Edge Infrastructure Review

Crusoe expands AI infrastructure race with 900 MW Abilene build for Microsoft

Stephen Mayhew — Fri, 17 Apr 2026 06:00:43 +0000

Neocloud provider Crusoe announced a new 900 MW AI factory campus in Abilene, Texas, dedicated to supporting large-scale AI workloads for Microsoft, with the total site capacity expected to reach 2.1 GW. The campus will consist of two new buildings and a powerplant on-site providing grid resilience as well supporting next-gen AI infrastructure. Construction has begun, with the first building planned to be operational by mid-2027, maintaining Crusoe’s rapid pace of infrastructure buildout. “Crusoe is building a new AI factory campus in Abilene, purpose-built for the demands of next-generation AI," says Chase Lochmiller, co-founder and CEO of Crusoe. “By integrating 900 megawatts of new on-site power generation, we will continue building the industrial foundation for American AI – at a velocity the industry has never seen." It will also be expected to provide thousands of jobs during the construction phase, and hundreds more that are permanent job opportunities that could seriously enhance North Abilene's economy. With integrated energy-efficient infrastructure, the new campus will come equipped with a ~900 MW on-site power plant, ultra-high-density compute and water-efficient cooling systems. The nearby Abilene campus serves as a large booster of local tax revenue, and that's expected to compound with its expansion. Crusoe's partnership with Microsoft shows how they are addressing the need for reliable and scalable infrastructures to satisfy increasing AI demands. Crusoe solutions deliver effective and energy first giga-scale AI infrastructure that enables the future of AI.

Aker BP taps Armada for offshore modular AI data center to process drilling data at the edge

Stephen Mayhew — Thu, 16 Apr 2026 06:00:16 +0000

Oil company Aker BP is collaborating with edge AI infrastructure solutions provider Armada to deliver a modular data center for offshore drilling on the Norwegian Continental Shelf. Galleon allows on-site processing and analysis of drilling data at the rigs, overcoming issues such as connectivity challenges and delays in decision-making. Running AI models locally facilitates more accurate prediction and prevention of equipment failures, reducing downtime. It also reduces the risk of cyber-attacks by lowering dependence on external networks. Aker BP will unify vendor applications based on a single edge platform, streamlining compliance and cost savings, while allowing operations to occur from remote locations. Deployment starts with a single reference Galleon on 1 rig laying down an installable repeatable blueprint. The project capitalizes on Armada’s history of providing modular AI infrastructure in remote environments, including a U.S. Navy deployment project from 2025. The partnership will improve operational efficiency, resilience, and autonomy of offshore drilling operations.

Leaseweb builds out European sovereign cloud with programmable networking and AI compute

Stephen Mayhew — Wed, 15 Apr 2026 10:00:22 +0000

Cloud services and Infrastructure-as-a-Service (IaaS) provider Leaseweb is advancing Europe's sovereign cloud initiatives through its European Cloud Campus project, part of the IPCEI-CIS program. Notable advancements include cloud infrastructure expansion, autoscaling, load balancing and privacy network storage; in addition to programmable virtual overlay networking for improved flexibility and security. “Over the past year, we’ve made significant progress in turning the vision of a sovereign European cloud into reality,” says Robert van der Meulen, director product strategy at Leaseweb. “From expanding our compute and container platforms to releasing open APIs and developer tools, these efforts demonstrate how industry and technology are coming together to build flexible, secure, and scalable cloud infrastructure for Europe. Looking ahead, we remain committed to advancing automation, network flexibility, and ecosystem integration, ensuring that Europe’s digital sovereignty is supported by practical, world-class infrastructure.” The company also launched new tools (open APIs, a Terraform provider, and platform monitoring systems) to help automate developer integration. In 2025, Leaseweb actively supported the cloud community by organizing events such as the Leaseweb Tech Summit and taking on contributions to 58 open-source projects. Emphasizing collaboration between industry stakeholders to build secure, scalable, and flexible cloud infrastructure for Europe. Leaseweb is the only Dutch cloud provider to join the IPCEI-CIS program directly and support Europe's digital sovereignty. Leaseweb's offerings are based on the core tenets of digital sovereignty and customer control over data. In early 2025 Leaseweb expanded its global cloud portfolio with NVIDIA GPUs to power AI and HPC workloads.

