Skip to content
Breaking
Most tourists skip this Japanese town. A Singaporean just spent seven figures opening a hotel there.Bursa Malaysia ends flat as investors await earnings for directionAWS launches RM1.5 billion cloud-skills training for Malaysian studentsNVIDIA launches NVLink Fusion to speed semi-custom AI factory buildsNVIDIA ships Groq 3 LPX to power Vera Rubin’s agentic AIUS Section 338 tariffs raise Canada's average US tariff to 6.27%Global food supply has become ‘collateral damage’ in conflicts: UNIran sanctions threaten RM1.8bn in Malaysian tradeIran warns new US sanctions will bring 'consequences' as rial hits record lowUN peacekeepers killed in South Sudan ambushSelangor declares Sept 1 a public holiday after Sukma successSelangor top Sukma medal tally - their best in 28 yearsCan Trump’s ‘economic D-Day’ force Iran to back down?Hot hatch record: Hyundai Ioniq 5 N delivers 641 hpDBS vs OCBC vs UOB: Which Bank Delivered the Best Results and Dividend?JCI Rises as Asian Markets Rally on U.S. Economic Data, China SignalsNew Economic Gateways: How Capital Is Redrawing Asia’s Urban MapFBM KLCI rebounds on US growth upgradeCGS International lifts KLCI 2026 target to 1,810 on ringgit strengthGamuda to export AI tunnel-boring system to US and IndiaMost tourists skip this Japanese town. A Singaporean just spent seven figures opening a hotel there.Bursa Malaysia ends flat as investors await earnings for directionAWS launches RM1.5 billion cloud-skills training for Malaysian studentsNVIDIA launches NVLink Fusion to speed semi-custom AI factory buildsNVIDIA ships Groq 3 LPX to power Vera Rubin’s agentic AIUS Section 338 tariffs raise Canada's average US tariff to 6.27%Global food supply has become ‘collateral damage’ in conflicts: UNIran sanctions threaten RM1.8bn in Malaysian tradeIran warns new US sanctions will bring 'consequences' as rial hits record lowUN peacekeepers killed in South Sudan ambushSelangor declares Sept 1 a public holiday after Sukma successSelangor top Sukma medal tally - their best in 28 yearsCan Trump’s ‘economic D-Day’ force Iran to back down?Hot hatch record: Hyundai Ioniq 5 N delivers 641 hpDBS vs OCBC vs UOB: Which Bank Delivered the Best Results and Dividend?JCI Rises as Asian Markets Rally on U.S. Economic Data, China SignalsNew Economic Gateways: How Capital Is Redrawing Asia’s Urban MapFBM KLCI rebounds on US growth upgradeCGS International lifts KLCI 2026 target to 1,810 on ringgit strengthGamuda to export AI tunnel-boring system to US and India
AI Edge

NVIDIA launches NVLink Fusion to speed semi-custom AI factory builds

NVIDIA unveiled NVLink Fusion, a platform that links custom XPUs to its proven AI infrastructure to cut deployment time and risk for hyperscalers building semi-custom AI factories.

Source: NVIDIA Newsroom · August 24, 2026 at 8:28 PM · AI-assisted report

Single-source
NVIDIA launches NVLink Fusion to speed semi-custom AI factory builds
Image: blogs.nvidia.com

KUALA LUMPUR, 25 AUGUST 2026 —

Listen to this article

DomainFork Audio · read aloud

Share

NVIDIA unveiled NVLink Fusion, a platform that links custom XPUs to its proven AI infrastructure to cut deployment time and risk for hyperscalers building semi-custom AI factories.

Market Impact

The technology addresses a bottleneck in bringing XPUs to market: the need to design not just accelerators but the entire factory—rack-scale architecture, networking, software and supply chains. NVLink Fusion provides pre-validated NVLink scale-up fabrics, rack designs and management tools so builders can focus innovation on silicon while relying on established infrastructure for the rest.

“The NVLink Fusion program lets customers choose the CPU architecture, performance level and software capabilities that best fit their workloads,” said Tim Wilson, vice president and general manager of data center silicon engineering at Intel.

The platform delivers up to three times lower XPU-to-XPU latency and ten times higher packet rates than Ethernet-based alternatives by extending sixth-generation NVLink across a 72-XPU domain. It includes NVLink-C2C links that deliver up to six times the energy efficiency of PCIe when connecting XPUs to NVIDIA Vera CPUs or other ecosystem CPUs, removing bottlenecks in agentic systems.

For end-to-end performance, NVIDIA GB300 NVL72 systems using NVLink Fusion show higher throughput and interactivity than non-NVL72 configurations. Future NVLink roadmap configurations scale to domains of up to 1,152 accelerators.

NVLink Fusion is supported by the NVIDIA MGX rack-scale architecture and the same supply chain used for MGX-based systems such as NVIDIA Vera Rubin NVL72. Manufacturing partners manage design and integration, while MGX suppliers provide rack, cooling, power and emerging 800 VDC designs.

“With Vera Rubin NVL72, we see almost 100% automation of system builds on the manufacturing line,” said Jack Luoh, head of product and solution at QCT and Quanta Computer. “Most of those investments can be leveraged if the XPU uses NVLink Fusion.”

Operators can deploy GPU and XPU systems in shared racks with unified networking, cooling and power, then reprovision capacity as workload demand, silicon supply and business priorities change.

“The NVLink Fusion program lets customers deploy a rack-level solution with an NVIDIA GPU, then decouple XPU development and pace it differently,” said Vince Hu, corporate senior vice president and general manager of the data center and computing business group at MediaTek.

NVLink Fusion aligns with the NVIDIA DSX reference architecture for AI factories, which codesigns buildings, power, cooling, compute and networking. The NVIDIA Omniverse DSX AI Factory Blueprint offers a digital twin and open reference design for gigawatt-scale AI factories.

“NVLink Fusion allows hyperscalers or custom ASIC designers to integrate their own CPU or XPU and bridge NVIDIA technology with third-party processes to create a unified rack-scale architecture,” said Lie-Szu Juang, chair and chief strategy officer at GUC.

Rack serviceability is built in. Reference compute trays use 100% liquid cooling with no fans, cables or hoses, and allow trays to be removed while the rest of the rack remains operational. NVLink Switch trays are also liquid cooled and support continued operation during service.

“With NVLink Fusion we can use the proven NVL72 rack design to shorten time-to-market and access multiple suppliers to deliver more systems to our customers,” said CC Lee, senior hardware development manager at Annapurna Labs, an Amazon company.

Software integration includes NVIDIA NCCL for distributed workloads, NVIDIA Dynamo and NIXL for disaggregation, and NVIDIA Mission Control for cluster management, telemetry and debugging.

For Malaysian operators, the launch lowers the barrier to deploying semi-custom AI infrastructure by letting local teams integrate XPUs with NVIDIA’s ecosystem while cutting upfront design risk.

Related: Amazon

Reporting based on NVIDIA Newsroom. Figures and claims are subject to revision as the story develops. DomainFork publishes editorial context, not investment advice — see our editorial standards.