When Qualcomm’s Snapdragon chip first debuted (even before the X1 Elite), some applications written for X86 chips from AMD and Intel wouldn’t run on Arm. Some of that was alleviated by Microsoft’s Prism emulator in Windows, which translated X86 instructions to Arm native code. But just like a conversation with someone who doesn’t speak the same language as you, translation slowed the process down. I was surprised to find that the RTX Spark family of laptops may have the same experience, though this time from a GPU perspective. Adobe Premiere needed a special beta to run. Without it, Premiere complained of an unsupported GPU. ComfyUI specifically supported the RTX Spark. I don’t think that games will have any issues with the RTX Spark, but I’m not entirely certain we’ve put this issue to bed, either. pcworld.com/article/3262052/…

The reported theft and publication of AfD email data shows how neglected communications infrastructure can become a concentrated source of political, financial and personal information. AfD has confirmed that a hosting provider suffered a compromise in September, although key questions about the intrusion, the perpetrators and the complete authenticity of the released material remain unresolved. According to Euronews, the archive reportedly contains about 421,000 emails associated with roughly 10 mailboxes, including those of senior party leaders. The scale alone is substantial, but the attachments and associated records present immediate risks of identity fraud and targeted phishing. Heise reported that the collection also included more than 41,000 PDF files, 3,480 DOCX documents and SQL database dumps. The material reportedly spans bank statements, identity documents, invoices, academic certificates, membership applications and regional member lists. That mix changes the character of the incident. This is not simply a cache of internal political correspondence. Identity documents and bank records can support impersonation or financial fraud, while membership information could expose individuals to harassment, pressure or targeted social engineering. Internal messages can also give attackers the context needed to make fraudulent requests sound convincing. Email archives preserve relationships as well as documents. They reveal who communicates with whom, how approvals work, which vendors are trusted and what language senior personnel use, providing the raw material for tailored phishing campaigns. insight.tmcnet.com/insight/a…

Shanghai Lin-gang UDC is testing a different answer to one of computing's most stubborn problems: how to remove heat without consuming large amounts of electricity and freshwater. Instead of relying on conventional cooling infrastructure, the project places sealed computing equipment beneath the sea and connects it directly to nearby offshore wind generation. The facility is designed for 24 MW of capacity. Its initial deployment comprises 192 server racks operating at 2.3 MW, according to New Atlas. Reported electricity savings are 22.8% compared with comparable land-based facilities, while more than 95% of the project's electricity reportedly comes directly from an adjacent offshore wind farm. That combination matters. Data centres increasingly compete for grid capacity, water and suitable development sites, particularly as artificial intelligence workloads raise rack densities. Locating computing infrastructure near offshore generation could reduce dependence on long-distance grid transmission while shifting some physical infrastructure away from congested urban areas. Lower cooling demand does not automatically make an underwater data centre environmentally benign. Shanghai Lin-gang UDC reportedly achieves a Power Usage Effectiveness rating below 1.15, compared with roughly 1.5 for conventional facilities, according to Newsday. PUE measures total facility energy against the energy used by computing equipment. A result closer to 1 indicates that less power is being consumed by cooling, lighting and other supporting systems. insight.tmcnet.com/insight/s…

CBRE is expanding its data center infrastructure business through an acquisition that strengthens its ability to deliver the physical guts of these facilities. The move places the commercial real estate services company closer to the engineering, power, and operational decisions shaping one of property's fastest-growing sectors. The strategic logic is straightforward. Data centers still involve land, buildings, and leases, but their economics depend heavily on electrical capacity, cooling systems, network connectivity, and construction expertise. By extending further into infrastructure delivery, the firm can participate in more of a facility's lifecycle rather than focusing primarily on transactions, valuation, or conventional property management. That matters because available space is no longer the only constraint. In many markets, the harder question is whether a project can obtain enough power, and when. A parcel may appear attractive on paper yet remain commercially impractical if grid interconnection takes years or if local infrastructure cannot support the planned computing load. The market numbers explain the strategic pivot. Supply across North American primary markets increased 36% year over year to 9,432 MW in 2025. Net absorption reached a record 2,497.6 MW, while vacancy dropped to a historic low of 1.4%. Average lease rates rose 6.5% to $194.95 per kW per month. Those figures point to a market growing quickly without creating much breathing room. Artificial intelligence workloads are adding demand, but so are cloud services, enterprise computing, and the continuing digitization of business processes. AI also changes facility design because dense accelerator clusters require substantially different power and cooling configurations from traditional enterprise servers. insight.tmcnet.com/insight/c…

