Financial service biz Moody's says the availability of power is still a major constraint on US datacenter growth as the pace of construction exceeds the speed at which the nation's grid infrastructure can add the requisite capacity. America is the epicenter of the datacenter construction boom, driven by the AI craze and demand for compute capacity to train ever-larger foundation models. The latest report from Moody's estimates the power needed by all those bit barns will reach 426 terawatt-hours (TWh) in 2030, citing data from the International Energy Agency (IEA). That is almost double the 2025 figure. It falls on the power sector to develop new resources to meet this extra demand as traditional power generation is struggling to keep pace. Moody's calculates that building the additional generation capacity will cost about $110 billion, adding an estimated $25 billion to $30 billion a year to electricity system costs. Yet efforts to supply the rapidly growing number of cloud and AI server farms are running up against lengthy permitting processes and long lead times for key equipment and materials, causing delays of up to seven years in some cases. Management consultancy Bain & Company highlighted the problem nearly two years ago when it warned that energy companies needed to significantly boost their generation capacity, reversing years of flat or declining demand. Transmission buildout is a parallel bottleneck, with $80 billion to $115 billion in grid investment through 2030 already approved or under construction across US power markets, Moody's says, with the potential to triple over the next decade. Investment bank Jefferies reported in June that half of the extra US datacenter capacity planned for 2026 is unlikely to come online this year, blaming power availability and grid connection setbacks, along with other issues including zoning and permitting challenges. Regions able to accelerate development, such as Texas, have experienced a wave of datacenter construction. The resulting pressure on local grids has prompted some authorities to clamp down on new connections for compute sheds. Moody's believes most new IT campus operators are still likely to seek grid connections to limit development costs and benefit from service reliability and shared redundancy. Yet many are also seeking alternatives, including on-site behind-the-meter (BTM) power generation to avoid delays in bringing facilities online. The document also notes that the growing energy demands from server farms are inflating power prices and customer bills in some regions of America, as The Register reported earlier this year. This has led to affordability concerns, fueling opposition to datacenters and prompting measures such as stricter approval frameworks and moratoriums on new projects. President Trump even introduced a Ratepayer Protection Pledge, under which tech giants and energy companies committed to shielding consumers from rising energy costs due to the datacenter boom. In the long term, Moody's believes that large-scale infrastructure investment will ultimately strengthen grid reliability and affordability, despite the risks. Extra generation capacity will alleviate longstanding supply constraints, it says, reducing price volatility during periods of extreme weather or high demand, while expanded transmission networks could improve access to lower-cost generation resources, particularly when datacenter developers or their tenants help finance the associated infrastructure. ®

Microsoft has released an out-of-band update for Windows 11 24H2 and 25H2 after September's record-breaking Patch Tuesday addressed almost 1,000 vulnerabilities across the company's products – and introduced several new problems. Those problems are related to USB Audio Class 1.0 devices, Remote Desktop Services, and host folders shared with certain Hyper-V-based Linux virtual machines via Plan9. While the Hyper-V and Remote Desktop Services issues are fixed, there is still work to do on USB audio. The update restores multichannel playback on USB Audio Class 1.0 devices that continued to work in stereo, but other affected devices may remain silent, have unresponsive volume controls, or display "This device cannot start (Code 10)" in Device Manager. Owners of devices still giving them the silent treatment were told: "Microsoft is working on a resolution and will update this documentation when more information is available." The cumulative update also includes protections for CVE-2026-62721, an elevation-of-privilege vulnerability in the Windows User-Mode Power Service. Separate September security updates for Excel also broke that rarely used function, Paste, on some installations. Microsoft has yet to issue a fix. Readers were impressively supportive of Microsoft's efforts to ensure its patches don't disrupt workflows. Paratiritis noted: "The issue with all these versions of Excel is understandable, considering such changes require testing, either manual or automated, which cannot be taken for granted for such cryptic functionality as Paste." Roj Blake suggested that perhaps somebody had tried submitting a bug report, but was unable to paste the data for some reason. ®

