Carbon-neutral AI compute

narrowly leads for carbon-neutral AI compute clusters.

Since Jul 8Updated Nov 14, 2026Based on 59 AI answersReviewed by

How we measure

Brands AI mentions

#BrandMentioned in
#3AWS54%

Renewable energy targets and tools for monitoring customer emissions.

Renewable-powered AI infrastructure that can use otherwise wasted flared gas.

#4IREN27%

Renewable-powered data centers for high-performance AI and HPC workloads.

Mentioned in, on ChatGPT and Google AI Mode

Questions and answers

  • ChatGPTNov 2, 2026

    1. 1Responsible Compute
    2. 2Canopy Wave
    3. 3ClusterTech
    4. 4Borealis Data Center
    5. 5Nscale
    6. 6Google Cloud
    7. 7AWS
    8. 8Microsoft Azure
    9. 9Oracle

    If you're moving AI training, fine-tuning, or inference workloads to green energy, there are several providers worth evaluating. The strongest candidates fall into two groups: specialist GPU/HPC providers using renewable-powered data centers and hyperscale cloud providers with carbon-free energy programs.

    One important distinction: renewable-powered compute is not automatically carbon-neutral compute. Some providers use renewable electricity directly, while others match consumption with renewable certificates or offset residual emissions.

    Providers worth evaluating

    • Origo og Borealis setja á fót fyrirtæki

    1. Responsible Compute — Iceland

    (100% renewable energy)

    • Hydropower and geothermal-powered HPC infrastructure.
    • Custom CPU/GPU clusters for demanding workloads.
    • Offers HPC through its partnership with Rescale.
    • Energy monitoring and emissions reporting.

    Best for: Dedicated enterprise HPC and AI infrastructure.

    Sources:

    Provider website


    • Iceland Data Centres - Sustainable Data Centre | atNorth

    2. Canopy Wave — Iceland and global locations

    (100% renewable at specified Iceland facility)

    • GPU clusters for AI and HPC.
    • Iceland facilities use hydroelectric and geothermal power.
    • Reports PUE as low as 1.2 at its Akureyri facility.
    • GPU deployment through dashboard or API.

    Best for: GPU-intensive AI training and inference.

    Source:

    Provider website


    • Crusoe expands Iceland data center capacity with atNorth, secures $175M credit facility from Victory Park Capital

    3. ClusterTech — Iceland

    (100% renewable energy claim)

    • GPU rentals and bare-metal compute.
    • Advertises NVIDIA H100, A100 and L40S GPUs.
    • Hourly billing and cluster provisioning.
    • Iceland-based renewable-powered data centers.

    Best for: Flexible GPU capacity without long-term commitments.

    Source:

    Provider website


    • Skýjaþjónusta IBM til liðs við Borealis Data Center | Feykir.is

    4. Borealis Data Center — Iceland and Finland

    (100% renewable energy claim)

    • Large-scale AI/HPC colocation.
    • Custom GPU infrastructure through Responsible Compute.
    • Reports PUE as low as 1.03.
    • Suitable for organizations deploying their own clusters.

    Best for: Enterprise-scale dedicated infrastructure.

    Source:

    Provider website


    • Nvidia-Backed AI Startup Nscale Raises $1.1 Billion for Data-Center Rollout - WSJ

    5. Nscale — Europe and North America

    • Purpose-built AI infrastructure and GPU cloud.
    • Large-scale GPU deployments.
    • Partnerships with renewable-powered Nordic data centers, including Verne in Iceland.
    • Strong candidate for large AI training workloads.

    Best for: Industrial-scale AI clusters.

    Source:

    Provider website

    Major cloud providers with green-energy programs

    These providers offer more extensive cloud ecosystems, but their energy claims should not be confused with a guarantee that every individual GPU cluster is carbon-neutral.

    ProviderSustainability approach
    Google CloudPublishes hourly carbon-free energy percentages by region, with a 2030 24/7 CFE goal.
    AWSReports matching 100% of electricity consumed with renewable sources in 2025.
    Microsoft AzureOffers sustainability guidance and carbon reporting for cloud workloads.
    Oracle Cloud InfrastructureRenewable-powered cloud infrastructure; targeting 100% carbon-free electricity for custom AI data centers by 2035.

    Google Cloud is particularly interesting because you can compare regional carbon-free energy percentages rather than relying solely on annual renewable-energy matching.

