By: Network-Switch.com | August 20, 2026
- 1. Quick Summary
- 2. NVIDIA Maps an 800 VDC Upgrade Path for Higher-Density AI Factories
- 3. Dell ObjectScale Pushes 9.83 PB of Raw Flash Capacity Into 2U
- 4. Indonesia Opens Its First University-Based NVIDIA AI Technology Center
- 5. NVIDIA Secures Gigawatt-Scale AI Infrastructure at Ohio's PORTS-Pike Campus
- 6. AMD Reports 4x Progress Toward Its Rack-Scale AI Energy-Efficiency Goal
- 7. Arista Runs Hands-On EOS, Routing, VXLAN and CloudVision Training in Canada
- 8. Dell Details GB300 Local AI for Secure and Disconnected Environments
- 9. Frequently asked questions (FAQs)
- 10. Sources
Quick Summary
Global enterprise infrastructure news from August 11 through August 20 focused on power delivery, storage density, sovereign AI, gigawatt-scale compute, rack-level efficiency and network operations. Seven new developments were selected for this edition, with previously covered topics excluded.
No distinct new Juniper or HPE Aruba standalone switch, router or wireless hardware launch met the selection threshold in this review window. Recent HPE networking items centered on customer deployments and skills programs, while Arista's qualifying fresh activity was a hands-on engineering event rather than a new hardware release.
NVIDIA Maps an 800 VDC Upgrade Path for Higher-Density AI Factories
NVIDIA published a new AI-factory power architecture update on August 11. Its 800 VDC design reduces the number of power-conversion stages between the grid and accelerated compute, targeting higher rack density and better power efficiency.
NVIDIA said an MGX-compatible 800 VDC power rack is planned for the second half of 2026 and can operate inside existing AC facilities. A later row power center is designed to support up to 2 MW per rack row. NVIDIA, Google and Microsoft have been developing the architecture through the Open Compute Project with more than 80 ecosystem companies.
Dell ObjectScale Pushes 9.83 PB of Raw Flash Capacity Into 2U
Dell's August 12 AI Data Platform update detailed ObjectScale support for KIOXIA 245.76 TB NVMe SSDs. Dell describes it as the first object-storage implementation to support drives at this capacity.
A software-defined ObjectScale configuration running on PowerEdge R7725 can use 40 of the 245.76 TB SSDs to provide 9.83 PB of raw capacity in 2U. The density is aimed at AI factories, analytics, content repositories, cyber resilience and large unstructured-data environments.
Indonesia Opens Its First University-Based NVIDIA AI Technology Center
Universitas Gadjah Mada, Indosat Ooredoo Hutchison, Indonesia's Ministry of Communication and Digital Affairs and NVIDIA opened the UGM Indosat NVIDIA AI Technology Center in Yogyakarta on August 14.
NVIDIA described it as Indonesia's first university-based AI technology center. The initiative gives researchers, students and developers access to accelerated computing, AI software and development resources while supporting locally relevant AI research and talent development.
NVIDIA Secures Gigawatt-Scale AI Infrastructure at Ohio's PORTS-Pike Campus
On August 17, NVIDIA and SB Energy announced that NVIDIA will be the exclusive AI compute infrastructure provider for the PORTS-Pike Technology Campus in Pike County, Ohio. OpenAI is the customer for the planned capacity.
The initial deployment is designed for 4.25 IT-GW, with an option covering additional capacity toward an 8 IT-GW campus. The site will use NVIDIA's DSX AI factory platform across GPUs, CPUs and networking. NVIDIA also announced a $1.5 billion investment in SB Energy, with campus capacity expected to come online in phases beginning in 2028.
AMD Reports 4x Progress Toward Its Rack-Scale AI Energy-Efficiency Goal
AMD said on August 18 that it has achieved an estimated fourfold improvement in AI energy efficiency from 2024 to 2026, ahead of the company's projected threefold target for this stage of its roadmap.
AMD is targeting a 20x increase in rack-scale energy efficiency for AI training and inference by 2030 versus a 2024 baseline. The company says the work spans silicon, memory, interconnects, networking, software and system-level co-design rather than treating GPU performance as an isolated metric.
Arista Runs Hands-On EOS, Routing, VXLAN and CloudVision Training in Canada
Arista held a Test Drive session in Saint John, New Brunswick, on August 18. The event gave network engineers dedicated virtual labs built around Arista EOS and modern network automation.
The technical agenda covered EOS CLI workflows, MLAG and vARP, BGP and OSPF routing, VXLAN and CloudVision provisioning and operations. This was a training and engineering event rather than a new hardware launch, but it reflects Arista's continued emphasis on a common programmable operating model across campus, data center, WAN and AI environments.
Dell Details GB300 Local AI for Secure and Disconnected Environments
Dell published an August 19 technical update on using the Dell Pro Max with GB300 for secure federal AI workloads. The system is built around the NVIDIA Grace Blackwell Ultra GB300 Superchip and is positioned for local inference where data cannot leave controlled environments.
Dell cites up to 20 petaflops of FP4 compute, 748 GB of coherent memory and support for models up to 1 trillion parameters. The update is a deployment-focused use case rather than a first announcement of the GB300 platform, with emphasis on keeping model execution, inputs and outputs local when cloud dependence is unsuitable.
Frequently asked questions (FAQs)
What were the most important enterprise infrastructure updates from August 11 to August 20, 2026?
Key developments included NVIDIA's 800 VDC AI-factory power architecture, Dell ObjectScale support for 245.76 TB NVMe SSDs, Indonesia's first university-based NVIDIA AI technology center, NVIDIA's gigawatt-scale PORTS-Pike campus agreement, AMD's rack-scale AI efficiency update, an Arista EOS and CloudVision engineering event, and Dell's secure local-AI guidance for GB300.
Why is NVIDIA moving AI factories toward 800 VDC power distribution?
Higher-voltage DC distribution can reduce conversion stages and power-delivery overhead as AI racks become denser. NVIDIA says its hybrid 800 VDC power rack can provide a transition path for existing AC facilities without requiring the entire building electrical system to be replaced.
How much raw capacity can Dell ObjectScale provide with 245.76 TB SSDs?
Dell says a 2U software-defined ObjectScale configuration can use 40 KIOXIA 245.76 TB NVMe SSDs to provide 9.83 PB of raw capacity.
What scale is planned for NVIDIA's PORTS-Pike AI campus in Ohio?
The initial deployment is designed for 4.25 IT-GW of AI-factory capacity. NVIDIA has an option covering additional capacity, while OpenAI is the customer for the planned 8 IT-GW campus.
What are the key specifications highlighted for Dell Pro Max with GB300?
Dell's August 19 secure-AI update highlights up to 20 petaflops of FP4 compute, 748 GB of coherent memory and support for AI models up to 1 trillion parameters, with local execution designed to reduce external data and connectivity dependencies.
Sources
- NVIDIA: Why Scaling AI Compute Performance Requires a New Power Architecture
- Dell: Dell AI Data Platform Delivers Breakthrough Storage Density
- NVIDIA: UGM, Indosat and NVIDIA Open Indonesia's First University AI Center
- NVIDIA: PORTS-Pike Technology Campus in Ohio
- AMD: AMD Tracks Ahead of Rack-Scale AI Energy-Efficiency Goal
- Arista: Test Drive - Saint John, New Brunswick, August 18, 2026
- Dell: Introducing the Dell GB300 for Secure Federal AI
- Dell: Deskside Agentic AI Platform