Local authorities across the US often lack the negotiating power and know-how to work effectively with developers proposing new datacenters in their communities. The Federation of American Scientists (FAS) is trying to fill that gap by arming officials with the right questions to ask and the bargaining tools to use. The AI industry has triggered a surge in server farms construction, leaving many local governments to negotiate with developers over massive new facilities, often for the first time. To help, FAS has put together a report combining an overview of what local governments are already doing with a look at Community Benefits Agreements (CBA) between authorities and developers. The document, Before Breaking Ground: A Local Government Guide to Better Data Center Policy and Community Benefits, can be downloaded from the nonprofit think tank’s website. One key recommendation is that state and local governments should not allow for tax exemptions or provide economic incentives to datacenter operators without the provision of opening these up for negotiation beyond five years. This is because the property value of the site may increase substantially while the authority ends up out of pocket, as The Register has previously reported. Some states are losing more than $1 billion per year in tax abatementsand rebates offered as incentives to developers. CBAs have been used in numerous contexts to ensure local community groups obtain the benefit of the bargain when negotiating with a developer over large infrastructure projects. Their purpose is to ensure developers act in the interest of the community, such as building or upgrading infrastructure, hiring locally, and abiding by sustainable practices. The report lists 10 CBAs, nut only one - between Lancaster, Pennsylvania and the developers of two AI campuses - is explicitly labeled a CBA and written specifically around community benefits. In general, they cover a number of policy areas, including taxes and financial contributions, energy and water use, noise, lighting and e-waste, plus legal compliance and potential remedies. The FAS recommends future agreements specify what happens during a sudden surge in energy demand, and whether the datacenter would generate its own power or draw from the grid. Likewise, agreements should include reporting requirements on the facility's water use, and the procedure for securing additional water beyond an average day’s demand. However, one activist has already filed a lawsuit against Amazon Web Services, alleging the company published false and misleading statements about water use at its facilities in Virginia. Authorities wishing to limit noise should consider adding a specific, measurable - and therefore enforceable - numerical figure in their CBA, plus yearly testing near to local residences regardless of whether complaints are received or not. When looking to specifying remedies, CBAs could include language allowing special or consequential damages, and contain a clause on alternative dispute resolution, like mediation, instead of relying on litigation for issue resolution. The FAS also advises that CBAs should supplement local authority ordinances – laws or regulations - rather than replacing them. The two also need to be written to reinforce each other because “a CBA cannot fix a zoning gap and a zoning ordinance cannot capture specific project commitments like a CBA can.” Another notable point in the report relates to decommissioning regulations remaining largely untested. FAS isn't aware of a single case where a datacenter had ceased operations or been formally decommissioned, meaning enforcement provisions in agreements and ordinances have yet to be tested. However, it recommends that authorities should approve a decommissioning plan before a datacenter gets the go-ahead. ®

Nscale accounts for nearly two-thirds of the equity funding raised by UK datacenter infrastructure companies, according to a report from Tracxn. The market research firm identifies the rent-a-GPU operation as the only unicorn – a privately held company valued at more than $1 billion – in Britain's datacenter infrastructure sector. Nscale leads the pack with $3.7 billion raised, according to Tracxn, out of $5.7 billion across the sector. Other notable players identified by Tracxn include GreenScale, with $1.3 billion; Kao Data, with $177 million; Verne, with $125 million; Iceotope, with $107 million; and Six Degrees Group, with $106 million. Iceotope is not a datacenter operator but a developer of liquid-cooling systems, hence its inclusion in Tracxn's broader infrastructure category. Tracxn counts 154 companies in the sector, 24 of which have raised equity funding. It says investment has surged in recent years amid growing demand for AI infrastructure. Tracxn's report, Datacenter Infrastructure – UK, covers British companies that operate server farms or provide supporting IT infrastructure, such as Iceotope. It excludes businesses focused on datacenter construction or real estate. Companies based in London attracted 93 percent of the funding tracked by Tracxn, highlighting the sector's geographic concentration. Tracxn also says that the Brit datacenter infrastructure sector has recorded 36 acquisitions but only two IPOs. TelecityGroup was the largest acquisition, absorbed by US giant Equinix for $3.8 billion. Reports covering the broader UK datacenter construction market provide a different perspective. Infrastructure specialist Onnec cites Barbour ABI figures showing nearly 100 UK datacenter projects worth more than £36 billion ($49 billion) in the pipeline. That construction activity includes substantial investment by overseas operators, a market excluded from Tracxn's figures. US-headquartered operators including Digital Realty, Vantage Data Centers, CyrusOne, and Equinix have a substantial UK presence, alongside cloud giants Amazon Web Services, Microsoft, and Google. Onnec's research also points to the risk of datacenter infrastructure failing to keep pace with investment in AI hardware. Its latest survey found that 43 percent of operators had needed infrastructure upgrades or remedial work after facilities went live. The survey covered 300 decision-makers across the UK, Ireland, and the Nordic countries, rather than the UK alone. ®

