Ampere Computing

Ampere Computing is now part of the SoftBank Group.

Ampere Computing designs and manufactures cloud native server processors and complete compute platforms focused on high performance, energy efficiency, and scalability from cloud to edge.

Ampere processors are built around single‑threaded, highly parallel CPU cores that deliver predictable, low‑latency performance for modern cloud workloads, including AI inference, data analytics, media, storage, and web services.

Customers and Use Cases

Ampere primarily serves:

• Hyperscale and public cloud providers
• Enterprise and SaaS data centers
• Telecom and networking providers
• Edge and embedded solution builders (for example in communications, transportation, and computer vision)

Common use cases include:

• AI inference and general AI compute
• Cloud native applications and microservices
• Data analytics and big data platforms
• Media streaming and content delivery
• Storage services and databases
• Web and API services

Technology Focus

• Cloud native, single‑threaded CPU architecture
• High core counts for dense, scale‑out workloads
• Consistent frequency and low jitter for predictable performance under load
• High energy efficiency to reduce power consumption and carbon footprint in data centers

Location

Ampere Computing LLC

4655 Great America Parkway, Suite 601 Santa Clara, CA 95054 United States

For general information and additional details about the company, visitors can use the "Contact Us" options on the website at https://amperecomputing.com.

• Sales contact page: https://amperecomputing.com/company/contact-sales
• The form collects basic personal and company information, country, project timeline, products of interest, and a free‑text message.
• There is an option to sign up for the Ampere newsletter when submitting the form.

Customers can use this page to request information about:

• Where to buy systems powered by Ampere processors
• Evaluating AmpereOne or Ampere Altra platforms
• Platform, solution, or partnership discussions

Other Inquiries

For non‑sales inquiries (such as general company questions, press, or other topics), visitors are directed from the sales contact page to other contact options available on https://amperecomputing.com.

Response Times and Support

The public website does not guarantee specific response times, support SLAs, or service levels. Any commitments on support, SLAs, or response times must be confirmed directly with Ampere or its partners.

• Deliver high inference throughput with many AI agents or models per server
• Maintain consistent performance under maximum load with low jitter
• Provide predictable multi‑tenant behavior for AI services
• Reduce power consumption and total cost of ownership for AI infrastructure

Key AI Products

AmpereOne M

• Optimized for high‑volume AI inference workloads
• Increased memory bandwidth vs. baseline AmpereOne
• Designed to host dense deployments of LLMs and SLMs with low cost per inference

AmpereOne

• General‑purpose cloud compute platform that also supports legacy ML and smaller language models
• Suitable for databases, web and application tiers, and other cloud services alongside AI workloads

Ampere Altra

• Used for AI in communications, telecom, and edge applications
• Focus on power‑efficient inference close to where data is generated

AmpereOne Aurora (future AI compute)

• Combines Ampere‑designed cores, interconnect mesh, and chiplet architecture with AI acceleration
• Positioned as a next‑generation, high‑efficiency AI compute solution

AI Software Stack and Resources

Ampere provides an AI software stack and ecosystem support, including:

• Ampere Optimized AI Frameworks (often abbreviated AIO), aligned with popular AI frameworks
• Support for frameworks such as TensorFlow, PyTorch, and ONNX on Ampere platforms
• Ampere Model Library and example workloads
• Technical documentation, whitepapers, solution briefs, and tutorials

Resources are organized in the AI and Developer sections of https://amperecomputing.com, where developers can:

• Download AI frameworks and tools
• Access AI workload briefs and performance resources
• Explore example implementations for GenAI, LLMs, and recommender systems

Target AI Use Cases

• Large‑scale AI inference (LLMs and SLMs)
• Recommender and personalization engines
• AI‑enhanced digital services and SaaS platforms
• AI at the edge in networking, communications, and embedded systems

Pricing, detailed sizing guidance, and deployment architectures are provided through Ampere and its partners and are not listed directly in this agent. Prospective customers should contact Ampere or its partners for current configuration and pricing information.

AmpereOne Platforms

AmpereOne platforms are general‑purpose, cloud and AI compute processors optimized for dense, scale‑out workloads.

Key characteristics

• Designed for high efficiency and container/VM density
• Supports both traditional cloud workloads and legacy AI models
• Built for predictable, single‑threaded performance

AmpereOne (flagship)

Most versatile and efficient for cloud and AI workloads.

Indicative specifications (from current public product briefs):

• 96–192 CPU cores (single‑threaded)
• 2 MB private L2 cache per core
• 64 MB system level cache
• 8‑channel DDR5 memory, up to 4 TB
• 128 lanes PCIe Gen5
• 200–400 W TDP

AmpereOne M

Optimized for large‑scale AI inference and dense AI environments.

Indicative characteristics:

• Higher memory bandwidth vs. AmpereOne for AI workloads
• 12‑channel DDR5 memory, up to 1.5 TB
• 96 lanes PCIe Gen5
• 250–425 W TDP

AmpereOne and AmpereOne M platforms are positioned for multi‑tenant AI compute, LLM and SLM hosting, and other demanding cloud workloads with a focus on low cost per inference and high utilization.