    For example, its published 2025 figures include:

    • Stockholm: 100%
    • Finland: 98%
    • Montréal: 97%
    • Oregon: 83%
    • Northern Virginia: 57%

    cloud.google.com

    My recommendation

    For an AI infrastructure migration, I would prioritize these three:

    PriorityProviderReason
    1Responsible ComputeEstablished renewable HPC infrastructure
    2Canopy WaveDirect GPU cloud offering with renewable Iceland facilities
    3NscalePotential for large-scale AI infrastructure

    If you're considering moving several hundred GPUs or more, I'd also investigate Borealis for dedicated infrastructure.

    What to verify before signing a contract

    I recommend asking every provider for these six items:

    1. Energy sourcing: Is the actual facility powered by hydro, geothermal, wind or solar?
    2. Carbon accounting: Are emissions measured using location-based or market-based Scope 2 accounting?
    3. 24/7 matching: Is renewable energy matched hourly or only annually?
    4. GPU efficiency: What are the facility's PUE and GPU utilization figures?
    5. Carbon reporting: Can they provide emissions per GPU-hour or per workload?
    6. Offsets: Are residual emissions offset, and are those offsets independently verified?

    Importantly, 100% renewable electricity does not necessarily mean zero lifecycle emissions or complete carbon neutrality.

    A question to narrow down the best providers

    What scale of AI infrastructure are you planning to migrate?

    1. Approximate GPU requirement


    2. Primary workload


    3. Preferred geographic region


    4. Preferred infrastructure

    [Find suitable providers]

    I can then narrow this down to providers with credible renewable-energy sourcing, enterprise GPU capacity, and verifiable carbon reporting rather than just general sustainability promises.

    Google AI ModeNov 2, 2026

    1. 1Google Cloud Platform
    2. 2Microsoft Azure
    3. 3Amazon Web Services
    4. 4NVIDIA H100
    5. 5CoreWeave
    6. 6Lambda
    7. 7Azure OpenAI Service
    8. 8Crusoe Intelligence Foundry

    Moving heavy AI and LLM training workloads to green energy requires looking at hyperscalers and specialized GPU cloud providers that match consumption with 100% renewable energy, track carbon intensity, or utilize ultra-efficient hardware architectures.

    The primary options for running carbon-conscious or renewable-backed AI compute clusters include the following major providers:

    • Google Cloud Platform (GCP): A pioneer in 24/7 carbon-free energy (CFE) matching. Google operates some of the industry's most efficient data centers (boasting an average PUE as low as 1.09) and allows you to deploy AI workloads (via custom TPUs or NVIDIA GPUs) in regions with the lowest Compute Carbon Intensity (CCI). They match electricity consumption with local renewable energy purchases on an hourly or annual basis.
    • Microsoft Azure: Has focused heavily on sustainable AI infrastructure to support massive multi-node training and Azure OpenAI services. Microsoft aims to be carbon negative, operates carbon-neutral data center operations, and lets enterprise customers track emissions tied directly to Azure subscriptions through their Sustainability Calculator.
    • Amazon Web Services (AWS): Reached a corporate goal of matching 100% of its global operations' electricity consumption with renewable energy well ahead of schedule. AWS provides high-performance EC2 UltraClusters for AI (featuring NVIDIA H100/B200 chips) backed by Amazon's massive portfolio of wind and solar PPAs (Power Purchase Agreements).
    • Specialized & Boutique GPU Clouds (CoreWeave, Lambda Labs, Crusoe): While hyperscalers match power via offsets and grid credits, specialized providers like Crusoe Energy take a different architectural approach by powering modular data centers directly off stranded or flared natural gas, or integrating heavily with dedicated renewable microgrids. Traditional high-density GPU orchestrators like CoreWeave are increasingly pressured to site data centers near clean energy corridors, though you will need to request specific region-by-region power audits for dedicated multi-node training clusters.

    To help narrow down the best fit, could you tell me:

    • What specific GPUs or scale (e.g., hundreds of H100/B200 or Google TPUs) do your clusters require?
    • Do you need hourly 24/7 carbon-free matching , or is an annual renewable energy credit (REC) / PPA offset sufficient for your compliance goals?
  • ChatGPTNov 8, 2026

    1. 1IREN
    2. 2Bitdeer
    3. 3Scala Data Centers
    4. 4Google Cloud
    5. 5Oracle
    6. 6Start Campus (SINES DC)

    If you're looking for data centers that host or operate AI GPU clusters powered by carbon-neutral or renewable energy, I'd focus on a mix of specialist AI infrastructure providers, renewable-powered data center operators, and hyperscalers investing in carbon-free electricity.

    There are some particularly interesting candidates in the US, Canada, and Europe.

