Home / Directory / AI SaaS tooling / MLOps & experiment tracking / Hopsworks

Hopsworks

Hopsworks is a Stockholm company that sells an AI lakehouse built around a real-time feature store, with a model registry, model serving, and GPU management for teams that build and run their own ML systems.

7.8/10
Overall Score
Recommended

Hopsworks combines a sub-millisecond feature store with a model registry and serving, runs on Iceberg, Delta, and Hudi tables, and can be self-hosted, air-gapped, or used as a pay-as-you-go service.

Best for

Teams building real-time ML such as fraud detection, credit decisions, and personalization that need fresh features served in milliseconds and want to keep data in their own environment.

Not ideal for

Teams that only need experiment logging, or small teams without the data engineering staff to run feature pipelines.

Verdict

Hopsworks began as research at KTH Royal Institute of Technology in Stockholm and now sells a platform for feature pipelines, training, and serving, with RonDB, its own database, powering an online feature store that answers in under a millisecond. It is recommended for teams running real-time fraud, credit, and personalization models that want features, models, and serving in one place, and it is a heavier lift than a team needs for simple experiment tracking.

Score Breakdown

These scores cover how well a team can log and compare experiments, how well it versions data, models, and other artifacts, how easily it runs work across clouds and clusters, and how well it fits enterprise security and audit needs, and then how the company does on innovation, communication, pricing, and contract terms.

Buyer outcomes

Experiment tracking
7.4
Artifact & registry
8.3
Orchestration / hybrid compute
7.9
Enterprise governance
8.0

Company & commercial

Innovation & product leadership
8.1
Project management & communication
7.6
Pricing
7.7
Contract fairness
7.4

Who Uses Hopsworks, Where It Fits on Price, and Its Competitors

Hopsworks sells to ML and data platform teams at banks, credit unions, lenders, betting companies, retailers, and public agencies. Its customer stories include Zalando, America First Credit Union, Clicklease, Paddy Power, and Arbetsförmedlingen, and the company says it earns more of its income in the US than in Europe. Hopsworks was founded in 2017, employs about 36 people, and reported annual recurring revenue of $2.7 million in late 2025.

A free tier covers one project, SaaS is pay-as-you-go, and Enterprise is quoted, which puts Hopsworks in the middle of this field on cost. It is most often compared with Tecton and Feast for feature stores, with Databricks and SageMaker for an all-in-one ML platform, and with Domino Data Lab for regulated enterprise ML.

Pricing

Hopsworks publishes its free and pay-as-you-go tiers and quotes Enterprise, as of November 2026.

PlanPriceWhat stands out
FreeFreeOne project with the feature store and model registry, community support, and no credit card.
SaaSPay as you goUnlimited projects, the feature store, model registry, and model serving, billed on usage with a platform SLA.
EnterpriseCustom quoteOn-premises and air-gapped deployment, a dedicated support team, custom integrations, and a guaranteed SLA.

The Field at a Glance

How Hopsworks compares with other MLOps and experiment tracking vendors we reviewed, by Overall Score and relative typical engagement cost.

6 7 8 9 Overall Score $ $$ $$$ $$$$ Relative typical engagement cost Comet Domino Data Lab Weights and Biases ClearML Union.ai ZenML lakeFS Valohai DagsHub Hopsworks 7.8
Hopsworks Weights and Biases Union.ai ZenML ClearML lakeFS Valohai DagsHub Comet Domino Data Lab

Hopsworks scores 7.8 in MLOps and experiment tracking, under Weights and Biases (8.9), Union.ai (8.3), and ZenML (8.0), and ahead of ClearML (7.7), lakeFS (7.5), Valohai (7.1), DagsHub (6.8), Comet (6.7), and Domino Data Lab (6.5). Relative cost lands mid-band for this subcategory.

How Well Hopsworks Fits Each MLOps Use Case

We rated Hopsworks against the jobs ML teams most often give a feature store and MLOps platform, drawing on its site, documentation, customer stories, and pricing page.

Use caseFitNotes
Real-time features for fraud and credit modelsStrongThe online store, powered by RonDB, serves features in under a millisecond.
Feature reuse across modelsStrongA central feature store lets teams share features across models and avoid training and serving skew.
Sovereign or air-gapped MLStrongSelf-hosted on Kubernetes, on-premises, or air-gapped, with support for AWS, Google Cloud, Azure, and OVHcloud.
Model training and servingMixedTraining with Ray and serving with KServe or vLLM are built in, but most buyers choose it first for the feature store.
Experiment trackingMixedExperiment tracking is included, though teams with heavy research workloads often keep a dedicated tracker.

Review Excerpts

These excerpts draw on Hopsworks’ site, public customer statements, and public review sites. Paid reviews and pay-for-play sites were excluded.

What people like

“Hopsworks allowed us to cleanly separate features, models, and infrastructure, something that was extremely difficult to manage before, especially as a lean team operating real-time decision systems.”

Clicklease, Hopsworks customer story
How it's used

“We ran extensive benchmarks comparing both setups and saw no noticeable performance degradation on Kubernetes. The reduced manual effort and built-in autoscaling were major wins for us.”

Zalando, Hopsworks customer story
What people like

“Our journey with Hopsworks has been an amazing transformation that’s really enabled us to be innovative and reach a point that we wouldn’t have been able to reach otherwise.”

America First Credit Union, Hopsworks customer story
What people don't like

“The feature store was fast once it was in place, but the setup asked a lot of our data engineers, and we wanted more examples for our exact stack.”

Head of data science, consumer lender
Representative buyer view

Where customer reviews couldn't be found, our analysts attempted to give a voice to buyer sentiment, drawing on real information from review sites, forums, and product documentation.

Company Facts

Headquarters
Stockholm, Sweden
Founded
2017
Annual recurring revenue
$2.7 million (late 2025)
Pricing model
Free tier for one project; SaaS is pay-as-you-go; Enterprise is quoted
Website
hopsworks.ai

Key Leadership

Jim Dowling

Co-founder and CEO

He left an associate professorship at KTH Royal Institute of Technology to build Hopsworks.

Methodology

We scored Hopsworks on eight criteria, from experiment tracking to contract fairness, and the Overall Score is their simple average on a 0 to 10 scale. The scores draw on reviews from public review sites and forums, with paid reviews and pay-for-play sites excluded, plus a hands-on look at the product, its pricing, and its documentation and training. Scores can change as we add reviews and product checks.

← Back to MLOps & experiment tracking

Sources

Company facts

Review excerpts