Oracle's 블로그는 OCI에서 Positron Atlas를 활용해 대규모 AI 추론을 효율적으로 수행하는 방안을 소개하고 있다.

As AI adoption accelerates, enterprises are increasingly focused on the cost and efficiency of running models in production. Delivering high-throughput, low-latency inference while managing power consumption and infrastructure costs has become a critical challenge.

Oracle Cloud Infrastructure (OCI) is expanding its AI compute portfolio with Positron Atlas, a purpose-built inference platform designed to deliver efficient, scalable inference for transformer-based workloads.

By combining Positron's memory-efficient accelerator architecture with OCI's high-performance infrastructure and Oracle Acceleron networking, customers gain another option for optimizing AI inference performance and cost.

Positron Atlas Inference Servers are purpose-built for large language models (LLMs), delivering real-time, low-latency performance for popular open-weight models. Positron Atlas is networked through Oracle Acceleron, OCI's suite of network software and architecture designed to remove networking bottlenecks for scaled inference applications, increase performance, enhance security, and simplify operations.

**Purpose-Built for Transformer Inference**

Positron Atlas is purpose-built for transformer architectures, which power today's popular LLMs. It uses a custom accelerator architecture optimized for LLM inference. Memory bandwidth is one of the primary bottlenecks in transformer workloads. With benchmarked memory-bandwidth efficiency of 93%, Positron accelerators offer a 3–5× TCO advantage for targeted workloads, due to power-efficient accelerators and fast request completion times.

This makes Positron Atlas on OCI well suited for:

- Cost-efficient deployment of smaller workhorse models such as GPT-OSS, Qwen 3.5, and Gemma 4

- Real-time AI agents, assistants, and coding copilot tasks

- Hot-model hosting and dynamic model switching with zero loading overhead

- High-throughput, continuous-demand inference APIs

**Rapid Deployment**

Software migration and adoption are seamless with OpenAI-compatible APIs. Teams can run existing models and integrate with popular open-source LLMs, accelerating time to production while minimizing engineering effort.

**Positron Atlas Instances on OCI**

OCI infrastructure, Oracle Acceleron, and Positron Atlas systems are designed to deliver strong performance and efficiency for production AI workloads, combining purpose-built accelerators with high-performance compute and networking:

- 8× Positron AI accelerators for high-throughput inference

- Tightly coupled, high-performance memory for up to 500B model weights

- OCI Acceleron hardware-level security isolation

- 144-core Intel Xeon 6 CPU (up to 4.4 GHz turbo)

- 1.5 TB system memory to support large models and data pipelines

- 30 TB of local NVMe storage for fast data access

- 200 Gbps networking for high-throughput communication

**Get Started with OCI Compute and Positron Atlas**

With the addition of Positron Atlas, OCI continues to broaden its AI infrastructure portfolio to give customers more flexibility, choice, and control. Whether leveraging general-purpose GPUs or purpose-built accelerators, customers can select the right platform for each workload, enabling more efficient, scalable, and cost-effective AI deployments. Positron Atlas Instances on OCI are available now. To get started, reach out to your OCI Sales Representative or contact Sales at Positron AI today.

**About Positron AI**

Positron AI builds hardware and software to make AI inference dramatically cheaper and more energy efficient. The company's shipping product, Atlas, is deployed at hyperscaler scale today, and its next-generation custom silicon, Asimov, tapes out in late 2026 with production in the second half of 2027. Asimov powers Titan, a multi-terabyte-memory inference system for long-context and next-generation AI workloads. Positron's energy-efficient architecture allows its systems to be deployed in air-cooled or liquid-cooled data centers at varying rack densities, without dependence on constrained HBM or CoWoS supply chains. Learn more at positron.ai.