At NVIDIA GTC 2026, HPE is focusing on how HPE ProLiant servers combined with NVIDIA Blackwell RTX PRO GPUs can help organizations run AI workloads from the edge to the data center. There are three main solution areas being highlighted: 1. **Edge AI with HPE ProLiant DL145 Gen11 + NVIDIA RTX PRO 4500 Blackwell Server Edition** - Compact, ruggedized edge server paired with a single-slot RTX PRO 4500 GPU with **32 GB of high-speed memory**. - Designed for environments where **space, power, and latency** are critical. - Supports **multi-instance GPU (MIG)** to securely isolate up to **two workloads per GPU**, enabling cost-effective sharing across: - Small model inference - Vector database acceleration - Data analytics - CUDA-based applications at the edge - Example use cases: real-time anomaly detection, predictive maintenance, on-device vision for safety, image triage in remote clinics, and live fraud detection at point-of-sale. 2. **Data center AI with HPE ProLiant DL380a Gen12 + NVIDIA RTX PRO 6000 Blackwell Server Edition** - Built for high-end AI workloads that need **maximum throughput and capacity**. - Configurations with **eight RTX PRO 6000 Blackwell Server Edition GPUs** have set records in: - **MLPerf Inference: Datacenter v5.1** (Llama 3.1-8B Server benchmark, submission ID 5.1-0051). - **STAC-AI LANG6** benchmarks for financial services workloads. - Achieved **46,060 tokens per second** among eight-PCIe GPU configurations, positioning the DL380a Gen12 as a strong option for large-scale inference and language-model workloads. 3. **Retail AI with HPE ProLiant DL380 Gen12 + NVIDIA RTX PRO 4500 Blackwell Server Edition** - HPE validated the **NVIDIA Retail Shopping Assistant AI Blueprint** on an HPE ProLiant DL380 Gen12 with **four RTX PRO 4500 Blackwell Server Edition GPUs**. - This reference architecture is aimed at: - More personalized, conversational shopping experiences - Better product discovery - Higher conversion rates and average order value - Reduced product returns through more relevant recommendations Across these platforms, HPE is emphasizing flexibility: customers can choose from a range of ProLiant models (DL145, DL365, DL385, DL380, DL345, ML350, DL380a) and GPU configurations to align with their specific AI workloads, budgets, and deployment footprints—from branch and edge sites to core data centers.

HPE and NVIDIA are positioning their joint solutions to meet the strict requirements of financial services institutions (FSIs), where latency, security, and reliability are central. **Key platform: HPE ProLiant DL380a Gen12 + NVIDIA RTX PRO 6000 Blackwell Server Edition** This system has been independently validated for financial services AI workloads through the **STAC-AI LANG6** benchmark. The results show it can handle production-grade FSI use cases such as market surveillance, fraud detection, and trading analytics. **Notable performance data from STAC-AI LANG6** (with RTX PRO 6000 Blackwell GPUs): - **Latency and throughput** - Up to **165 inferences per second**. - **Median latency under 200 ms**, which supports real-time decisioning for: - Market surveillance - Fraud detection - Trading and order routing - **Streaming performance for language models** - Around **40 words per second** in language-model streaming scenarios, suitable for: - Risk simulations - Scenario analysis - Automated review of filings and alerts - **Accuracy under load** - Maintains **more than 90% accuracy** even under heavy load, which is important for: - Fraud screening - Regulatory reporting - Compliance monitoring In addition, the DL380a Gen12 with eight RTX PRO 6000 GPUs achieved **46,060 tokens per second** in the **MLPerf Inference: Datacenter v5.1** Llama 3.1-8B Server benchmark (submission ID 5.1-0051). This demonstrates its ability to scale large language model inference for FSI use cases like: - Intelligent trade surveillance assistants - Automated document and filing analysis - Real-time customer service and advisory tools For workloads that must remain on-premises due to regulatory or data-sovereignty requirements, these ProLiant platforms allow FSIs to: - Keep sensitive data (transactions, customer records, trading strategies) inside their own data centers. - Use high-throughput, low-latency inference for time-sensitive decisions. - Balance performance and compliance by combining speed, accuracy, and secure infrastructure.

Retailers can use HPE ProLiant servers with NVIDIA Blackwell GPUs to move from basic keyword search toward more intelligent, conversational shopping experiences. **Reference design: NVIDIA Retail Shopping Assistant AI Blueprint on HPE ProLiant DL380 Gen12** HPE has validated the **NVIDIA Retail Shopping Assistant AI Blueprint** on an **HPE ProLiant DL380 Gen12** configured with **four NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs**. This combination is designed as a scalable, high-performance platform for retail AI. **What this enables for retailers** 1. **Conversational, context-aware product discovery** - Customers can ask complex, natural-language questions instead of short keyword queries. - Example: “Help me plan a soccer-themed birthday party for 10 kids, including decorations, snacks, and small gifts,” rather than searching item by item. - The assistant can curate full sets of products for projects or events (e.g., backyard makeovers, seasonal décor, DIY projects). 2. **More personalized recommendations** - The blueprint is designed to support personalization, using customer context and preferences to: - Suggest complementary products (cross-sell) - Offer suitable upgrades (upsell) - Tailor recommendations to style, budget, or use case 3. **Operational and business benefits** - **Higher conversion rates** by making it easier for customers to find what they need in a single interaction. - **Increased average order value** through more complete baskets and relevant add-ons. - **Reduced product returns** by recommending items that better match customer intent and use cases. 4. **Scalability and deployment flexibility** - The DL380 Gen12 with RTX PRO 4500 GPUs provides a data center platform that can scale across regions and channels (web, mobile, in-store). - Retailers can run these assistants on-premises or in their own data centers to keep tighter control over data and integration with existing systems. In short, by combining HPE ProLiant DL380 Gen12 servers with NVIDIA Blackwell GPUs and the Retail Shopping Assistant AI Blueprint, retailers can reimagine how customers discover and buy products, while tying AI initiatives directly to measurable outcomes like revenue, conversion, and customer satisfaction.