Elon Musk’s Terafab is rapidly expanding its search for semiconductor suppliers, with Tesla and SpaceX approaching Korean manufacturers across equipment, materials, facility components and chip-testing products. The outreach goes well beyond HPSP and Hanmi Semiconductor, the two Korean suppliers previously associated with the project. According to TheElec, Terafab personnel visited around 10 Korean booths during the opening morning of SEMICON West in San Francisco last week. The companies approached included ENF Technology, Dongjin Semichem, Hanmi Semiconductor, Unisem and MiCo. That kind of concentrated activity suggests procurement is moving from broad planning into supplier qualification. Several companies have reportedly signed nondisclosure agreements. One received an NDA by email on the evening of Terafab’s booth visit, an unusually compressed timeline for a semiconductor project. Hanmi Semiconductor is discussing TC bonders for Terafab, while HPSP has agreed to provide one high pressure hydrogen annealing tool for the research and development line. MiCo’s cleaning unit Komico signed a three-way NDA with Tesla and SpaceX about a month ago. ISC, meanwhile, is exchanging samples for rubber test sockets. The list spans several layers of semiconductor manufacturing. TC bonders support advanced packaging processes, annealing systems modify material properties under controlled conditions, and test sockets connect packaged chips to testing equipment. Process chemicals, piping and cleaning systems are less visible, but fabs cannot operate without them. A giant factory is still dependent on thousands of highly specific parts. insight.tmcnet.com/insight/t…

AI demand continues to expand, semiconductor suppliers are posting record profits, and major technology companies are committing hundreds of billions of dollars to new infrastructure. Yet shares of NVIDIA and other AI-linked businesses can still decline. The apparent contradiction reflects a shift in investor priorities: markets are focusing less on whether AI adoption will happen and more on how much it will cost, when it will produce revenue, and which businesses will retain the economics. The demand indicators remain formidable. Gartner forecast worldwide generative-AI spending at $643.9 billion in 2025, up 76.4% year over year (source). That expansion supports continued investment in accelerators, servers, networking equipment, and data centers. It also highlights a complication. Much of the spending surge is concentrated in infrastructure, where upfront costs arrive well before many end-user applications generate material returns. The broader market is growing as well. IDC estimated that AI solutions represented a $307 billion market in 2025 and projected spending would reach $632 billion by 2028, implying a compound annual growth rate of roughly 27% (source). Those figures point to substantial commercial adoption. They do not, by themselves, show how revenue and profit will be distributed among chip designers, cloud providers, software developers, and enterprise customers. Demand can be real while a stock is still expensive. NVIDIA can continue receiving exceptional data-center orders, for example, while investors lower the valuation they are willing to assign to future earnings. Expectations, competition, and potential margin pressure all influence that calculation. A strong quarter may therefore confirm the AI buildout without resolving concerns about what happens when supply increases or customers become more selective. Capital expenditure is now at the center of that debate. J.P. Morgan Asset Management said five major U.S. technology companies were expected to spend $697 billion on capital expenditure in 2026, $173 billion more than estimated at the start of the year (source). It also estimated that AI-related capital expenditure would rise from 33% of hyperscalers' operating cash flow in 2023 to 93% in 2026. insight.tmcnet.com/insight/i…

