Goldman Sachs Alternatives, Cleanhill Partners sell PCS manufacturer EPC Power for US$4.4 billion

energy-storage.news Sep 4, 2026

Goldman Sachs Alternatives, Cleanhill Partners sell PCS manufacturer EPC Power for US$4.4 billion

Curated by Shubham · TechCapsules · Sep 4, 2026

AI-summarised brief · reviewed before publication

Goldman Sachs Alternatives and Cleanhill Partners sold EPC Power Corp, a California‑based manufacturer of software‑defined inverters and power conversion systems, to Flex for US$4.4 billion. The deal, announced on 3 September, will close in Q4 2026 and is financed through a mix of debt and equity. EPC Power’s technology, including Agile Grid Forming and solid‑state transformers, targets renewable energy, battery storage, and data‑centre power markets. Flex plans to integrate EPC Power into its Cloud and Power Infrastructure unit, slated for a public spin‑out in Q1 2027.

💡 Why It Matters

· The acquisition signals a strategic shift toward high‑voltage, AI‑driven data‑centre power solutions, positioning Flex to capture growing demand for resilient, scalable energy infrastructure.

Read full article on energy-storage.news

Oracle expands HPE Juniper Networking deal to support AI infrastructure build-out

Oracle has expanded its existing collaboration with HPE Juniper Networking to support the global build-out of AI infrastructure within its data centers. The multi-year agreement involves the deployment of HPE Juniper’s PTX and MX routing platforms, alongside QFX and EX switching platforms, across Oracle’s worldwide facilities. This expansion also includes comprehensive multi-year networking support services. HPE executives stated the partnership advances a high-performance architecture designed to manage traffic, power, space, and operational demands in large-scale AI environments. Currently, HPE Juniper equipment supports key elements of Oracle Cloud Infrastructure’s data center and edge networks. As Oracle expands its AI superclusters, the new QFX switches will provide necessary high-density connectivity, dynamic load balancing, and advanced congestion management for large RoCEv2 AI backend networks, addressing increasing demands on bandwidth, latency, and fault recovery.

💡 The deal underscores the critical need for specialized networking hardware to handle the extreme bandwidth and latency requirements of modern AI superclusters. It validates HPE Juniper’s technology as a foundational component for next-generation cloud infrastructure scaling.

tekedia.com · Sep 4

How AI Data Centers Are Making Everything More Expensive

The rapid expansion of artificial intelligence data centers is driving up costs across multiple sectors, extending economic pressure beyond the technology industry. These facilities require massive amounts of electricity, water, land, and construction materials. Advanced AI models demand thousands of processors operating continuously, causing power consumption to outpace grid capacity in some regions. Utilities are investing billions in new infrastructure, with costs potentially passed to consumers through higher electricity bills. AI companies compete with households and manufacturers for limited power resources. The construction boom increases demand for concrete, steel, and copper, straining supply chains and raising material prices. Land values are rising as tech firms seek large parcels near power sources. Additionally, housing costs may increase in communities hosting these developments due to an influx of workers. Water usage for cooling systems presents further challenges, particularly in regions already facing water scarcity. This multifaceted resource demand creates significant economic ripple effects.

💡 AI infrastructure development is shifting from a tech-sector expense to a broad-based inflationary force affecting basic utilities and housing. Consumers face higher living costs as grid upgrades and material shortages are distributed across the wider economy.

taipeitimes.com · Sep 4

Data center spending to reach US$31.6tn by 2050

Global data center spending is projected to reach US$31.6 trillion by 2050, driven by surging artificial intelligence demand, according to a new PricewaterhouseCoopers report. This unprecedented investment cycle dwarfs historical infrastructure booms like railways and the internet, potentially hitting US$50 trillion if AI adoption accelerates. Tech giants and smaller providers are rapidly expanding facilities worldwide, with hardware comprising the bulk of expenditures. The United States will capture nearly half of the total spending, followed by the Asia-Pacific region and Europe. However, local opposition has already blocked or delayed at least 75 projects worth US$130 billion in early 2024 due to environmental and social concerns. Power availability remains the critical constraint shaping investment locations, as affordable, reliable, and low-carbon electricity is essential for sustaining this growth. China and India are expected to drive significant incremental demand. Annual spending is forecast to rise from US$800 billion this year to US$1.8 trillion by 2050, fundamentally resetting global capital expenditure patterns.

