WAIC Live | GenstorAIge AI90 Joins Hands with Weier Innovation, Consumer-Grade GPU Delivers Enterprise-Class Performance

2026 WAIC – Weier Innovation Booth
On July 17, the 2026 World Artificial Intelligence Conference (WAIC) was held in Shanghai.
GenstorAIge's AI90 — FaaS (Flash as a Service) for AI architecture and AI SSD series products made their debut. At the same time, a Weier Innovation GPU workstation equipped with the GenstorAIge AI90 solution was running a live large model inference demo, immediately drawing a crowd of professional visitors and becoming one of the most popular highlights of the exhibition area.
GenstorAIge AI90 Solution: Running Large Models with Small Memory
The core challenge in current AI inference is the severe imbalance between compute and storage: the 'memory wall' leaves expensive compute chips in a perpetually 'starved' waiting state, with effective compute utilization for mainstream GPUs at only 30%–60%.
Meanwhile, as large model inference capabilities and AI Agents accelerate their evolution, the rapid expansion of KV Cache further exacerbates this contradiction.
The GenstorAIge AI90 solution, running stably in a real-world environment at the booth, is a systematic engineering approach to this bottleneck. The core innovation of AI90 is its FaaS for AI Cache architecture. Through a self-developed management middleware, it hierarchically offloads the rapidly expanding KV Cache from HBM to a high-performance AI SSD (PT200Z), building a three-tier storage pool of 'HBM + DRAM + SSD'.
Simultaneously, intelligent P2P interconnect technology optimizes the data path between GPUs, enabling efficient direct GPU-to-GPU data transfer without involving the CPU or host memory.

GenstorAIge CEO Sun Dan and Weier Innovation Partner Observe AI90 Data
At the event, the GenstorAIge AI90 solution was deeply integrated with a Weier Innovation workstation. The measured data from the AI90 acceleration solution was outstanding:
• Throughput surged from approximately 120 tk/s in a traditional architecture to about 610 tk/s, a 5.1x performance increase.
• Time to First Token for a 64K context dropped dramatically from 27.99 seconds to 0.564 seconds, a nearly 50x performance improvement.
• GPU memory utilization increased from 30%–40% to 85%–95%, a 2-3x improvement.
• 'The Longer, The Better' — The longer the input sequence, the more pronounced the acceleration effect of AI90. It is perfectly suited for complex scenarios like long document analysis, code comprehension, and multi-turn conversation history.
GenstorAIge × Weier Innovation: AI90 Completes System-Level Validation for AI Computing Center Scenarios
At WAIC 2026, GenstorAIge integrated its AI90 'compute-storage convergence' solution and AI SSD (PT200Z) storage foundation into Weier Innovation's GPU server environment. The joint teams completed full-link tuning of KV Cache offloading, intelligent scheduling, and data read/write paths.

The synergy efficiency between compute and storage is becoming a decisive variable for token throughput. The collaboration between GenstorAIge and Weier Innovation validates the maturity of the AI90 solution and its depth of integration with edge-side computing infrastructure.

Conclusion
While the industry is still paying a premium for HBM to cover high compute costs, the GenstorAIge AI90 solution leverages high-efficiency AI SSD storage, offering customers a lower-cost, achievable engineering path through a systematic 'compute-storage convergence' approach.
GenstorAIge's goal is to generate more tokens for every dollar of compute investment, making large models affordable for every SME and developer.

GenstorAIge CEO Sun Dan and Weier Innovation Partner Group Photo
|About GenstorAIge:
GenstorAIge is dedicated to building the data foundation for an intelligent world. It is the only full-stack 'store-connect-manage' vendor in China that simultaneously provides enterprise SSDs, PCIe Retimer/Redriver connectivity chips, and integrated AI training and inference solutions. Founded in 2023, its core team has over 20 years of deep expertise in the storage and server industries and has rapidly entered the supply chains of several leading internet and OEM companies. At WAIC 2026, GenstorAIge launched AI90 — its FaaS for AI architecture, an innovative 'compute-storage convergence' solution that unleashes the potential of large models on consumer-grade GPUs, driving AI computing toward universal access.
|About Weier Innovation (TIGERWAY):
Founded in 2018, Weier Innovation (TIGERWAY) is a hardware provider integrating ODM/OEM customization and industry IT solutions, and a National High-Tech Enterprise. The company focuses on server R&D, design, and manufacturing, with products covering AI servers, GPU servers, Xinchuang servers, hyper-converged appliances, and AI clusters. Its self-developed GPLUSTER series of high-density computing servers supports mixed installation of various GPU and NPU architectures, accommodating up to 8 compute cards in a 4U chassis. These are widely used in AI computing centers, information security, government, and large state-owned enterprises.