Stack Infrastructure is moving forward with plans for a massive data center campus at the Berry Hill megasite in Pittsylvania County, Virginia. The proposed development represents up to $100 billion in investment over roughly 30 years, placing it among the most ambitious digital infrastructure projects contemplated in the United States. The scale of the project is immense. According to Virginia Business, Stack Infrastructure’s plan covers approximately 2,990 acres and could produce 21 data center buildings. Two on-site natural gas facilities would support the campus, helping address the challenge of securing enough reliable power for high-density computing workloads. At full build-out, the campus could require approximately 3 gigawatts. That is not simply a large corporate energy purchase; it is the kind of sustained load that directly influences regional transmission planning, generation investment, and utility rate discussions. The $100 billion headline figure represents a long-term goal rather than an immediate capital commitment. Stack Infrastructure’s binding performance agreement establishes a minimum investment of $73.5 billion, with at least 2,050 permanent jobs created within 20 years. Those positions are expected to carry an average annual wage of $80,500. Longer-range plans refer to as many as 2,500 jobs. Even those employment numbers need context. Data centers can generate considerable property value without employing as many people as manufacturing campuses of comparable cost. For Pittsylvania County, however, the fiscal impact will be substantial. The completed project could produce up to $48 million annually in property-tax revenue at full build-out, giving local officials potential funding for schools, public safety, transportation, and other services. insight.tmcnet.com/insight/s…

Two planned facilities totaling roughly 400 MW are moving Aligned Data Centers’ Chandler campus into another phase of development. The expansion comes during an eventful period for the company, which was acquired for $40 billion this summer as demand for large-scale computing infrastructure continued to climb. Google and Oracle are reportedly associated with the development, according to the Phoenix Business Journal, although the specific nature of their Chandler arrangements has not been independently confirmed. That distinction matters. Data-center projects can involve dedicated hyperscale buildings, wholesale colocation agreements, reserved capacity or phased commitments that change as power becomes available. Still, the reported involvement of Google and Oracle points to the type of demand reshaping metropolitan Phoenix. Both operate at a scale where electricity, network connectivity and expansion options can outweigh proximity to traditional technology centers. Aligned Data Centers occupies a different but complementary position, developing and operating infrastructure that can accommodate large deployments without requiring every customer to build an entire campus independently. The proposed 400 MW is substantial. For perspective, capacity at that level resembles a utility-scale industrial development rather than a conventional commercial real estate project. The headline number also does not mean every megawatt will come online at once. Large campuses are often delivered in stages as buildings, substations, transmission connections and customer deployments are completed. Available land is only one part of the equation. Power procurement has become the central constraint for many North American data-center projects. Permitting schedules, transmission upgrades, equipment lead times and local utility planning can determine when a technically viable campus begins serving workloads. insight.tmcnet.com/insight/a…

Pangea Connected has launched Pangea Amplify, adding a commercial funding initiative to its Partner Programme as channel businesses pursue a growing pipeline of cellular IoT projects. The channel-only cellular IoT connectivity specialist is positioning Pangea Amplify as a way for partners to reduce the initial financial burden associated with winning and deploying new business. That burden can include hardware procurement, connectivity activation, installation, device management, and the early delivery costs of a recurring managed service. It addresses an awkward feature of the IoT business model. Customers may pay through multiyear service contracts, but partners often incur substantial costs before those recurring payments begin. Even a commercially attractive deployment can therefore put pressure on working capital, particularly when hundreds or thousands of connected devices need to be purchased, configured, and installed. Pangea Amplify gives partners another route for structuring those opportunities. The broader aim is to help partners package cellular connectivity alongside devices and deployment services, rather than selling SIMs or data plans as isolated products. Specific eligibility requirements and commercial terms will be important for partners evaluating the initiative. Cellular IoT connectivity is expanding, but connectivity alone usually represents only one part of the customer requirement. Businesses also need suitable hardware, coverage planning, configuration, installation, monitoring, security, support, and sometimes integration with existing operational systems. Channel partners that can bring these components together have more room to develop recurring revenue, but they also carry more delivery risk. insight.tmcnet.com/insight/p…

