Edge Infrastructure Review

Panthalassa lands $140M from Peter Thiel to build ocean-powered AI data centers at sea

Stephen Mayhew — Fri, 15 May 2026 10:00:19 +0000

Panthalassa, an ocean technology company, secured $140 million in Series B funding. Peter Thiel led the round of funding with additional ongoing investment from John Doerr and others. Funding will be used to fabricate and deploy autonomous, ocean-powered AI computing systems called Ocean-3 nodes. “There are three sources of energy on the planet with tens of terawatts of new capacity potential: solar, nuclear, and the open ocean,” says Garth Sheldon-Coulson, co-founder and CEO of Panthalassa. “We’ve built a technology platform that operates in the planet’s most energy-dense wave regions, far from shore, and turns that resource into reliable clean power. We’re now ready to build factories, deploy fleets, and provide a sustainable new source of energy for humanity.” It enables power generation from ocean waves and utilizes it to perform AI inference computing at sea by these nodes. Panthalassa aims to address the growing demand for energy and computing while reducing the environmental impact of traditional data centers. Commercial deployments are planned for 2027, although the company's technology has been validated at prototype scale. The timing of this announcement aligns with broader industry pressures around power scarcity, grid congestion, land constraints, and the exploding energy demands of AI infrastructure, all of which are reshaping data center strategy globally. Investors like Peter Thiel and John Doerr are clearly betting that future AI capacity may increasingly require unconventional energy and deployment models beyond terrestrial hyperscale campuses. While the concept remains highly experimental and commercial deployment is still several years away, Panthalassa reflects a growing industry thesis: AI infrastructure expansion may ultimately depend as much on energy innovation and geographic diversification as on chips and cloud platforms themselves.

Mindgrove, Pinetics target edge AI, identity systems with Indian-designed chips

Stephen Mayhew — Fri, 15 May 2026 01:29:16 +0000

Mindgrove Technologies is partnering with Pinetics Technology Solutions to bring Indian-designed processors into biometric and digital identity infrastructure, positioning domestic semiconductor technology for edge AI and embedded security workloads.

The partnership aims to integrate Mindgrove’s system-on-chip (SoC) platforms into identity and authentication hardware including biometric terminals, smart cards, secure access systems and related edge devices.

The move reflects India’s broader push to expand sovereign semiconductor capabilities beyond mobile and consumer electronics into infrastructure-grade systems tied to digital identity, secure credentials and embedded AI.

Pinetics, which develops hardware and embedded systems for identity and security applications, will work with Mindgrove to integrate the chips into production-ready platforms for authentication and identity verification deployments.

The companies say the processors are designed to support low-power AI inference and secure on-device processing, reducing dependence on cloud infrastructure for real-time identity and authentication workloads.

The partnership also highlights growing demand for localized compute architectures in government and regulated-sector deployments, where data sovereignty, supply-chain control and domestic manufacturing are becoming procurement priorities.

India has increasingly positioned semiconductors and digital public infrastructure as linked strategic sectors, with identity systems, payments infrastructure and edge AI deployments creating potential demand for domestically designed silicon.

Mindgrove said the collaboration is intended to accelerate adoption of Indian-designed processors across embedded security and identity markets while reducing reliance on imported chip platforms.