💡 The scale of this infrastructure shift rivals the entire US GDP, yet it faces immediate friction from local communities concerned about resource consumption. This tension between rapid technological expansion and regional resistance creates a volatile landscape where power grid capacity, rather than just capital availability, will dictate the pace of global AI deployment.

forbes.com · Sep 4

Before You Buy More Hardware, Check What Your Cluster Is Really Using

Artificial intelligence demand has severely strained the global hardware market, with Big Tech projected to spend approximately $700 billion on AI infrastructure in 2026. This surge has driven a 130% increase in combined DRAM and solid-state drive prices, creating a cost structure that enterprises must navigate for the next two to three years. Supply chain disruptions are evident, with lead times for large memory orders extending beyond 40 weeks. Some organizations have faced cancelled orders as hyperscalers prioritize their contracts, forcing others to request significantly higher budgets for the same hardware capacity. Yasmin Rajabi, Chief Operating Officer at CloudBolt Software, highlights that many platform teams mistakenly believe their clusters are maxed out. She advises companies to audit actual cluster usage before purchasing additional hardware, suggesting that optimizing existing resources may alleviate the pressure of inflated costs and scarcity rather than expanding physical infrastructure.

💡 Organizations risk severe financial inefficiency by blindly scaling hardware during a supply crisis. Auditing existing cluster utilization prevents unnecessary capital expenditure when memory prices are inflated by 130%.

prnewswire.com · Sep 3

Cango Inc.’s AI Subsidiary, EcoHash, Announces Operational Milestones: Completes Dedicated AI Infrastructure…

Cango Inc. announced that its AI subsidiary, EcoHash, has achieved significant operational milestones at its owned 50-megawatt facility in Georgia. EcoHash completed infrastructure modifications for a dedicated AI compute section, which currently supports up to 3 megawatts of capacity with room for expansion. The company deployed high-density compute containers, utilizing a modular design that standardizes power, cooling, and networking to accelerate scaling. By leveraging existing mining infrastructure, EcoHash minimized retrofit work and rapidly activated an initial batch of GPU servers. Consequently, EcoHash has commenced commercial GPU compute services using these servers. Initial clients include GPU cloud platforms and AI-native infrastructure providers. The company is also engaging prospective customers in decentralized compute networks and advanced AI research laboratories. Simon Tang, Cango’s CFO, highlighted this launch as a demonstration of the commercial viability of their plug-and-play container solution, marking a key step in transitioning from infrastructure preparation to active revenue-generating AI operations.

💡 By repurposing existing mining infrastructure for AI compute, Cango bypasses the lengthy construction timelines typical of new data centers, offering a faster path to market for high-density GPU services. This modular approach allows the company to rapidly scale capacity while minimizing capital expenditure on site retrofits.

channellife.com.au · Sep 3

Synology launches first Australia symposium on AI data

Synology has launched its inaugural Insight Symposium in Australia, targeting IT professionals and senior executives to address enterprise data management within the context of artificial intelligence. The event, held on September 3, 2026, aims to bridge the gap between AI adoption enthusiasm and the robust data environments required to support it. Keynote speakers include Joanne Weng, Director of International Business, and Jacqueline Graciella, Country Manager for the ANZ region. Other notable participants are Nate Cochrane, Editor-in-Chief at Pixntx, and Johan Sulaiman, Chief Technology Officer at Heart Research Institute. The agenda focuses on three critical challenges: preparing fragmented data for AI use, determining optimal data and workload locations, and enhancing resilience against cyber threats. This symposium reflects a broader market shift from trialing AI products to scrutinizing the underlying infrastructure that stores operational and customer information. Synology positions storage, protection, and data management as central to boardroom discussions on AI strategy, emphasizing practical solutions over broad technological predictions. The forum seeks to facilitate open dialogue among industry leaders to exchange strategies for building stronger data foundations, addressing the growing disparity between rapid AI adoption and the necessary data governance frameworks in Australian businesses.

💡 The event underscores a critical pivot in enterprise technology strategy, where infrastructure readiness now dictates the viability of AI initiatives. By prioritizing data governance and security, Synology challenges organizations to address foundational weaknesses before scaling artificial intelligence deployments.