Canadians who lose money through impersonation scams and manipulated payments are intensifying pressure on Ottawa to make banks shoulder more of the financial burden. The push targets a difficult category of fraud: transactions technically authorized by customers who were deceived into sending the money. That distinction shapes how banks investigate claims. When criminals take over an account or process an unauthorized payment, existing protections may offer a clearer path to reimbursement. Cases involving social engineering are messier because the customer completed the transfer, disclosed information, or followed instructions from someone posing as a bank employee, government official, relative, or business contact. Victims argue that this approach puts too much responsibility on individuals while underestimating what banks can see across their payment systems. Financial institutions monitor transaction size, destination, timing, device information, account history, and other risk signals. Customers generally see only the story presented by the scammer. And the losses are climbing. Canadians reported more than $704 million in fraud losses in 2025, according to Canadian Anti-Fraud Centre figures cited by Finance Canada. Reported losses since 2022 exceeded $2.4 billion, although authorities estimate that only 5% to 10% of fraud is reported. The payment channels involved are central to the policy debate. Government figures show that reported 2024 losses included $154 million sent through wire transfers and $36.7 million through Interac e-Transfers. These systems move money quickly, which is useful for legitimate commerce but leaves a narrow window for stopping or recovering fraudulent transfers. insight.tmcnet.com/insight/c…

Insurance buyers should evaluate AI compliance services by determining whether a provider can connect each automated insurance decision to its model, data, approval, security controls, and regulatory evidence. For example, an underwriting model changes a risk score, but the compliance team cannot reconstruct which data version, model release, or approval record produced the decision. That is the practical problem driving demand for insurance compliance services. The work extends beyond writing policies: buyers need controls connecting policy administration, claims, pricing, customer communications, and model governance. The exposure is expanding. Fenwick’s review of the NAIC’s published AI and machine-learning survey results reported adoption or planned adoption reached 92% among surveyed health insurers, 88% among auto insurers, 70% among home insurers, and 58% among life insurers. The surveys were released between 2022 and 2024 and cover different insurance segments rather than one directly comparable industrywide sample. The health-insurance findings also showed that nearly one-third of surveyed insurers did not regularly test models for bias or discrimination. A useful scope starts with specific decisions. Buyers can catalog models used for prior authorization, utilization management, claims triage, fraud detection, underwriting, and pricing. Each inventory record should identify the model owner, training-data source, production version, validation date, affected product, and applicable jurisdiction. That detail matters. A broad objective such as "improve AI compliance" gives providers little to design against. A testable requirement, such as linking every adverse underwriting recommendation to its model version and data lineage (the recorded origin and movement of data) creates a clear technical and regulatory target. insight.tmcnet.com/insight/h…

A buyer-centered professional-services website helps prospects identify the right service, assess the firm’s evidence and fit, understand the engagement, and take an appropriate next step without first speaking to sales. A professional-services website can appear polished while still making purchasing difficult. A prospective client may understand that a firm offers managed IT services, cybersecurity, and Voice over Internet Protocol (VoIP), yet remain unable to determine which service applies, what the engagement includes, or how the technologies work together. That gap matters because vendor research increasingly occurs before a buyer contacts sales. Gartner reported that 70% of B2B buyers preferred a completely digital, self-service buying experience, 67% preferred an experience without a sales representative, and 45% used generative AI during vendor research. The firm also found that buyers used an average of seven information sources during a purchase. The website consequently needs more than a broad "services" page. A managed IT page might identify supported environments such as Microsoft 365, Windows Server, Azure, VMware, and common endpoint platforms. A cybersecurity page should distinguish managed detection and response (MDR), which combines continuous threat monitoring with expert investigation, from vulnerability scanning, email security, and incident response. A VoIP page can explain Session Initiation Protocol (SIP) trunking, number porting, Enhanced 911 (E911), call queues, and Microsoft Teams integration. Consistency is equally important. Service names, technical claims, and qualification criteria should match across the website, downloadable PDFs, search snippets, social profiles, and sales collateral. insight.tmcnet.com/insight/h…

Amazon has made big bets before. It spent heavily on cloud computing, fulfillment centers, and its delivery network long before they became some of the company's most valuable assets. Investors questioned those investments at the time, but Amazon ultimately built enormous, profitable businesses from them. AI is testing this playbook at a far larger scale. Amazon is committing hundreds of billions of dollars to data centers, servers, and power before investors know how quickly those investments will generate revenue. businessinsider.com/amazon-2…