Akamai pushes AI inference to the edge with orchestrated GPU grid across 4,400 sites

Stephen Mayhew — Tue, 14 Apr 2026 10:00:49 +0000

Akamai launched the Akamai Inference Cloud late last year, the first global-scale implementation of NVIDIA AI Grid, enabling distributed AI inference across 4,400 edge locations. Akamai empowers a platform that delivers AI systems using NVIDIA AI infrastructure and optimizes workload routing with Akamai's network to offer the best possible latency, cost, and performance. Intelligent orchestration optimizes the cost-efficiency and response time of AI applications via improved “tokenomics” in Akamai’s AI Grid, resulting in throughput gains. "AI factories have been purpose-built for training and frontier model workloads and centralized infrastructure will continue to deliver the best tokenomics for those use cases," says Adam Karon, COO and general manager, Cloud Technology Group, Akamai. "But real-time video, physical AI, and highly concurrent personalized experiences demand inference at the point of contact, not a round trip to a centralized cluster. Our AI Grid intelligent orchestration gives AI factories a way to scale inference outward, leveraging the same distributed architecture that revolutionized content delivery to route AI workloads across 4,400 locations, at the right cost, at the right time." It reduces latency by processing requests at the edge to support use cases for AI in real time, such as gaming, financial services, media, and retail. To do so, Akamai has thousands of NVIDIA RTX PRO 6000 GPUs as part of its infrastructure with high density compute capabilities to offer enterprise scale GPU services for large-scale AI workloads and multi-modal inference. The platform empowers enterprises to deploy adaptive, context-aware AI agents in both centralized and distributed architectures through this model. Early adoption is evident across the gaming, finance and media sectors; a recent $200 million service agreement announced last month by Akamai validates enterprise demand.

How AI-RAN paves the way to performance and profits

Rob Hughes — Mon, 13 Apr 2026 11:54:05 +0000

By Rob Hughes, Head of Wireless Marketing for 1Finity, a Fujitsu company. As data demands escalate, today’s networks are growing and evolving at a rapid pace. Mobile networks, in particular, have changed substantially in response to expectations of readily available hyperconnectivity at our fingertips everywhere we go. At the same time, mobile network operators (MNOs) are facing a confluence of pressures and challenges. Market consolidation, increasing competition, rising supply chain costs and shrinking profits all contribute to an urgency to offer profitable new services while reducing total cost of ownership (TCO). Growing reliance on artificial intelligence (AI) offers promise for MNOs to address these challenges. The important question is not whether AI belongs in mobile networks, but how operators will turn AI into both performance gains and new revenue.

The Power of AI

With the evolution to 5G, the mobile radio access network (RAN) has seen significant changes. Increased throughput speeds have accelerated the adoption of video streaming and encouraged a wide range of content rich applications that have driven increased demands on the network. As a result, MNOs are looking for solutions to improve spectral efficiency and capacity while reducing TCO. Indeed, we have seen quite a bit of positive enthusiasm around harnessing the power of AI automation to streamline management and improve RAN performance. With AI optimization, MNOs can boost network performance to deliver an outstanding customer experience. Taking this a step further, there is genuine interest among MNO pioneers to fully integrate AI resources into the RAN infrastructure and implement AI-RAN technology. With the technical breakthrough offered by AI-RAN, MNOs can further enhance RAN performance by increasing network throughput, reducing energy consumption and improving spectral efficiency. While these network enhancements are exciting, even more enticing are the business opportunities presented by AI-RAN.

