AI

PwC has published an analysis projecting that $31.6 trillion in capital expenditure will flow through AI-era infrastructure investment globally. The report maps where that spending concentrates, covering data centers, semiconductors, energy systems, and network infrastructure. PwC frames this as a structural economic shift rather than a cyclical investment cycle.

The National Institute of Standards and Technology has released guidance on securing AI data center architecture, covering security posture assessment and emerging standards applicable to facilities running large-scale AI workloads. The publication addresses how the physical and logical structure of AI data centers differs from conventional facilities and what those differences mean for security controls. NIST guidance typically informs federal procurement requirements and is widely adopted as a baseline by private sector operators.

Equinix, Together AI, and Nvidia have announced a joint partnership to launch a product called Inference Exchange, which is designed to give enterprise customers streamlined access to AI inference capacity across Equinix's global data center footprint. The arrangement combines Equinix's colocation infrastructure, Together AI's model serving software, and Nvidia's GPU hardware. Financial terms were not disclosed.

ChatGPT, Claude, and Grok experienced simultaneous service outages, according to Data Center Dynamics, in an event that briefly disrupted access to three of the most widely used AI platforms. The coincidence of outages across separate operators raises questions about shared infrastructure dependencies or common upstream failures. No single cause has been publicly confirmed by all three companies.

Data Center Knowledge reports that self-improving AI systems, which retrain or refine themselves autonomously, could drive compute demand well beyond current forecasts and strain data center capacity in ways that static workload models do not capture. The analysis highlights that recursive improvement cycles generate unpredictable bursts of training demand rather than the steady growth operators plan for. Capacity planners are warned that traditional utilization curves may become unreliable.

Data Center Knowledge identifies fiber network capacity as a potential constraint on data center expansion, arguing that the industry's focus on power and land overlooks limits in high-density fiber routing to and within large campuses. The analysis points to long lead times for new fiber builds and permitting hurdles for buried conduit as compounding factors. Operators racing to deliver AI capacity may find connectivity, not just power, delaying commissioning timelines.

Amazon is planning a new subsea cable system linking the United States and Japan, Data Center Dynamics reports. The cable would expand Amazon Web Services' transpacific network capacity, supporting data center operations and cloud traffic on both ends of the route. Subsea cable investments of this scale typically involve hundreds of millions of dollars in capital expenditure. The project would add redundancy and throughput to one of the world's highest-traffic data corridors as AI workloads increase demand for low-latency international connectivity.

AI inference demand is reshaping data centre infrastructure requirements across networking, storage, power delivery, and cooling, according to ET Datacenters, as inference workloads differ substantially from training in their latency sensitivity, traffic patterns, and hardware configurations. Operators are finding that infrastructure optimized for batch training runs cannot efficiently handle real-time inference at scale, driving significant retrofitting and new-build specification changes. The shift is accelerating procurement cycles for specialized accelerators and high-bandwidth networking equipment.

Data center and power company Iren has secured a new contract with what it described as a leading frontier AI lab, confirming it is on track to fully exit the cryptocurrency mining business by the end of 2026. The company did not name the client, but the deal marks a strategic pivot from crypto hosting toward AI compute workloads. Iren's transition reflects broader industry pressure on crypto-adjacent operators to reposition their infrastructure for AI revenue.

IREN, formerly Iris Energy, has made its first delivery of AI cloud computing capacity to Microsoft under a contract valued at up to $9.7 billion. The milestone represents the initial phase of one of the largest AI infrastructure supply agreements between a specialist compute provider and a hyperscaler. Further capacity tranches are expected to follow as IREN expands its GPU data center footprint.

Amazon Web Services is wiring a new transpacific subsea cable route connecting the United States to Asia, specifically designed to support AI workload traffic. The cable is intended to reduce latency and increase bandwidth for AI inference and training jobs routed between US data centers and Asian markets. AWS has not disclosed the full construction cost or the cable's exact landing points.

Dell'Oro Group projects global data center capital expenditure will exceed $3 trillion cumulatively by 2030, driven by sustained AI infrastructure buildout. The research firm says AI workloads are the primary engine behind the spending wave, with hyperscalers and cloud providers committing to multi-year capacity expansions. Dell'Oro tracks hardware, power, and facility spending together to arrive at the aggregate figure.

JLL's Global Data Center Outlook 2026 report projects global data center capacity will nearly double to approximately 200 gigawatts as AI infrastructure demand accelerates. The report covers supply and demand dynamics across major markets, noting that new construction pipelines remain strained despite record-level development activity. JLL identifies power availability as the single largest constraint on further capacity growth.

The Trump administration is actively promoting AI data center buildouts across the United States despite organized opposition from local communities in affected areas. The administration has positioned federal support for AI infrastructure as a national priority, creating tension with municipal and neighborhood-level resistance to new facilities. The dynamic reflects a broader conflict between federal economic and technology policy goals and local land use authority.