Town has joined the ranks of privately held startups valued at $1 billion or more after announcing a $108 million Series B funding round on Monday. Forerunner Ventures and Menlo Ventures co-led the investment, giving the AI assistant startup substantial new resources at a moment when investors are racing to identify durable winners in enterprise automation. The valuation reflects how quickly expectations have moved ahead of proven financial returns across the broader AI agent market. Enterprise demand is growing. Spending on enterprise generative AI reached an estimated $37 billion in 2025, up from $11.5 billion in 2024, according to Menlo Ventures. That expansion helps explain why companies such as Town attract large funding rounds while the category is still taking shape. Adoption comes in several forms. Using a chatbot in one department is different from allowing an autonomous agent to update customer records, approve a transaction, contact a supplier, or alter an operational workflow. The latter carries more potential value, but it also introduces more risk. A 2026 market overview from ToolGlance describes an environment in which enterprise AI adoption continues to expand while organizations work through deployment, integration, and operational questions. Separately, BERI has examined Gartner and IDC data concerning why enterprise agent initiatives stall. Those implementation barriers matter because Town is entering a field crowded with well-capitalized competitors and incumbent platforms. insight.tmcnet.com/insight/t…

Speed and price are becoming central battlegrounds in enterprise artificial intelligence, particularly for agents that make repeated model calls while completing a task. MiniMax entered that contest on Monday with M3.1 Flash, an updated AI model designed for agentic workflows, coding, computer use, and cybersecurity. The Chinese AI startup is pitching M3.1 Flash as a substantial step beyond its predecessor. According to Reuters, MiniMax claims the model offers nearly three times faster online inference while reducing usage costs by 70%. Those figures could attract businesses experimenting with agents but struggling to justify the cost of running them at scale. An ordinary chatbot might answer after one or two model calls. An agent may plan a task, retrieve information, invoke software tools, inspect the result, revise its approach, and repeat the process. Each step consumes tokens and adds latency. Small differences in model pricing and response time can therefore become much larger differences when multiplied across thousands of workflows. A model that responds quickly enough to support interactive work, while costing less per workflow, can sometimes be more useful than a stronger model that is slower or more expensive to operate. MiniMax is reporting a peak output rate above 200 tokens per second and a median rate above 170 tokens per second. M3.1 Flash also offers a one-million-token context window, giving developers room to supply large code repositories, extensive documentation, lengthy interaction histories, or several business records within a single session. insight.tmcnet.com/insight/m…

The September 2026 CXO Monthly Roundup places artificial intelligence at both ends of the security equation. Attackers are using AI to accelerate parts of their operations, while defenders are introducing agentic systems intended to investigate alerts and coordinate responses with less manual intervention. That tension runs through the roundup's coverage of AI-assisted attacks in Asia, an Agentic SecOps launch, the 2026 Ransomware Report, and the Frontier AI Readiness Report. For business and technology leaders, these are not isolated developments. Together, they point to a security environment in which speed, identity, access, and automation increasingly overlap. AI-assisted attacks do not need to introduce entirely new tactics to create problems. They can make familiar techniques more efficient. More convincing phishing messages, faster reconnaissance, multilingual social engineering, and automated vulnerability analysis can increase the volume of credible activity reaching employees and security teams. The regional focus on Asia also matters. Enterprises operating across multiple countries often manage different regulatory obligations, cloud environments, languages, suppliers, and levels of security maturity. That creates uneven visibility. An attack detected quickly in one business unit may remain difficult to investigate in another because identity records, endpoint telemetry, and network data are held in separate systems. Agentic SecOps is emerging as one response. Unlike conventional automation based mainly on fixed playbooks, agentic security tools are designed to interpret context, select steps, and carry out portions of an investigation. In practice, that could mean gathering evidence from several systems, assessing whether an identity appears compromised, recommending containment, or initiating a narrowly authorized response. insight.tmcnet.com/insight/c…