DeepInfra closes $107M Series B to expand global AI inference cloud

Stephen Mayhew — Thu, 14 May 2026 10:00:05 +0000

DeepInfra, a cloud platform for AI inference, has raised $107 million in Series B funding. The investment will be utilized to broaden the global capacity of DeepInfra, as AI demand is now transitioning from research prototyping toward production-scale inference. “When we launched nearly four years ago, we believed inference would become the dominant driver of enterprise AI workloads and we are now at this inflection point,” says Nikola Borisov, co-founder and CEO, DeepInfra. “What’s happening now is incredibly exciting – open-source models are rapidly reaching parity with proprietary systems, unlocking a new wave of innovation at a fraction of the cost and enabling widespread adoption. At the same time, agent-based systems are driving continuous, high-volume demand. Inference is no longer a thin layer – it’s the system constraint that will define the majority of workloads. Most cloud platforms weren’t built for this always-on, distributed model, so we built DeepInfra from the ground up to deliver better economics, performance, and security.” DeepInfra has agent and open-source powered AI workloads with nearly 5 trillion tokens processed every week. The platform is purposely built in an agentic era to provide vastly superior economics, performance and security for high-throughput AI workloads. DeepInfra is also an early partner of NVIDIA in the open AI ecosystem. The investment round was co-led by 500 Global and Georges Harik. DeepInfra aims to be a leading infrastructure provider for the next phase of AI development. As inference becomes the dominant compute workload, regional and distributed AI infrastructure are gaining importance, and a new generation of specialized “neocloud” providers is emerging to serve GPU-intensive AI applications more efficiently than legacy cloud architectures. DeepInfra’s focus on economics, performance, and distributed deployment aligns closely with current market trends around AI factories, inference-at-scale, and decentralized GPU infrastructure as noted in a recent report by Structure Research.

Vultr, SUSE and Supermicro target sovereign AI boom with unified cloud-to-edge infrastructure stack

Stephen Mayhew — Wed, 13 May 2026 10:00:31 +0000

Vultr, SUSE and Supermicro are partnering to create a unified cloud-to-edge architecture for AI deployments. The solution addresses the challenges of deploying AI workloads in distributed environments, focusing on latency, cost, and consistency. The architecture consists of three layers: Cloud and near-edge (Vultr), metro edge (Supermicro) and the control layer (SUSE Edge). "As AI moves into its next phase, the next challenge is data sovereignty and geographic proximity,” says Kevin Cochrane, CMO at Vultr. "By combining our global reach with regional GPU acceleration, we are helping enterprises extend their primary cloud regions directly to the edge. This partnership ensures that no matter where data is created, the sovereign infrastructure to process it is already there and ready to scale." SUSE’s technology will better enable GitOps-driven workflows for managing AI across cloud and edge. The partnership aims to make large-scale AI deployments practical by combining Kubernetes and specialized edge hardware The combination of Vultr’s globally distributed GPU cloud, Supermicro’s edge-optimized accelerated hardware, and SUSE’s Kubernetes GitOps control plane creates a practical reference architecture for enterprises deploying AI across factories, retail, telecom, healthcare, and sovereign environments. The key takeaway is that the market is maturing beyond experimentation: enterprises now want operationally unified AI infrastructure spanning centralized cloud regions, metro edge sites, and on-prem deployments, while abstracting away the complexity of managing distributed GPU infrastructure at scale. This aligns closely with broader market dynamics showing rising demand for edge inferencing, sovereign AI, and hybrid AI infrastructure ecosystems.

IREN acquires Mirantis in $625M push to become a full-stack AI cloud platform

Stephen Mayhew — Tue, 12 May 2026 10:00:43 +0000

IREN has announced its plan to acquire Mirantis in a bid to strengthen its AI cloud delivery capabilities. This acquisition aims to strengthen IREN’s deployment, operational awareness and visibility, customer support and market access. Mirantis is a cloud infrastructure and Kubernetes solutions provider with more than 1,500 enterprise customers. “IREN’s core advantage is execution from securing power to building data centers, deploying GPUs and bringing compute online at scale,” says Daniel Roberts, co-founder and co-CEO of IREN. “Mirantis builds on our existing capabilities and strengthens how that compute is deployed, managed and operated for customers.” The deal enables IREN to get AI infrastructure up and running more quickly, while servicing current customers. The consideration for the transaction will be made in IREN ordinary shares in a value of about $625 million. Mirantis will remain a standalone subsidiary. Infrastructure providers are increasingly recognizing that orchestration, Kubernetes management, deployment automation, and enterprise-grade support are becoming strategic control points in the AI value chain. By bringing Mirantis’ cloud-native and Kubernetes expertise in-house, IREN is positioning itself to evolve from a pure infrastructure provider into a vertically integrated AI cloud platform capable of accelerating customer onboarding and improving operational efficiency at scale. The move also underscores how AI infrastructure providers are racing to abstract growing infrastructure complexity for enterprises, particularly as hybrid, distributed, and inference-heavy AI deployments expand globally.