Pinpoint the Real Opportunity

AI-RAN offers significant monetization potential and competitive advantages for network operators. In a recent Future of AI survey by Heavy Reading analysts, however, only 38 percent of MNOs cited ‘the ability to offer AI as a service’ among the top three benefits they expected from AI-RAN. For some network operators the evolution to AI-RAN is seen as a challenge to surmount, requiring effort and financial investment. Yet, AI offers the ability to deliver promising new services - particularly through AI-as-a-Service (AIaaS) and GPU-as-a-Service (GPUaaS) - that can create the new revenue streams needed to pay off 5G and AI-RAN investments. In fact, the GPUaaS market is expected to exceed $25 billion by 2030 with an annual growth rate of 26.5 percent. That’s because cloud-based GPU resources are increasingly desirable for AI training, batch AI/data analytics, video rendering and simulation/engineering workloads.

Disrupt the Cycle

Historically, emerging network technologies have faced a "chicken and egg" dilemma regarding investment. Software developers typically hesitate to create content until infrastructure is widespread, while MNOs often delay deployment until there are sufficient revenue-generating applications to justify the cost. The nature of AI-RAN infrastructure enables MNOs to disrupt this traditional cycle. This can be achieved by leasing excess compute capacity within the network, generating revenue without waiting for a broad base of applications that need the low latency or data sovereignty of AI-RAN to become available. Much like a high-end restaurant must maximize table turnover to offset premium real estate costs, MNOs can monetize their valuable AI-RAN resources to pay for themselves. Because base station traffic fluctuates throughout the day, the compute capacity of a RAN distributed unit (DU) is built to handle the absolute peak traffic of the busiest day of the year, plus a margin for future expansion. Consequently, significant compute capacity remains idle most of the time. Today that excess capacity sits idle, but that doesn’t need to be the case. Given the growing market for GPUaaS, MNOs have a promising opportunity to generate near-term revenue while they gradually introduce new AI services.

Fast-track to Revenue

Current GPUaaS and AIaaS solutions enable highly flexible billing models, including hourly or even minute-level options, without long-term obligations. This allows MNOs to sell GPU capacity contracts that fluctuate according to daily demand cycles. Moreover, MNOs maintain operational control by reserving the right to lease this capacity only when it is not required for their own internal needs, taking full advantage of idle capacity. By leveraging AI resources to monetize excess capacity, MNOs can effectively fast-track to AI-RAN immediately, reaping the performance benefits while generating revenue to cover initial costs. As the MNO’s own demand for AI compute resources grows, they can then gradually scale back the amount of capacity available for lease.

The Road to AI-RAN

The transition to AI-RAN requires more than just new technology - it requires new business models, skillsets and mindsets. We are just at the beginning of a long journey that will bring a host of new innovations along the way. And as with any new technology, a small number of pioneers and innovators are pushing the envelope and deploying something new in order to spur AI-RAN advancements. Although the pioneers are ready to implement AI-RAN now, the majority of deployments will be a year or two behind, but that’s par for the course. Network change can be slow to materialize, such as was the case with 5G Standalone (SA) networks, but adoption will happen in time. Those MNOs willing to start the journey now will have more time to learn both the AI-RAN technology as well as the necessary changes in operations and business models. That head start will put them firmly ahead of the competition on the path to intelligent, next-generation AI-powered networks.

About the author

Rob Hughes is Head of Wireless Marketing at 1Finity, overseeing solution and marketing strategy for the company’s wireless portfolio. With over 20 years’ experience in telecommunications, he has a broad range of expertise in wireless, optical, wireline, enterprise, open networking, analytics and cable solutions. Rob has a unique perspective based on both business and technical experience. He studied Engineering at University of Victoria, British Columbia, and taught Marketing Strategy at the University of Texas.