Barton Associates’ acquisition of LocumsCollective reflects a broader change in how healthcare staffing deals are evaluated. LocumsCollective, a five-year-old managed service provider, or MSP, was sold to Barton Associates in May. The transaction places an MSP model inside a larger healthcare staffing business at a time when hospitals and regional health systems remain under pressure to fill clinical roles and manage labor costs. An MSP typically coordinates staffing suppliers, job orders, compliance processes, reporting, and workforce planning for healthcare clients. In practice, the model can give staffing organizations a clearer view of demand across facilities, specialties, and shifts. LocumsCollective therefore brings Barton Associates more than a set of customer relationships. It potentially adds workflow knowledge, recruiter expertise, credentialing processes, and data about how contingent clinical labor is sourced and deployed. Workforce availability increasingly shapes the value of healthcare assets. The Oklahoma Hospital Association found that vacancy and turnover remained elevated across nursing, behavioral health, radiology, respiratory therapy, and allied-health positions during 2025, even as some measures improved. Scarcity extends well beyond registered nurses, creating operational gaps that can affect scheduling, service capacity, and expansion plans. Reporting based on West Virginia Hospital Association data identified more than 4,500 vacancies across 13 critical occupations at West Virginia hospitals. Nursing represented approximately 3,600 of those openings. For organizations serving rural markets, acquiring recruiting reach and workforce-management capabilities can be faster than building them internally, although integration introduces its own risks. A staffing acquisition can lose value quickly if experienced recruiters, account managers, and credentialing specialists leave. Their relationships and institutional knowledge often sit outside formal databases. Barton Associates will consequently need to consider workforce retention alongside technology consolidation, client migration, and operating efficiency. Much of LocumsCollective’s intrinsic value relies directly on the professionals who understand its unique processes and customer needs. insight.tmcnet.com/insight/b…

The acquisition of Zurich-based Akenza brings together two parts of the industrial data chain that are often managed separately: collecting information from physical processes and moving that information into applications where it can support decisions. The transaction expands the acquirer's Industrial IoT and digital portfolio beyond its established measurement business. The organization supplies level, pressure and point-level measurement systems used in process industries, while Akenza provides an enterprise IoT platform designed to connect heterogeneous devices, networks and business systems. In practical terms, the companies are pursuing an end-to-end path from sensor readings to cloud applications. The VEGA announcement describes the combination as a way to unite process automation expertise with IoT capabilities and eventually offer scalable solutions from a single source. That positioning matters because industrial IoT projects rarely begin with a blank sheet of paper. Plants may contain equipment from several generations, multiple sensor suppliers and a mix of operational technology and enterprise software. Connecting those environments can involve protocol translation, device management, data normalization and application integration. A technology-agnostic platform can help coordinate that work without requiring every asset to originate from the same manufacturer. Gathering more data is not, by itself, a business result. Industrial operators need trustworthy measurements, reliable transport and a practical way to route information into maintenance, production, compliance or resource-management workflows. Combining the field-level instrumentation with Akenza’s platform could reduce some of the handoffs between those layers. insight.tmcnet.com/insight/v…

Flip Electronics has entered an authorized distribution partnership with Synaptics, expanding access to components used across embedded computing, wireless connectivity and sensing applications. The agreement has a particular focus on hard-to-find, aged, last-time-buy, end-of-life and obsolete products supporting automotive and Internet of Things deployments. That focus distinguishes the deal from a conventional catalog expansion. Evertiq reports that the distributor will support Synaptics products for automotive and IoT customers, two markets where deployed equipment may remain in service far longer than the commercial life of an individual semiconductor. A connected industrial controller, vehicle subsystem or building sensor can operate for years after its original components leave active production. When one part becomes unavailable, an OEM may face a difficult choice: locate remaining inventory, qualify an alternative or redesign the affected board. Each route carries engineering, compliance and procurement implications. Obsolescence is rarely just a purchasing problem. A replacement component can require firmware changes, electrical validation, thermal testing and renewed regulatory work. Even a seemingly minor substitution may consume engineering capacity that an OEM would rather direct toward new products. Flip Electronics specializes in this later stage of the component lifecycle. Its addition to the Synaptics channel complements established distributors including Arrow Electronics, Avnet and WT Microelectronics, while giving customers another authorized route for sourcing legacy inventory. Electronics For You characterized the partnership around obsolete-component distribution, an increasingly visible requirement as connected products remain deployed across fragmented hardware generations. insight.tmcnet.com/insight/f…

No words. Pun intended.