Compute Cloud

Cuda 12.6 Update News December 2025 _verified_ -

Deploy powerful GPUs for demanding AI workloads. Choose from on-demand or reserved clusters aimed at expansion

Cloud GPU Illustration

Run workloads on a range
of high performance GPUs

Access global pools such as B200, H200, MI300, etc. Deploy where AI workloads require, fully compliant and uncompromising.

Cuda 12.6 Update News December 2025 _verified_ -

Heterogeneous Memory Management (HMM) has been a long-term goal for CUDA, allowing GPUs to access CPU memory transparently. However, "oversubscription"—running out of GPU memory and spilling to system RAM—has historically resulted in steep performance cliffs.

For those operating on the 12.6 update path as of December 2025, the following features define the experience: Nsight Compute Release History - NVIDIA Developer cuda 12.6 update news december 2025

NVIDIA CUDA-GDB CUDA Toolkit 13.2 Release. Latest. This tag represents the state of CUDA-GDB sources for the CUDA Toolkit 13.2 rel... GitHub CUDA GPU Compute Capability - NVIDIA Developer Table_title: CUDA GPU Compute Capability Table_content: header: | Compute Capability | Data Center | row: | Compute Capability: 12... NVIDIA Developer CUDA 12.6 Update 2 Support · Issue #2142 · OE4T/meta-tegra Feb 9, 2026 — Heterogeneous Memory Management (HMM) has been a long-term

If you are still on 12.6, plan a migration to CUDA 12.9 (for compatibility) or CUDA 13.x (for future hardware and performance). The CUDA 12.x series is winding down, and all December news points toward CUDA 13.1 as the main event. Latest

Previously, developers saw kernel execution times in isolation. The updated Nsight Systems now integrates with popular front-end frameworks (PyTorch 2.6, JAX) to map GPU kernels directly to the computational graph of the model.

GPU

NVIDIA H100 PCIe

from $2.45 /hr

Scale with high performance H100 GPUs on our reserved cloud.

GPU

NVIDIA HGX B200

Pricing on request.

Scale with high performance B200 GPUs on our reserved cloud.

GPU

NVIDIA HGX B200

Pricing on request.

Scale with high performance B200 GPUs on our reserved cloud.

GPU

NVIDIA HGX B200

Pricing on request.

Scale with high performance B200 GPUs on our reserved cloud.

GPU

NVIDIA HGX B200

Pricing on request.

Scale with high performance B200 GPUs on our reserved cloud.

Spin up an instance in seconds

Choose how you work. Lightx gives you full control to run AI workloads where they matter most, with the speed and flexibility to scale globally and meet compliance at every step.

Secondary

Dashboard

Deploy your infrastructure quickly with our web dashboard

CLI tool

Handle your compute programmatically with our CLI tooling

API

Use our API to build applications and manage infrastructure

On-demand clusters

Deploy scalable GPU clusters
on-demand in minutes

Launch production ready clusters in minutes with our on-demand platform. Manage your cluster with our API, CLI or dashboard and scale as needed with our resizing & volume storage features.

View pricing
End-to-end clusters

Large scale GPU cluster
builds for AI and HPC

Our end-to-end cluster infrastructure service provides access to the latest enterprise grade NVIDIA GPUs. From research labs and startups to enterprise teams, we can provide clusters to faciliate your needs.

Learn more
Virtual machines

Flexible virtual machines for
agile workloads

Launch virtual machines in seconds and scale resources to match your needs. Ideal for development, testing, and lightweight production, with minute-based billing, high-speed NVMe storage, and full flexibility to configure as you grow.

View pricing
Bare metal

Dedicated bare metal for
performance and control

For teams that need direct access to hardware, our bare metal servers deliver maximum performance without virtualization. Tailor your setup for AI training, complex processing or compliance‑driven workloads, with infrastructure designed for reliability and customisation.

Learn more

Run AI workloads with
confidence

Maintain full control over your GPU operations with intelligent monitoring, cost management, and enterprise‑grade compliance.

Realtime monitoring

Access live GPU metrics including utilisation, latency and memory so your operations team can identify and address issues before they impact performance or SLAs.

Global compliance

Certified UK, EU and US deployments ensure GDPR, NHS, PCI‑DSS and HIPAA alignment, keeping data securely within your chosen region.

Cost intelligence

Per‑second billing, exportable reports and trend dashboards help you forecast spend, manage resources efficiently and demonstrate return on investment.

Flexible commitments

Scale on demand or choose longer‑term plans for predictable costs with the flexibility to adjust tiers as your workloads evolve.

