Sustainable, affordable cloud compute
Our decentralised network delivers high performance compute that supplements your existing cloud platforms to save you money and carbon.
TRUSTED BY
Compute designed to save you money and carbon
heata is a distributed compute network that puts its waste heat to work, cutting energy use and cost.
Deploys within or alongside your cloud
Use the standard tools you already run, such as Kubernetes, Docker containers, Jupyter notebooks or your provider's hybrid cloud services (e.g. ECS Anywhere, Azure Arc). Your data continues living in your primary cloud platform.
Offload work that doesn't need a data centre
Move stateless, parallel workloads such as containers, AI inference, batch processing, CI/CD pipelines and ML training onto heata, and keep data centre capacity for what needs it. See our services pages for more.
SLA-backed capacity
Everything you access directly, such as control planes and orchestration, lives in the cloud. The distributed workers are interchangeable and managed by heata, with SLAs and capacity guarantees.
Secure by design
Each client's workloads run in isolated, encrypted virtual machines on private VPN networks, backed by Cyber Essentials Plus certification and independent penetration testing. More on our security.
Groundbreaking sustainability
heata delivers groundbreaking ultra low carbon compute
Morgan Stanley expects data centre emissions to triple by 2030, from around 200 to 600 million tonnes of CO2e a year, as AI drives demand. That puts big cloud's sustainability goals out of reach. We can be smarter about how we build compute infrastructure.
Running our workloads on the heata network
has demonstrated that we can provide significant
benefits to both Centrica and households.Peter Simons
Head of Cloud Architecture Centrica
Up to 50% cheaper than the hyperscalers
Flexible pricing, from as little as 1p per core hour. Talk to us and we'll size a plan against your actual workload.
Why heata costs less
heata doesn't carry the costs of a data centre. There's no land, planning or construction, no new grid connection, no electrical or power systems, and no cooling systems or cooling energy. We pass those savings on to you.
SUPPORTED BY
heata compared to major cloud providers
heata is building compute infrastructure in a different way.
We avoid the huge embodied carbon of building a data centre by reusing existing infrastructure and we turn a data centre problem (heat) into a benefit (free hot water).
We need less power and we have fewer overheads, which means we are cheaper too.
 |  |  |  | Learn more | |
|---|---|---|---|---|---|
Lower costs | |  | | | Pay for what you use, avoid the hard to manage hidden costs of the public cloud. |
Ultra low carbon | | | | | heata uses energy twice avoiding cooling and offsetting a hot water heating energy cost. Hyperscalers also use market-based reporting to massively under-report their carbon impact. |
Re-uses heat | | | | | Heat is a useful resource, every day hyperscalers waste all the heat they create. |
Avoids cooling | | | | | Hyperscale data centres spend extra energy on cooling, and some also use water to get rid of heat. |
Saves water | | | | | Some data centres use large amounts of water for cooling. heata uses none. |
Uses less energy | | | | | Large data centres demand huge amounts of power, putting significant strain on the grid. |
heata compute services
Services for stateless, parallel workloads, from containers and VMs to CI pipelines and machine learning. Each deploys within or alongside your existing cloud.
Kubernetes
A standard, dedicated Kubernetes API. No proprietary tooling, no custom SDKs, no lock-in.
VM Clusters
Our lowest-priced compute. Disposable Ubuntu VMs with full root, in clusters that heal themselves.
Action Runners
Hosted GitHub Actions runners. Change one line of your workflow and save around 25%.
Ray via Anyscale
Add heata as a compute provider inside your Anyscale account and keep everything else.
Hybrid and custom setups
Connecting through ECS Anywhere, Azure Arc or another hybrid tool, or not sure which service fits? We'll help you choose.
CERTIFIED
Frequently asked questions
Around 750kg of CO2e a year per unit, compared with running the same compute in a conventional data centre and heating the same water separately. The estimator above applies the same figures to whatever amount of compute you need.
Very secure, we’re working with SUSE, leveraging their secure edge platform, providing enterprise grade security in physically exposed public infrastructure, eg mobile base stations.
Storage is inaccessible if removed from appliance; LUKS / AES-256 Fully Encrypted Drive, key secured with industry standard TPM 2.0 (ISO/IEC 11889) on motherboard.
Compute customer’s jobs are unaware of, and cannot be accessed by other customer’s jobs through VM isolation. Client's can request single tenancy if preferred.
Compute job’s access to the network will be removed once a job is finished or in event of suspicious behavior; remote access via unique revocable AES-256 private keys.
Compute loads are on a Private Subnet, isolated from the home network; all network traffic goes out through VPN out to the wider internet. Compute jobs are unable to access the host’s wifi and we use per unit AES-256 encrypted VPNs to the control server.
Standard, open tooling, so there is nothing proprietary to learn and nothing to rewrite. Compute runs on Kubernetes (K3s), with a dedicated cluster per client and no shared control plane, so kubectl, Helm, ArgoCD, Rancher and Flux all work unchanged.
Alongside that there are VM clusters with full root access, GitHub Actions runners, Ray through Anyscale, and Kubeflow for notebooks and pipelines. The units themselves run SUSE's immutable Linux.
Better than 99.9% monthly uptime on the cloud access node you deploy through, and better than 99.5% committed capacity per job cluster. Both are backed by service credits.
Spot capacity is priced lower and carries no SLA, so it suits work that can be interrupted and retried.
Anything that packages as a container or a VM and is content to be spread across many small machines: high-volume repetitive jobs, machine learning training and inference, CI pipelines, research and data processing, computational finance and risk analysis, and climate modelling.
Work that needs one very large machine, or very low latency to a single place, is a poorer fit.
Read about sustainable compute
- Cloud compute
Jupyter notebooks article
- 3D rendering
Team Spotlight: 3D Lead, Jay Patel
.webp%3F2026-09-10T16%253A03%253A12.747Z&w=3840&q=100)
