Pliops Unveils Breakthrough AI Performance Enhancements

XDP LightningAI Delivers More Than 2.5X End-to-End Performance Boost for NVIDIA Dynamo and vLLM Production Stack at Dell World Technologies 2025, Maximizing GPU Server Efficiency

XDP LightningAI

LAS VEGAS, May 20, 2025 (GLOBE NEWSWIRE) -- Booth #568,  Pliops, a leader in AI infrastructure optimization, is set to showcase its groundbreaking XDP LightningAI solution at Dell Technologies World 2025 in Las Vegas. In collaboration with Dell OEM Solutions, Pliops is redefining AI performance possibilities. Pliops XDP LightningAI solutions enable businesses to achieve up to 4X more transactions per server, tackling rack-level power constraints while also simplifying deployment and lowering costs.

At its booth, Pliops will demonstrate how its XDP LightningAI solution, running on NVIDIA Dynamo and the vLLM Production Stack, delivers more than 2.5X end-to-end (E2E) performance improvements. This breakthrough redefines efficiency and significantly lowers the Total Cost of Ownership (TCO) for enterprise AI deployments.

Unlike traditional approaches that require DRAM, complex network storage, or additional indexing layers, XDP LightningAI simplifies AI infrastructure with a streamlined, single-namespace architecture, eliminating metadata overhead and ensuring seamless integration with vLLM.

“The demand for scalable, high-performance AI infrastructure has never been greater,” said Ido Bukspan, CEO of Pliops. “Through our collaboration with Dell OEM Solutions, XDP LightningAI is redefining AI infrastructure efficiency. With XDP LightningAI, there is no DRAM. No network storage complexity. No indexing. No metadata. Just one unified data flow, for maximum efficiency with minimal deployment friction."

HBM-Class Performance, Without HBM Costs

XDP LightningAI delivers performance on par with HBM, without the cost or supply constraints. Leveraging Pliops’ storage architecture, XDP LightningAI effectively provides GPUs with an abundant supply of HBM-level memory, unlocking unprecedented efficiency for AI inference workloads. Enterprises gain HBM-class performance without the need for expensive memory upgrades, maximizing existing infrastructure investments.

Rack-Level Power Efficiency

As AI workloads grow in complexity, power constraints at the rack level have become a major bottleneck for enterprises scaling their GPU infrastructure. XDP LightningAI directly addresses this challenge by reducing power consumption per inference task, enabling organizations to:

• Achieve up to 4X more transactions per GPU server while staying within existing power budgets.
• Optimize power efficiency across AI inference workloads, reducing cooling and infrastructure costs.
• Extend the lifespan of AI deployments by maximizing performance-per-watt, ensuring sustainable scalability.
  

Cost Savings and Dollar/Token Efficiency

XDP LightningAI delivers significant cost savings compared to traditional AI infrastructure solutions by accelerating transactions per GPU server. XDP LightningAI directly improves dollar/token efficiency, allowing enterprises to process more tokens per dollar spent. This translates into lower inference costs, making high-scale AI workloads financially sustainable without compromising performance.

Pliops FusionX stack

At the core of XDP LightningAI is the Pliops FusionX stack, a breakthrough technology that reduces cost, power consumption and computational overhead by optimizing LLM inference workflows.

Traditional LLM inference systems repeatedly process the same context for each request, leading to wasted compute cycles, excessive power consumption and unnecessary GPU overhead. FusionX eliminates this inefficiency by enabling context reuse, ensuring that each context is processed only once.

By reducing computational overhead, FusionX allows enterprises to scale AI inference workloads effortlessly and efficiently, accelerating performance while lowering infrastructure costs.

Exclusive Demos and Industry Validation

Adding to the excitement surrounding the show, StorageReview will publish the results from their lab testing, further showcasing the real-world performance advantages of XDP LightningAI. Commented Brian Beeler, CEO of StorageReview, “As large language models continue to scale, so do the infrastructure challenges around inference performance and efficiency. Pliops XDP LightningAI addresses a critical pain point by enabling fast, scalable KV cache offload without compromise. Our benchmarking with Pliops and NVIDIA Dynamo demonstrates how this solution can dramatically improve GPU throughput, bringing a new level of efficiency to real-world AI deployments of any scale.”

Visit Pliops at Dell Technologies World 2025 to experience firsthand how XDP LightningAI is transforming AI infrastructure efficiency.

Connect with Pliops 
Read Blog 
About Pliops 
Visit Resource Center, XDP LightningAI Solution Brief 
Connect on LinkedIn 
Follow on X

About Pliops 
A winner of the FMS 2024 most innovative AI solution, Pliops is a technology innovator focused on making data centers run faster and more efficiently. The company’s Extreme Data Processor (XDP) radically simplifies the way data is processed and managed. Pliops overcomes I/O inefficiencies to massively accelerate performance and dramatically reduce overall infrastructure costs for data-hungry AI applications. Founded in 2017, Pliops has been recognized multiple times as one of the 10 hottest semiconductor startups. The company has raised over $200 million to date from leading investors including Koch Disruptive Technologies, State of Mind Ventures Momentum, Intel Capital, Viola Ventures, SoftBank Ventures Asia, Expon Capital, NVIDIA, AMD, Western Digital, SK hynix and Alicorn. For more information, visit www.pliops.com. 

Media Contact:
Stephanie Olsen
Lages & Associates
(949) 453-8080
[email protected]

A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/17216ac4-bad6-463f-81af-e43b4a41181d