Alimak Group taps Avassa to scale industrial edge AI and containerized applications

Stephen Mayhew — Mon, 11 May 2026 10:00:05 +0000

Avassa has been chosen by Alimak Group to power its edge native computing projects. Using IoT, real-time data processing and AI applications at the edge is possible through this collaboration to make a new generation of intelligent connected systems. By using the Avassa edge platform, Alimak Group can manage and observe container workloads running on its equipment and facilities, providing secure access to all data. “Industrial companies are rapidly moving towards software-defined capabilities at the edge, and Alimak Group is truly manifesting that shift,” says Olle Hillström, CEO of Avassa. “With the Avassa edge platform, Alimak Group can accelerate its development of intelligent, connected solutions while maintaining the control, security, and scalability required in complex, distributed environments.” Alimak Group will develop a modular edge architecture for control systems, advanced analytics and AI solutions. Its aim is to increase the operational efficiency, security and provide new data-based services. The platform will allow the reliable management of applications across distributed environments. The announcement also reflects a broader market trend: manufacturing and industrial sectors are among the fastest adopters of on-premise edge computing because of requirements around data localization, operational resiliency, and low-latency decision making. This positions Alimak Group to evolve from a traditional industrial equipment company into a connected systems platform provider with recurring digital services layered onto its physical infrastructure.

Amazon buys Globalstar to extend cellular coverage and connect edge to AWS cloud

Stephen Mayhew — Mon, 11 May 2026 06:00:33 +0000

Amazon will acquire Globalstar to enhance its Amazon Leo satellite network, enabling direct-to-device (D2D) services. This acquisition will enable Amazon Leo to expand cellular coverage beyond terrestrial networks and enable connectivity in remote and edge locations. “There are billions of customers out there living, traveling, and operating in places beyond the reach of existing networks, and we started Amazon Leo to help bridge that divide,” says Panos Panay, SVP of devices & services, Amazon. “By combining Globalstar’s proven expertise and strong foundation with Amazon’s customer-obsession and innovation, customers can expect faster, more reliable service in more places—keeping them connected to the people and things that matter most. We’re excited to support Apple users through the Leo D2D system, and look forward to working with mobile network partners to help extend coverage to every corner of the planet.” In the deal Amazon picks up Globalstar’s satellite operations, infrastructure and assets, and this includes MSS spectrum licenses that come with global authorizations. Over time, this will be integrated with the AWS cloud, giving end users the ability to connect from the edge to the cloud through virtual private connections. Amazon and Apple have also finalized plans to offer satellite connectivity for the most recent iPhone and Apple Watch models. The new D2D system is expected to launch in 2028. The agreement seeks to foster innovation, increase connectivity alternatives and narrow the digital divide. The deal also allows Globalstar investors to receive cash or Amazon shares and the merger is subject to regulatory approvals and should close sometime in 2027.