South Korea is assembling a sovereign AI investment program that SemiAnalysis estimates at roughly one trillion dollars, with Nvidia positioned as the primary hardware beneficiary. SK Hynix, despite being the leading supplier of high-bandwidth memory used in Nvidia GPUs, is expected to see limited direct financial uplift from the program's structure. The analysis points to procurement frameworks and partnership terms that favor US chip suppliers over domestic memory manufacturers.

Data Center Frontier's DCF Trends Summit examined how rapid AI hardware iteration is shortening the useful life of data center equipment, putting pressure on IT asset disposition practices and creating new challenges for resale, recycling, and secure destruction. Speakers noted that GPU generations are cycling faster than traditional server refresh timelines, making ITAD planning more complex. The compressed lifecycle also affects depreciation accounting and sustainability commitments tied to equipment reuse.

Amazon Web Services announced it will launch its first cloud infrastructure region in the Kingdom of Saudi Arabia by December 2026. The region will give Saudi customers local access to AWS compute, storage, and database services, supporting the country's Vision 2030 digital transformation goals. The move places AWS in direct competition with other hyperscalers expanding into the Gulf.

OpenAI has purchased Macs by the tens of thousands to support AI workloads, a procurement scale that analysts say repositions Apple as a meaningful AI infrastructure vendor. The buying program reflects OpenAI's search for diverse compute capacity beyond GPU clusters, with Apple Silicon offering high memory bandwidth suited to certain inference tasks. The move draws Apple into the AI infrastructure supply chain in a way its consumer-focused narrative has not previously emphasized.

QumulusAI is expanding its GPU cluster deployments but says available powered capacity, not GPU supply, is the primary constraint on its growth, according to Data Center Knowledge. The company is actively sourcing powered shell space across multiple US markets as it competes with larger cloud providers for scarce megawatts. The bottleneck reflects a broader industry pattern where compute procurement now depends on power availability rather than hardware lead times.

The United States has mobilized three major technology companies to compete in South Africa, the continent's second-richest economy, against a Huawei proposal to build a 2,008-chip AI data center. The move is framed as a direct geopolitical counter to Chinese infrastructure investment in Africa. South Africa represents a strategically significant market for AI compute deployment on the continent.

Nvidia has paused certain cloud revenue-sharing arrangements with some partners, according to a report from Data Center Dynamics. The move could affect the economics of partnerships where Nvidia shares in the revenues generated by cloud providers running its GPU hardware. No specific partners or dollar figures were disclosed in the report. The pause raises questions about how Nvidia is restructuring its commercial relationships as it forecasts record growth.

Nvidia reported that its revenue doubled compared to the same period last year, driven by sustained demand for its data center GPU products. The company also issued a forecast for record growth in the coming quarters, reflecting continued AI infrastructure investment by hyperscalers and enterprise customers. Specific quarterly figures were reported by Data Center Dynamics. The results cement Nvidia's position as the primary revenue beneficiary of the current AI compute buildout cycle.

Anthropic has signed a $45 billion compute capacity agreement with Nscale, according to a report cited by Data Center Dynamics. The deal represents one of the largest single compute procurement contracts disclosed in the AI industry. Terms of the agreement and the timeline for capacity delivery were not fully detailed in the report.

Nebius has secured $5.75 billion through convertible notes to fund AI data center construction and GPU fleet expansion. The raise is one of the largest single financing rounds for an AI infrastructure company in 2026, reflecting continued investor appetite for dedicated GPU compute capacity outside the major hyperscalers. Nebius operates GPU cloud services and plans to deploy the capital across multiple markets.

Hyundai is building a proprietary AI campus rated at 100MW and equipped with 50,000 GPUs, according to reporting by Tech Times. The automaker is constructing the facility to run AI workloads in-house rather than renting capacity from public cloud providers. The move positions Hyundai alongside a small number of non-tech industrial companies that have elected to own dedicated AI compute infrastructure.

Chris Malone, OpenAI's head of data centers, has left the company, according to Data Center Dynamics. Malone's departure is the latest in a series of senior infrastructure exits at OpenAI as the organization accelerates its buildout of AI training and inference capacity under the Stargate program. No successor has been named publicly. The loss of a senior data center executive during a period of rapid infrastructure scaling raises questions about execution continuity on OpenAI's largest facility commitments.

OpenAI has detailed its Jalapeño custom AI chip, disclosing a 700W thermal design power figure that places it among the most power-intensive processors yet announced for AI workloads. The specification signals a significant infrastructure challenge for any facility deploying the chip at scale, requiring purpose-built cooling and power distribution systems. Data Center Dynamics reported the details, which follow earlier reports that Jalapeño outperforms Nvidia Blackwell on certain inference tasks.