    One important distinction: 100% renewable energy does not necessarily mean 100% carbon-free electricity every hour. Some operators use renewable energy certificates (RECs) or annual power-purchase agreements (PPAs), while others have more direct connections to renewable generation.

    Here are the companies I'd put on your research shortlist.

    My shortlist: 6 companies to investigate

    • IREN plans 75MW liquid-cooled AI data center in Texas - DCD

    1. IREN

    (Strong US candidate)

    United States, Canada, Europe, Australia

    • Purpose-built NVIDIA GPU infrastructure for AI training and inference.
    • More than 5 GW of secured power.
    • Claims 100% renewable energy at its facilities, including through RECs.
    • Offers GPU cloud, bare-metal compute, and data center infrastructure.

    Why investigate: One of the strongest direct matches if you want to actually rent AI compute rather than simply lease a data center building.

    Sources:

    IREN website


    • Start Campus selected by Nscale for the European Union’s First and Largest Deployment of NVIDIA Blackwell Ultra AI Infrastructure in support of Microsoft - subcablenews.com

    2. Start Campus

    (Strong sustainability candidate)

    Sines, Portugal

    • 1.2 GW planned campus capacity.
    • 100% renewable electricity according to the operator.
    • Seawater cooling, reducing freshwater consumption.
    • Designed for high-density AI, cloud and HPC workloads.
    • First facility, SIN01, operational since 2024.

    Why investigate: Particularly compelling if you're interested in European AI infrastructure with a sustainability-focused design.

    Sources:

    Start Campus website


    • Solgt til Singapore-basert selskap

    3. Bitdeer

    (Interesting renewable-powered project)

    Tydal, Norway

    • Developing a 121 MW AI/HPC data center.
    • Powered by 100% renewable energy, according to its announcement.
    • Long-term AI infrastructure lease with NVIDIA GPU deployment.
    • Norway offers an attractive renewable electricity environment.

    Why investigate: An interesting option for large-scale GPU infrastructure in a renewable-rich Nordic market. Verify the current commissioning and customer-access arrangements.

    Source:

    Bitdeer website


    • Scala e Vertiv Assinam Contrato Estratégico para o Fornecimento de Soluções de Gerenciamento Térmico na América Latina

    4. Scala Data Centers

    Brazil and Latin America

    • 100% renewable and certified electricity, according to its sustainability reporting.
    • Long-term renewable power purchase agreements.
    • Significant hyperscale infrastructure portfolio.
    • Net-zero ambitions aligned with SBTi methodology.

    Why investigate: A promising option if you want to explore large-scale AI infrastructure in Latin America.

    Note that renewable certificates and PPAs are not necessarily equivalent to hourly carbon-free electricity.

    Source:

    Scala website


    • Luchtfoto Eemshaven Google Data Center | Siebe Swart

    5. Google Cloud

    Global

    • Operates AI infrastructure using TPUs and GPUs.
    • Has a 24/7 carbon-free energy target for every grid by 2030.
    • Achieved approximately 65% carbon-free energy matching across data centers and offices in 2025.
    • Invests in new renewable and other carbon-free electricity generation.

    Why investigate: Particularly relevant if you want managed AI compute with substantial clean-energy procurement.

    Important: its 65% figure is a global average, not a claim that every AI data center is 65% carbon-free.

    Sources:

    Google Cloud


    • Data Centers Become Flashpoint In Midterm Races Across The Nation

    6. Oracle Cloud Infrastructure

    United States, including Texas

    • Significant AI infrastructure investments.
    • Investing in more than 1.7 GW of Texas wind generation.
    • Working toward 100% carbon-free electricity for its AI data centers.
    • Relevant for enterprise AI and large-scale GPU workloads.

    Why investigate: An interesting candidate if you're interested in enterprise-grade AI compute and large power procurement agreements.

    Important: the wind investments support its goal; they do not establish that every facility is already carbon-free around the clock.

    Source:

    Oracle Cloud

    How I would evaluate their carbon-neutral claims

    I would not take a company's "100% renewable" marketing claim at face value.

    There are four levels worth distinguishing:

    Energy modelWhat it actually means
    24/7 carbon-freeClean electricity matched to consumption every hour
    Direct renewable supplyOn-site or directly contracted renewable generation
    Annual renewable matchingRenewable generation or certificates offset annual consumption
    Carbon offsetsEmissions compensated through external projects

    The first two are generally the strongest starting points for your research.

    Google's 24/7 framework is a useful benchmark for understanding hourly matching.

