.
4. It is mobile-first with a thin backend
If 90% of the work is in an iOS and Android app and the backend only stores some data, a backend-as-a-service or a lightweight API may be enough. Rails shines when there is real business logic on the server: accounts, payments, workflows, admin tools and integrations.
5. Your team and hiring market are built around another stack
Technology is also a people decision. If your in-house team are strong TypeScript or .NET engineers, switching them to Rails has a real cost. Ruby developers are also fewer than JavaScript or Python developers, and they are well paid: in the 2025 Stack Overflow survey, Ruby showed the highest median salary of all languages (analysis of the survey data). If you plan to build a large in-house team in a market with few Rails engineers, factor that in.
6. It is a static site or a simple content website
A marketing site, a landing page or a blog does not need an application framework. A website builder or a static site generator is cheaper to build and host. Rails starts to pay off when users log in and do things.
Quick decision table: Rails or something else?
Use this as a starting point, not a verdict. Most real products combine several rows.
| Project type | Rails as the main stack? | Better primary choice |
|---|---|---|
| SaaS, B2B platform, marketplace | Yes, a strong fit | — |
| Internal tools, CRM, admin systems | Yes, a strong fit | — |
| MVP that needs to reach revenue fast | Yes, a strong fit | — |
| Business app with AI features via APIs | Yes, a strong fit | — |
| ML model training, data science core | No | Python |
| Chat, live multiplayer, millions of open connections | Only around the core | Elixir/Phoenix, Erlang |
| Video processing, heavy computation | Only around the core | Go, Rust or managed services |
| Mobile-first app with a thin backend | Optional | Backend-as-a-service or a lightweight API |
| Marketing site, blog, landing page | No | Website builder or static site generator |
Myths that are not real reasons to avoid Rails
Some objections come up in almost every first call. They sound reasonable, but they rest on information that is years out of date.
"Rails doesn't scale." Shopify's Rails monolith peaked at 31 million API requests per minute and 53 million database reads per second on Black Friday 2025 (DHH). Your product will hit business limits long before it hits Rails' limits.
"Ruby is slow." Ruby 4.0, released in December 2025, ships with a new just-in-time compiler (ruby-lang.org). For a typical web application, response time depends far more on database queries and caching than on the language.
"Rails is dead." Rails World 2026 was the largest Rails conference yet, with more than 1,000 developers from 58 countries (Rails Foundation). Rails 8.1 shipped in October 2025, and work on 8.2 is under way.
"Rails isn't for AI." Most business AI products call models through APIs, and Rails handles that well. Rails is also one of the most convenient stacks for AI coding agents: in the Rails Foundation's 2026 benchmark, the best model solved 92% of real Rails tasks (Rails Foundation).
"We'll need a separate frontend team." Not necessarily. With Hotwire, the same Rails team builds interactive interfaces without a separate JavaScript application.
The usual answer: Rails plus the right specialist service
The choice is rarely "Rails or not". In most serious products, Rails runs the business application, and one or two specialized services handle the parts it is not built for. Two examples from our own projects:
- Healthcare AI. For a US healthcare startup, Rails and React run the product, while speech recognition and the language model run as a separate Python service on local GPUs. Each part uses the best tool, and patient data never leaves the client's environment.
- Restaurant data platform. For an international restaurant-tech company, Rails sits alongside Elixir/Phoenix services and vector search to handle large volumes of data and multilingual AI content.
This approach gives you Rails' speed and low cost for most of the product, and specialized performance only where it pays off. The key is to decide the boundaries early, during discovery, instead of discovering them in production.
A practical rule: start with Rails for the whole product unless one of the six cases above clearly applies from day one. It is much cheaper to move one proven bottleneck into a specialized service later than to build everything in a specialized stack up front.
FAQ
When should you not use Ruby on Rails? Avoid Rails as the main stack when the product is built around ML model training, needs millions of simultaneous real-time connections, is dominated by CPU-heavy processing, is mobile-first with a thin backend, or is a simple content website.
What is better than Rails for machine learning? Python. It has the ML libraries and the specialists. A common setup is a Python service for the models and Rails for the application, users and billing around it.
Is Rails good for real-time applications? For real-time features such as notifications, live dashboards and chat inside a business app, yes. For products where real-time is the entire workload, such as large chat platforms or multiplayer games, Elixir/Phoenix is usually a better core.
Are Ruby on Rails developers expensive? They are among the best paid: Ruby showed the highest median salary in the 2025 Stack Overflow survey. In return, Rails teams are usually smaller, which often makes the total cost of a product lower.
Should I migrate away from Rails? Only if one of the specific limits above is hurting your business. Usually it is cheaper to upgrade to Rails 8 and move one bottleneck into a specialized service than to rewrite the product.
Sources
- DHH: Six billion reasons to cheer for Shopify (1 Dec 2025)
- Discord: How Discord scaled Elixir to 5,000,000 concurrent users
- Stack Overflow Developer Survey 2025: language and salary trends (analysis)
- ruby-lang.org: Ruby 4.0.0 released (25 Dec 2025)
- Rails Foundation: Rails World 2026 recap
- Rails Foundation: Agents on Rails, the first benchmark report (13 Aug 2026)