SemiAnalysis published a detailed total cost of ownership and architecture analysis comparing Nvidia's Vera Rubin NVL72 against the GB200 NVL72 for inference workloads. The analysis examines per-token economics, memory bandwidth, and interconnect architecture across both platforms, providing operators and hyperscalers with a framework for capital allocation decisions. The comparison arrives as data centers face pressure to justify rack-level power expenditures exceeding hundreds of kilowatts per unit.

SemiAnalysis published an analysis concluding that OpenAI's internally developed chip, codenamed Jalapeño, surpasses Nvidia's Blackwell architecture in key performance metrics. The report details how Jalapeño is designed specifically for large-scale AI training workloads, potentially reducing OpenAI's dependence on third-party silicon suppliers. If accurate, the development signals a significant shift in the custom chip race among hyperscalers and AI labs.

CNBC has reported that small-scale data centers designed for residential deployment may become a feature of American homes in coming years, driven by demand for edge AI compute closer to end users. The concept involves compact, consumer-facing infrastructure capable of running localized AI workloads without relying on centralized cloud facilities. No specific product launch date or named company was identified as the primary developer in the snippet. The idea represents a potential structural shift in how AI compute is distributed across the grid.

Keel has decommissioned all of its US-based cryptocurrency mining facilities as the company pivots entirely to AI compute infrastructure. The company did not specify how many sites were closed or their combined capacity, but the action represents a complete exit from the cryptomining business in favor of AI workloads. Keel joins a growing list of former mining operators repurposing or retiring assets to capture AI demand.

Broadcom executives have said hyperscalers are increasingly sidelining telecommunications companies for certain large-scale AI infrastructure projects, opting to build or source network and compute capacity directly rather than routing through telco partnerships. The shift reflects hyperscalers' drive to control more of the stack as AI workloads grow. Broadcom, which supplies custom AI chips and networking silicon to several major hyperscalers, stands to benefit as direct buildouts accelerate.

Alibaba Cloud has announced it will increasingly rely on internally developed chips for its data center infrastructure going forward, reducing dependence on third-party suppliers. The move aligns with a broader trend among major hyperscalers to build proprietary silicon tailored to their specific AI workloads and infrastructure architectures. Alibaba has been developing custom accelerators and networking chips, and this announcement signals those efforts will now take a central role in its capacity planning.

Network infrastructure provider Zayo has secured a long-term fiber capacity agreement with Corning to support AI-driven network buildout. The deal gives Zayo priority access to Corning's fiber supply as demand for high-bandwidth connectivity between AI data centers and compute clusters intensifies. Specific contract terms and capacity volumes were not disclosed.

Google's deepening bet on custom silicon from Marvell is prompting data center planners to revisit capital expenditure assumptions built around standard GPU procurement cycles. The shift toward application-specific integrated circuits changes rack density, power draw per chip, and cooling requirements, all of which feed into facility design budgets. The arrangement reflects a broader hyperscaler trend of moving away from merchant silicon to control performance and cost at scale.

Quantitative trading firm Hudson River Trading has contracted CoreWeave to provide GPU compute for its research platform, the companies confirmed. The deal represents a significant customer win for CoreWeave as it expands its client base beyond AI-native companies into the financial services sector. Terms of the agreement were not disclosed, but the partnership positions CoreWeave as infrastructure for latency-sensitive, high-performance workloads outside traditional AI model training.

Bitdeer AI has deployed an Nvidia GB300 NVL72 cluster at a data center in Malaysia, according to Data Center Dynamics. The GB300 NVL72 is Nvidia's current-generation AI training system, built around Blackwell Ultra GPUs configured in a 72-GPU rack-scale unit. The deployment represents one of the first confirmed GB300 NVL72 installations in Southeast Asia.

Amazon Web Services has released new AI data center designs that support a sixfold increase in rack density compared to previous generations. The designs represent a significant architectural shift aimed at accommodating the power and cooling demands of next-generation AI training and inference hardware. AWS did not disclose specific deployment timelines or capital figures tied to the new configurations.

Data Center Knowledge examines the growing challenge of adapting existing data center facilities to meet the power density, cooling, and structural requirements imposed by AI workloads. Operators are weighing the costs of full retrofits against new builds, with neither option straightforward given supply chain delays and permitting backlogs. The retrofit market is gaining traction as a near-term solution for operators who cannot wait for greenfield capacity to come online.

Accelerating AI infrastructure buildouts are pushing global construction capacity to its limits, according to Data Center Knowledge. Shortages of skilled labor, structural steel, specialized electrical equipment, and long-lead generators are extending project timelines across North America, Europe, and Asia. The strain is intensifying as multiple hyperscalers simultaneously pursue gigawatt-scale campuses.