StorMagic and HiveRadar target off-grid compute with mobile edge data centers

Stephen Mayhew — Mon, 13 Apr 2026 10:00:17 +0000

StorMagic and HiveRadar have partnered to deliver a mobile edge computing solution combining StorMagic's SvHCI software and HiveRadar's Portable Edge Data Center (P-EDC). It is a solution purpose-built for secure, robust and high-performance IT infrastructure in remote, mobile and off-grid environments. "Whether operating in disaster recovery zones, forward-operating environments, mobile command centers, remote film sets or distributed retail sites, our customers can now deploy a secure, high-performance virtualized environment in minutes," says Rahul Narsimhan, CEO, HiveRadar. "Together, StorMagic and HiveRadar customers will benefit from rapid deployment and travel-ready mobility, which is especially beneficial for defense, emergency response, energy, retail, entertainment and industrial operations users." Key features include data-at-rest encryption, VM import for workload migration, caching for performance, real-time health monitoring, and low-latency field compute and storage. With 5G and satellite integration, it also supports off-grid operations, quick deployment and secure connectivity. Customers benefit from high availability, failover capabilities and easy virtualization of mobile infrastructures. While HiveRadar specializes in ruggedized IT infrastructure solutions, StorMagic builds lightweight, edge-optimized virtualization solutions. The joint solution is now available via StorMagic's global partner and reseller network. Late last year StorMagic and SNUC launched rugged HCI appliances for demanding edge deployments.

Lambda doubles down on NVIDIA stack with 10,000+ Blackwell GPUs and CPO networking push

Stephen Mayhew — Fri, 10 Apr 2026 10:00:12 +0000

Lambda recently announced it’s becoming a launch partner for the NVIDIA Vera CPU platform and NVIDIA STX. The GPU-native AI infrastructure provider will deploy NVIDIA Quantum-X800 infiniBand photonics co-packaged optics in an AI factory with 10,000+ NVIDIA Blackwell Ultra GPUs. Lambda's bare metal instances made its way out of the lab and into the core cloud offering, giving users direct access to hardware while avoiding virtualization overhead for distributed AI training workloads. Designed for launching thousands of parallel AI environments, the NVIDIA Vera CPU platform enables maximally high memory bandwidth which optimizes reinforcement learning and agentic AI workloads. The NVIDIA STX is a modular architecture for AI storage that augments inference, analytics, and training with next-gen hardware optimized KV-cache management. Co-Packaged Optics (CPO) networking enables faster, cost-efficient AI infrastructure suitable for large-scale AI factories, alleviating major efficiency bottlenecks found in current approaches. “The race to build AI factories isn’t won on GPU counts alone,” says Dave Salvator, director of accelerated computing at NVIDIA. “Network architecture is what determines whether those systems can perform at scale. Getting this right is what allows AI infrastructure to power services used by hundreds of millions of people around the world.“ Lambda oversees one of the largest deployments of NVIDIA Quantum-X800 CPO switches, highlighting how critical network architecture is when scaling AI systems. These announcements further bolster Lambda's AI infrastructure platform, which empowers frontier labs, enterprises, and hyperscalers with proven and energy-efficient workhorses built for reliability at scale. Lambda continues its mission to make AI compute ubiquitous, leveraging a decade-long collaboration with NVIDIA to advance its Superintelligence Cloud platform.

DDN and Zadara target sovereign AI deployments with multi-tenant NVIDIA factory stack