SPAN pushes AI compute into homes to bypass grid bottlenecks with XFRA

Stephen Mayhew — Fri, 08 May 2026 10:00:12 +0000

SPAN announced XFRA, a distributed data center solution to address growing AI compute demand. By tapping into existing, underutilized power capacity in residential and small commercial spaces XFRA can deliver scalable compute power in a fraction of the time compared to new energy infrastructure buildouts. SPAN is working with NVIDIA to use liquid cooled GPU Server Edition RTX PRO 6000 Blackwell series GPUs. “SPAN’s unique and differentiated intellectual property in power controls enables us to improve the utilization of existing grid infrastructure,” says Arch Rao, founder and CEO of SPAN. “We have successfully deployed this capability to accelerate home electrification, unlock new home construction, and increase utility grid utilization. Now, distributed compute is the next logical extension of our technology. By building on our core strengths in power optimization and collaborating with industry leaders like NVIDIA, we are collapsing the speed-to-power gap to deliver gigawatts of cost-effective compute capacity at unprecedented speed.” The solution intends to bridge the "speed-to-power" gap for AI workloads, which are consuming electricity at an astonishing pace. XFRA offers benefits for scalers, homeowners, and utilities, creating a win-win scenario across the energy and compute ecosystem. First deployments are scheduled for sometime in 2023, with a target of being in the gigawatt range by 2027.

Nokia-Blaize alliance targets GPU limits with hybrid edge AI push in Southeast Asia

Stephen Mayhew — Thu, 07 May 2026 10:00:44 +0000

Blaize, Nokia, and Datacomm Diangraha are collaborating to deploy hybrid AI infrastructure across Indonesia and Southeast Asia. This partnership seeks to help accelerate AI adoption by combining Nokia's networks, Blaize's energy-efficient AI compute and Datacomm's regional expertise. "The combination of Nokia's validated networking infrastructure and Blaize's energy-efficient inference compute is uniquely positioned for the enterprise edge,” says Dinakar Munagala, CEO, Blaize. “GPU economics simply do not scale to thousands of distributed sites but Blaize does. This is not about replacing GPUs; it is about deploying the right compute where it delivers the most value." The partnership combines Nokia's advanced infrastructure for GPU-compute telco workloads with Blaize's platform for low-power enterprise edge AI processing. Datacomm reports increasing demand for AI inference solutions in Indonesia, highlighting the market's rapid growth. The alliance is rolling out in the public sector as well as geospatial and enterprise, with Indonesia as the target market for expansion. The hybrid approach enables the best fit compute platform (GPU or Blaize edge AI) for the right workload. Nokia and Blaize recently signed an edge AI inference MOU targeting APAC networks earlier this year.

Aranya exits stealth with GPU orchestration play as inference infrastructure shifts up the stack

Stephen Mayhew — Wed, 06 May 2026 10:00:17 +0000

Aranya, a cluster-scale operating system, emerged from stealth to address the growing demand for AI inference infrastructure. Aranya announced partnerships, including one with Hydra Host, deploying its ClusterdOS across 1,700+ GPUs. ClusterdOS enables Kubernetes to become self-healing by providing a straightforward way to manage AI inference infrastructure. The technology helps eliminate the need for dedicated platform teams, cutting setup time for partners. "At Aranya, we believe inference is the new mining. Just as crypto defined the last era of GPU-scale compute, inference is the core value-extracting workload of the AI era. The infrastructure demands are just as unforgiving: the clusters have to run, and they have to run at scale. For bleeding-edge companies that simply cannot afford downtime, we've built Aranya around that reality, with technical depth from GPU orchestration all the way up the stack," says Christian Bhatia Ondaatje, co-founder and CEO of Aranya. "Partnering with top AI inference companies like Hydra Host proves our technology inside some of the most demanding inference environments in the industry as we work towards what comes next: giving every engineering team the agency to own, operate, and expand their inference infrastructure." The company is also building Vibecluster, a team-level inference scaling management tool. Hydra Host claims it has seen a 90% reduction in cluster downtime using Aranya's technology, reflecting growing demand for software-defined abstractions that can unify fragmented AI stacks and reduce reliance on specialized platform teams. Strategically, this aligns with the rise of neocloud and GPU-centric service providers, where differentiation is shifting up the stack from raw compute to orchestration, reliability, and time-to-market. In that sense, Aranya is less a point solution and more a signal of where value is consolidating in the AI era - not just in owning GPUs, but in making them usable, scalable, and continuously available for inference at production scale.