Fully managed service

Our engineering team manages provisioning, patching and optimisation from start to finish, so your team can focus on delivering results without worrying about the underlying infrastructure.

Enterprise Cloud

Financial Intelligent Computing Industry Cloud

By sharing AI computing resources, we aim to help small and medium-sized financial institutions reduce construction thresholds and costs, achieving inclusive technology. We support the intelligent transformation of the financial industry, break technical barriers, and promote the rapid implementation of applications such as intelligent risk control and customer analysis.

Challenges and pain points

01

Cost-Benefit Analysis

How to ensure service quality while reasonably controlling costs and providing cost-effective solutions for large institutions providing financial intelligent computing industry cloud services to small and medium-sized financial institutions is an important challenge faced by large institutions.

02

Financial Compliance Challenges

With stricter regulation, large institutions need to ensure that cloud services in the financial intelligent computing industry are both technologically advanced and fully compliant with domestic and foreign financial regulations to safeguard customer data security and avoid compliance risks.

03

Weak operation and maintenance capabilities

How to ensure service quality while reasonably controlling costs and providing cost-effective solutions for large institutions providing financial intelligent computing industry cloud services to small and medium-sized financial institutions is an important challenge faced by large institutions.

Heterogeneous Memory Management (HMM) has been a long-term goal for CUDA, allowing GPUs to access CPU memory transparently. However, "oversubscription"—running out of GPU memory and spilling to system RAM—has historically resulted in steep performance cliffs.

For those operating on the 12.6 update path as of December 2025, the following features define the experience: Nsight Compute Release History - NVIDIA Developer

NVIDIA CUDA-GDB CUDA Toolkit 13.2 Release. Latest. This tag represents the state of CUDA-GDB sources for the CUDA Toolkit 13.2 rel... GitHub CUDA GPU Compute Capability - NVIDIA Developer Table_title: CUDA GPU Compute Capability Table_content: header: | Compute Capability | Data Center | row: | Compute Capability: 12... NVIDIA Developer CUDA 12.6 Update 2 Support · Issue #2142 · OE4T/meta-tegra Feb 9, 2026 —

If you are still on 12.6, plan a migration to CUDA 12.9 (for compatibility) or CUDA 13.x (for future hardware and performance). The CUDA 12.x series is winding down, and all December news points toward CUDA 13.1 as the main event.

Previously, developers saw kernel execution times in isolation. The updated Nsight Systems now integrates with popular front-end frameworks (PyTorch 2.6, JAX) to map GPU kernels directly to the computational graph of the model.

Solutions

LightX DCI Solutions enable hyperscale and enterprise customers to move data securely, seamlessly, and in real time across geographically distributed infrastructures.

Al Cluster

Engineered for Al clusters and data-driven networks, LightX 100G to 1.6T modules enable breakthrough throughput with power efficiency, critical for distributed Al/ML training and real-time data processing. With intelligent thermal management and backward compatibility, LightX accelerates deter- ministic connectivity for Al factories and cloud-scale networking while reduc. ing TCO versus standard modules.

DCl Solutions

LightX DCl solution enable hyperscale and enterprise customers to move data securely, seamlessly, and inreal time across geographically distributed infrastructures,Built on advanced OTN (Optical Transport Network) technology and managed through the Gaussian Network Management System (NMS), our DCl architecture delivers deterministic latencycarri er-grade reliability,and comprehensive network visibility.

About LightX

LightX is a full‑stack cloud and connectivity provider built for the AI era. We deliver high‑performance infrastructure that powers compute‑intensive and latency‑sensitive workloads across industries. Our portfolio spans four core domains — Compute Cloud, Enterprise Cloud, AI Cluster Solutions, and Data Center Interconnect (DCI).

The LightX Compute Cloud provides elastic GPU and CPU compute optimized for AI, HPC, and large‑scale data processing. Enterprise Cloud solutions serve mission‑critical sectors including quantitative trading, education, and healthcare — combining reliability, security, and regulatory compliance.

Through our AI Cluster Solutions, we offer end‑to‑end design, deployment, and operations supported by our proprietary orchestration platform, dramatically improving resource utilization and scalability. Our DCI Solutions leverage advanced OTN transmission and a network‑management platform to deliver ultra‑low‑latency, high‑throughput interconnects between data centers.

At LightX, we believe infrastructure is intelligence. By unifying compute, cloud, and interconnect into a single, high‑performance ecosystem, we empower enterprises to accelerate innovation, scale AI workloads, and connect the future of intelligent computing.