Stephen Mayhew — Thu, 09 Apr 2026 10:00:08 +0000

DDN and Zadara announced a partnership to deliver high-performance AI infrastructure for sovereign clouds and multi-tenant AI factories, leveraging NVIDIA reference designs. The integration combines the DDN EXAScaler AI data platform with Zadara’s cloud-native, AI-optimized infrastructure for scalable, secure, and efficient AI deployments. The solution solves critical problems of enterprise AI such as real-time infrastructure complexity, GPU performance, compliance and multi-tenant governance that hinder faster deployment of AI. “NVIDIA reference designs are accelerating the broad adoption of AI factories. Yet organizations need a cloud-native platform that can operationalize them simply and efficiently in real-world, multi-tenant environments while complying with data sovereignty regulations,” says Yoram Novick, CEO at Zadara. “DDN EXAScaler brings the high-performance AI data foundation needed to meet demanding sovereign and enterprise requirements while Zadara delivers the orchestration, isolation, and policy control required to deploy AI infrastructure quickly, securely, and efficiently.” Zadara’s platform decouples AI infrastructure operations with performance, compliance and tenant isolation in accordance with NVIDIA best practices. This partnership powers NVIDIA-enabled AI factories for secure multi-tenancy, predictable performance and policy-based orchestration with rapid time to import. The integrated solution enables sovereign AI and multi-tenant environments with high throughput, GPU-aware scheduling, and compliance controls. This initiative aims to make enterprise AI more accessible, affordable and practical by minimizing operational complexity while improving time-to-value. The partnership between DDN and Zadara reinforces the need for AI factories as a cornerstone of enterprise, telcos, and service providers.

Premio targets multi-camera edge AI with new Jetson Orin systems

Stephen Mayhew — Wed, 08 Apr 2026 10:00:45 +0000

Rugged edge AI and embedded computing provider Premio debuted two new models to its JCO-Series of rugged NVIDIA Jetson Orin edge AI computers at ISC West 2026. Designed for next-gen vision AI applications, these models power multi-camera deployments used in multiple industries including security, transportation and smart infrastructure by supporting up to four GMSL2 cameras. “As AI-driven vision systems continue to transform industries such as security, transportation, and smart infrastructure, organizations need reliable computing platforms capable of processing data directly at the edge," says Dustin Seetoo, VP of product marketing at Premio. “With the JCO-1000-ORN-B and JCO-1000-ORN-C, we are expanding the JCO-1000-ORN Series to support multi-camera AI deployments while maintaining the rugged performance required for real-world edge environments.” These supercomputers come equipped with NVIDIA Jetson Orin NX/Nano modules, which provide up to 157 TOPS of AI performance and ruggedized I/O for edge environments. Highlights will include dual LAN or rugged M12 LAN connectors, high-speed NVMe M.2 storage, 4G/5G and Wi-Fi support including CAN Bus and extended operating temperature range (-20°C to 55°C). The systems are small form factor, vehicle-ready with 9–36VDC input and certified to CE, FCC and UL global standards. The new models will be available for order by mid-Q2 2026.

Hosted.ai raises $19M to tackle GPU underutilization and reshape AI infrastructure economics

Stephen Mayhew — Tue, 07 Apr 2026 22:42:11 +0000

AI infrastructure startup hosted·ai secured $19M in seed funding to simplify and optimize GPU infrastructure for AI, led by Creandum with participation from other investors like Repeat VC and People Ventures. The startup focuses on the inefficiencies of GPU infrastructure, including high costs, low utilization — with average usage at just 40% — and limited access to regional suppliers. hosted·ai's software stack includes: hosted·ai: A GPUaaS platform improving GPU utilization by up to 5x, reducing costs, and enabling resource sharing. packet·ai: A neocloud service leveraging optimized GPU infrastructure for competitive pricing. GPUaaS.com: A matchmaking service connecting enterprises with GPU providers for scalable solutions. The company will target the “GPU waste” problem and allow regional service providers a piece of this AI infrastructure pie. “The GPU market has a waste problem, not a scarcity problem,” says Ditlev Bredahl, CEO of Hosted.ai. “We’ve spent 25 years building infrastructure software that makes service providers competitive and the GPU opportunity is the biggest we’ve seen. This funding lets us move faster: more platform, more partners, more regions. We’re building the operating system for the GPU economy, and this round puts us in a strong position to do exactly that.” Some of the founding team members have previous experience working with infrastructure technologies at VMware, NVIDIA or XenSource. Founded in 2024 and launched in 2025, hosted·ai is a global company offering products across the US, EMEA and Asia-Pacific. Their mission is to revolutionize the economics of GPU computing so that AI infrastructure becomes affordable and accessible for developers and companies alike.