Ampere Altra Platforms

Ampere Altra and Altra Max platforms are cloud native processors used widely in telecom, networking, autonomous vehicles, and edge AI applications.

Key characteristics

• Focus on low power and high density from cloud to edge
• High core counts for scale‑out services
• Designed for consistent, predictable processing even in power‑constrained racks

Indicative specifications (from current public product briefs):

• 32–128 CPU cores (single‑threaded)
• 1 MB private L2 cache per core
• 16–32 MB system level cache
• 8‑channel DDR4 memory, up to 4 TB
• 128 lanes PCIe Gen4
• 45–250 W TDP

These platforms are suited for:

• Communications and telecom workloads
• Edge and embedded AI services
• Media processing and content delivery
• General purpose cloud native compute where power efficiency is a priority.

Future AI Compute: AmpereOne Aurora

Ampere has announced AmpereOne Aurora, combining Ampere‑designed CPU cores, a proprietary mesh and chiplet interconnect, and integrated AI acceleration. It is positioned as next‑generation AI compute for high‑performance, energy‑efficient inference at scale.

For the most up‑to‑date specifications, customers should refer to the latest product briefs and documentation on https://amperecomputing.com.

Ampere‑based instances are available from multiple public cloud providers. Examples listed on the website include:

• Oracle Cloud Infrastructure (OCI)
• Google Cloud (Tau T2A VMs)
• Microsoft Azure (Dpsv5 and Epsv5 VMs featuring Ampere Altra)
• Alibaba Cloud
• Tencent Cloud
• JD Cloud
• UCloud
• Hetzner
• IONOS
• Scaleway
• Leaseweb
• And other regional and specialized providers noted in the partner ecosystem pages

These clouds generally offer Ampere‑based virtual machines or services targeted at scale‑out, cost‑efficient workloads.

Systems and Platform Partners

Ampere processors are available in server platforms from major system vendors. Examples include:

• Supermicro (servers using AmpereOne and Ampere Altra processors)
• GIGABYTE
• Hewlett Packard Enterprise (HPE ProLiant RL300 Gen11)
• Additional hardware partners such as 7StarLake, ASRock, Foxconn Industrial, Hyve, Wiwynn, and others listed on the ecosystem page

These partners provide a variety of rack‑mount and specialized systems for data center and edge deployments.

Integrators and Distributors

Ampere collaborates with:

• Systems integrators that configure and deploy Ampere‑based infrastructure for end customers
• Distributors that make Ampere‑powered platforms available through channel partners

Examples of distributors listed on the site include Arrow, Avnet, Edom Technology, Newegg, and TD SYNNEX.

How to Buy

• Customers typically purchase Ampere‑based compute capacity either:
• As cloud instances from public cloud providers, or
• As physical servers and systems from hardware partners, integrators, and distributors.
• Ampere's website provides "Partner Ecosystem" and "Where to Buy" entry points that list current options and links.

Specific prices, SKU‑level configurations, and regional availability are managed by cloud providers and hardware partners and are not listed in this agent. Prospective buyers should follow the partner links on https://amperecomputing.com or contact Ampere for sales assistance.

Artificial Intelligence

• High performance, power‑efficient AI inference
• Support for major AI frameworks such as TensorFlow, PyTorch, and ONNX
• Suitable for GenAI, LLMs, and a wide range of inference‑driven services

ARM Native Applications and Cloud Gaming

• Native support for 32‑ and 64‑bit ARM/Android applications
• Demonstrated ability to host very high numbers of 3D cloud gaming instances per server (figure referenced on the website as 120+ instances per server)

Data Analytics

• High core counts and single‑threaded performance for scale‑out big data workloads
• Consistent frequency and predictable performance for analytics pipelines

Edge and Embedded

• Fits more compute and AI into tight power and thermal envelopes
• Target applications include communications, transportation, and computer vision
• Supports SWaP‑constrained (size, weight, and power) deployments

Media Services

• Suitable for media processing and streaming workloads
• Cloud native design helps reduce jitter and latency
• Helps isolate workloads from noisy‑neighbor effects

Storage

• High performance storage services on Kubernetes and other cloud infrastructure
• Focus on data management and scalable storage performance

Web Services

• Optimized for common web and API workloads
• Supports modern containerization and cloud native orchestration
• Aims to lower hardware costs while maintaining performance

Developer and Evaluation Programs

Ampere Ready Software

Ampere maintains a catalog of workloads and applications that are validated on Ampere platforms.

• Users can search by application, name, or availability
• Helps match specific applications to Ampere‑based solutions

Trial Access to Ampere

• Free trial platforms and compute resources are offered for evaluation of Ampere cloud native processors
• Access is requested through forms on the website

Ampere Developer Program

• Developer newsletter and program with updates on tools, resources, and events
• Supports the full lifecycle from design and build through deployment on Ampere platforms

For the latest solution briefs, validated workloads, and trial access details, visitors should refer to the Solutions section at https://amperecomputing.com/solutions.

Ask anyway — the agent will say plainly that it is not published rather than guess.