Chip-maker and accelerator clouds
Clouds that give access to a company’s own accelerator chips, not only GPUs.
Definition, not a sourced statistic
What it is
Some chip makers and cloud companies design their own accelerators, such as TPUs or other custom chips, and offer them as a cloud service. You reach the chip through that company’s cloud or API. Software support differs from GPUs, so code may need changes to run well.
Who it suits
- Teams whose frameworks and models already run well on that chip.
- Inference workloads that suit a specialised design.
- Buyers who want an alternative supply route to GPUs.
Common contract shapes
- Cloud instances or pods. Rent the custom accelerator by time inside the company’s cloud.
- Inference API. Call models running on the chip, billed per token or per call.
- Dedicated or on-premises systems. Some makers also sell or place systems for one customer.
Providers use their own names for these shapes, so compare what is promised, not what it is called. The contract shapes guide explains each one.
What to ask
- Egress and data transfer. What does it cost to move data and trained weights out of that cloud?
- Service levels. Is the custom accelerator covered by the same service level agreement as ordinary instances?
- Data residency and access. In which regions is the chip available, which may be fewer than for GPUs?
- Security certifications. Do the provider’s certifications cover the accelerator service?
- Export controls and sanctions. Are the chip and its systems covered by export controls for your location?
- Exit. Will your models and code run elsewhere without the custom chip, and at what effort?
- Which frameworks and precisions are supported, and how much porting work is expected?
The full list is on the printable buyer's checklist. For rules on chips and where they can go, see export controls and compute thresholds.
On this site
Entries listed here because their own summary or checked facts on this site fit this type. An entry can fit more than one type, and a provider missing here may still offer it. This is a list, not a ranking, and in name order. We link to providers for prices.
Providers
- Cerebras: A maker of wafer-scale AI systems that also sells cloud inference, dedicated capacity and on-premises systems.
- Google Cloud (Compute Engine): Google's virtual machine service. Besides GPUs, it is the place to reach Google's own Cloud TPU accelerators for machine learning.
- Groq (GroqCloud): An inference provider built on its own Language Processing Unit chips, offering GroqCloud for developers.
- SambaNova: An AI hardware and inference company that runs its own RDU chips, sold as racks and offered as a cloud inference service.
Programmes
- Google TPU Research Cloud Global
Gap Trainium and other chips are not listed here because no provider page on this site yet records them.
Related
- Glossary: Accelerator, Training and inference.
- Guide: How to get GPUs this week
- Chip and accelerator entries in the supply chain
- Compare offer types
Other offer types: Hyperscale cloud instances · GPU clouds and bare metal · Marketplaces and aggregators · Decentralised and community networks · Regional and sovereign clouds · Managed inference and serverless GPU · National and institutional HPC, and free or funded programmes · Colocation, wholesale, powered shell and build-to-suit · On-premises hardware, and leasing or financing it · Capacity brokers, reserved contracts, secondary and spot markets · Edge and telecom sites, and converted sites
Definitions and questions written in our own words, checked on 3 October 2026 and due for review by 3 January 2027. Sourced facts stay on the provider, programme and policy pages this page links to. This is general information to help you ask better questions. It is not legal, tax or procurement advice. For export controls, sanctions and contracts, check the primary rules and talk to a qualified adviser. Something wrong? Tell us.