    I'd also examine:

    • PUE: How efficiently the facility uses electricity.
    • WUE: How much water cooling consumes.
    • Grid emissions: The actual carbon intensity of the electricity grid.
    • Additionality: Whether the operator's energy purchases help create new clean generation.
    • GPU density: Whether the facility can support modern NVIDIA H100, H200, B200 or GB200 clusters.

    My recommended starting points

    Depending on what you're trying to accomplish:

    Your objectiveStart with
    Rent renewable-powered GPU computeIREN
    Lease AI-ready data center capacityStart Campus
    Large Nordic AI infrastructureBitdeer
    Latin American infrastructureScala
    Managed AI cloud servicesGoogle Cloud
    Enterprise AI infrastructureOracle

    One additional consideration: the Data Center Air Pollution Tracker publishes a useful comparison of how US hyperscalers source their electricity and their exposure to fossil-fuel generation.


    One question that would help me narrow this down

    Are you looking at these companies as a potential customer, investor, or infrastructure partner?

    That distinction matters considerably:

    • Customer: Find operators where you can actually rent H100/H200/B200 GPU clusters and obtain verifiable renewable-energy reporting.
    • Investor: Identify companies with renewable power agreements, secured megawatts, AI infrastructure growth and defensible economics.
    • Infrastructure partner: Find operators seeking land, renewable power, colocation, GPU infrastructure or joint-venture partners.

    I can build a more detailed comparison, including actual AI compute capacity, geographic locations, energy sourcing, GPU availability, and the strength of each company's carbon-neutrality claims.

    Google AI ModeNov 8, 2026

    1. 1Google Cloud
    2. 2Microsoft Azure
    3. 3AWS
    4. 4Digital Realty
    5. 5Equinix
    6. 6Iron Mountain Data Centers
    7. 7EdgeConneX
    8. 8Infomaniak
    9. 9OVHcloud

    Finding true carbon-neutral or 24/7 carbon-free energy (CFE) for heavy AI compute clusters requires looking closely at how providers match power —whether they rely on real-time local clean grids, major Power Purchase Agreements (PPAs), or carbon offsets.

    While many major tech firms are heavily expanding into natural gas to feed massive training demands, several hyperscalers and specialized operators lead in green and renewable infrastructure:

    1. Hyperscalers & Cloud AI Platforms

    • Google Cloud: Widely regarded as a leader in operational efficiency and clean energy matching. Google has matched 100% of its annual electricity use with renewable energy globally and is working toward 24/7 carbon-free energy on every grid where it operates by 2030, utilizing high-efficiency data centers (PUE ~1.10) and massive wind/solar PPAs.
    • Microsoft Azure: Has been carbon-neutral since 2012 and operates on a public pledge to become carbon-negative by 2030 . Like others, it heavily leverages large-scale renewable procurement and clean-energy investments (though specific heavy-compute campuses still face regional grid mix challenges).
    • Amazon Web Services (AWS): Reached high levels of renewable energy matching globally through hundreds of wind and solar projects, and aligns its scaling compute clusters with the Climate Pledge (targeting net-zero carbon by 2040).

    2. Colocation & Dedicated AI Infrastructure Providers

    • Digital Realty: Offers scalable, high-density, AI-ready data halls and matches 100% of the electricity used across its entire European portfolio with renewable energy, actively participating in the EU Climate Neutral Data Centre Pact. They also build tailored AI deployments (such as work with GPU cloud providers like CoreWeave).
    • Equinix: Achieved roughly 96% renewable energy coverage globally across its International Business Exchange (IBX) data centers, invests heavily in facility energy efficiency, and targets net-zero operations by 2040 under SBTi-approved guidelines.
    • Iron Mountain Data Centers: Powered by 100% renewable energy , features low PUE metrics, and is a signatory to the Climate Neutral Data Centre Pact aiming for strict carbon neutrality goals.
    • EdgeConneX: Reached 100% carbon neutrality for Scope 1 and Scope 2 emissions globally and frequently partners on dedicated high-density AI and edge compute infrastructure.

    3. Specialized / European Green Hosts

    • Infomaniak: A standout European/Swiss option running on 100% local hydro and solar with verified ultra-low PUE (around 1.08), ideal if you need green sovereignty in Europe.
    • OVHcloud: Offers strong European enterprise cloud and AI infrastructure backed by high renewable energy integration and custom industrial water-cooling techniques.

    If you can share what kind of workload you are running (e.g., massive LLM training vs. lower-density inference) or your preferred geographic region (US, Europe, Asia), I can narrow down which of these providers fits your exact cluster and compliance needs.

Also asked as

  • Which companies provide AI compute clusters that run on carbon-neutral energy? We want to move our AI operations to a greener infrastructure.