SuperComputing 2026: From AI Infrastructure to Token Factory
Event News

SuperComputing 2026: From AI Infrastructure to Token Factory

2026.10.01
INFINITIX SC26 Chicago, HPC Unites BOOTH: 6111

From AI InfrastructureTo Token Factory

Software-Defined GPU & AI Infrastructure,
Built to Power Every Token.

Nov 17–19, 2026|McCormick Place|Chicago

About INFINITIX

Turning GPU Infrastructure into Scalable AI Services

INFINITIX provides AI infrastructure software that enables enterprises to manage and optimize heterogeneous GPU and NPU resources.

As a global NVIDIA Solution Advisor, we help organizations simplify compute management, improve resource utilization, and build scalable AI infrastructure.

  • Awards: 2025 AMD GPU Ecosystem, COMPUTEX Best Choice and TAIA Awards
2003Founded in Taiwan
250+Enterprises and organizations served
8Markets: Taiwan, Japan, Korea, Hong Kong, China, Singapore, Malaysia, Thailand
On display at booth #6111

Two platforms, from GPU pool to AI service

AI-Stack: Software-Defined Infrastructure Management for GPU
GPU Resource Management

 ▶ AI-Stack

Software-Defined GPU & AI Infrastructure for HPC & AI

Run Slurm and AI workloads on one shared GPU pool, from a single card to multi-node clusters.

  • Intelligent Scheduling
  • Slurm on Kubernetes
  • Cross-node training & inference
  • Multi-tenant quotas & monitoring

Maximize GPU utilization. Accelerate AI innovation.

ixCSP: Turn AI Infrastructure into a Token Factory
AI Cloud Operations Platform

 ▶ ixCSP

Turn AI Infrastructure into Token Factory

Launch and run scalable AI services on your own infrastructure.

  • Multi-tenant cloud operations
  • Metering & billing
  • AI model marketplace
  • GaaS · MaaS · TaaS

From compute capacity to scalable AI services.

AI-Stack for HPC

Keep Slurm. Pool every GPU.

AI-Stack runs Slurm on Kubernetes, so HPC users keep sbatch and srun while training, inference and development share one GPU pool.

Traditional Slurm

Great at HPC, hard to operate

  • Fine-grained parallel scheduling
  • Native MPI / OpenMPI support
  • Manual bare-metal setup
  • No self-healing or scaling
  • Hard to share with AI work

Pure Kubernetes

Great at operations, weak at batch

  • Portable containers
  • Self-healing, elastic scaling
  • Pods start one by one, risking deadlock
  • Pods spread out, adding latency
  • No HPC job management
AI-Stack

Slurm on Kubernetes

HPC scheduling on a cloud-native foundation

  • Keep existing Slurm workflows
  • Slurm clusters share one GPU pool with other AI workloads
  • Kubernetes automates node provisioning, self-healing, and scaling
  • Easy move from bare-metal to cloud-native

Two-tier scheduling, one GPU pool

Slurm schedules jobs. AI-Stack allocates every GPU.

Workloads
Slurm cluster Created on AI-Stack · sbatch / srun
Queues & priority Backfill Gang allocation
Kubernetes-native AI work Scheduled directly by AI-Stack
Distributed training (DeepSpeed) Training jobs Dev & inference
AI-Stack scheduling
Gang all pods start together (default) Binpack fill nodes first, cut fragmentation (optional)
Kubernetes foundation
Containerized Self-healing Elastic scaling Unified network & storage
AI Infrastructure
NVIDIA & AMD GPUs NPUs InfiniBand / RoCE Shared storage

Governance HPC centers need

Multi-tenant projects & quotas Usage & cost reports Idle GPU auto-reclaim LDAP / AD & 2FA GPU, power & idle-time dashboards Full audit logs, on-prem
90%↑ GPU utilization
2 wks → 1 minDev environment setup
10×Workload throughput
Architecture

How they fit together

One stack from GPU pool to paying user. AI-Stack runs the compute, ixCSP turns it into services.

01 · AI-Stack

Worker Cluster (GPU Pool)

The compute plane. GPU resources are scheduled, partitioned, and pooled across nodes and high-speed storage.

02 · AI-Stack

Management Cluster

The control plane. Multi-tenant governance, quotas, monitoring, and FinOps across workloads.

03 · ixCSP

Cloud Service Gateway

The service layer. Connects models and compute resources with API access, metering, and billing.

Event information

SC26, Chicago

The International Conference for High Performance Computing, Networking, Storage, and Analysis.

Sun15
Mon16
Tue17
Wed18
Thu19
Fri20
Exhibition days Conference days
Exhibition

November 17–19, 2026Full conference: November 15–20

Venue

McCormick PlaceChicago, Illinois, USA

Booth

#6111

Contact our team

Meet us at booth #6111

Reserve a 1:1 meeting for AI-Stack or ixCSP with our leadership team from Taiwan, Japan and Korea.

Book a Meeting / Partner with Us ↗

Tell us your preferred day (Nov 17, 18 or 19) and the product you’re interested in. Walk-ins are welcome.

In accordance with applicable personal data protection laws, we are committed to safeguarding your personal data and providing you with control over how it is used.

By clicking “Accept All,” you consent to our use of cookies to enhance your browsing experience, help us analyse website performance and usage, and deliver relevant marketing content. You may manage your cookie preferences below. By clicking “Confirm Selection,” you agree to the settings currently selected.

Manage cookies

Privacy Preference Center

In accordance with applicable personal data protection laws, we are committed to safeguarding your personal data and providing you with control over how it is used.

By clicking “Accept All,” you consent to our use of cookies to enhance your browsing experience, help us analyse website performance and usage, and deliver relevant marketing content. You may manage your cookie preferences below. By clicking “Confirm Selection,” you agree to the settings currently selected.

View privacy policy

Manage Consent Settings

Necessary Cookies

Enable All

Necessary cookies are required for the proper operation and security of this website, as well as for remembering your cookie preferences. These cookies cannot be disabled through this preference centre. You may still block cookies through your browser settings; however, doing so may affect the availability or functionality of certain features on this website.