Google plans to invest at least €13 billion ($15 billion) in digital infrastructure and supporting projects in Finland over the next two years. The cloud and search services giant says it is deepening its long-term commitment to Finland and to Europe through additional datacenters and investments in low-carbon electricity generation, onshore wind, and battery storage. Google already has a massive campus at Hamina, on Finland's south coast, which houses seven separate data halls. This is just along the coast from Loviisa, where one of Finland's two operating nuclear power plants is sited. The firm says it intends to expand its facilities at Hamina, but is also constructing new datacenters at sites near Kajaani, Muhos, and Vaala, further north in the country. However, it declined to specify how many facilities it is building, or how much extra infrastructure capacity the projects will add to Google's global network. Google said it chose the sites after consulting Business Finland and grid operator Fingrid, focusing on locations in northern and central Finland with existing grid infrastructure and access to carbon-free electricity. Local energy partnerships also form part of the Chocolate Factory's plans. It has signed a 22-year power purchase agreement (PPA) with Fortum to support the extension and upgrade of the energy company's Loviisa nuclear plant. Fortum said the plant could not continue operating beyond 2030 without this investment. The Google agreement provides revenue certainty for that work, which is intended to keep the plant – currently responsible for about 10 percent of Finland's electricity – operating through 2050. According to Reuters, the agreement will give Google access to as much as 50 percent of the plant's generating capacity. Fortum said the deal will begin at a lower level in 2028 before reaching 50 percent from 2030 through 2049, which Google confirmed to The Register. The cloud and search firm says it has secured additional onshore wind PPAs with green energy firms Valorem and Suomen Hyötytuuli, bringing the new-to-the-grid onshore wind capacity it has contracted in Finland to 629 MW. Google has also contracted for a 94 MW battery system near its Kajaani site, which is expected to connect to the grid in late 2027. The system is intended to balance supply and reduce price volatility during cold, windless periods. "This investment underscores Google's commitment to grow our presence responsibly, pairing the expansion of our technical infrastructure with new energy capacity, grid enhancements, and energy affordability initiatives," commented Ruth Porat, president and chief investment officer of Alphabet and Google. Google did not answer when The Register asked why it had chosen Finland for this expansion. Obvious attractions include its extensive carbon-free electricity supply – almost 96 percent of the country's generation is described as carbon-neutral – as well as established grid infrastructure near the selected sites. ®

Two runways are currently closed at Miami International Airport after an Amazon-branded cargo plane overran a runway, resulting in five fatalities. The leased 767, which bears the “Prime Air” brand Amazon uses for its air freight and drone operations, landed but was unable to stop before the end of the 2850-meter runway 30. The plane appears to have exited the runway and the airport grounds, crossed a road, and come to a halt in what looks like a carpark. As it ran across grass beyond the runway, the plane’s nose pitched down into the ground, its starboard engine caught fire, and the craft sustained extensive damage. Local authorities have confirmed that five people died, and three are severely injured. A statement from Amazon spokesperson Kelly Nantel says the company is”heartbroken to learn that five people lost their lives in today’s incident at Miami International Airport.” The statement identifies the operator of the plane as 21Air, a company that leases cargo planes and crew to its customers. 21Air CEO Keith Winters said “We are devastated by the accident involving one of our aircraft in Miami today. Our deepest condolences are with the families and loved ones of those who lost their lives. Our immediate priorities are supporting those affected, assisting the authorities, and ensuring that accurate information is communicated as it becomes available.” At the time of writing, the crash is just over 12 hours in the past, so investigators’ work has just begun. A video purportedly showing the plane landing depicts dark storm clouds and rain around the airport. That weather could conceivably have created gusty conditions that made landing unusually challenging. Federal records indicate the plane rolled off the Boeing production line in 1994. According to Planespotters, its first owner was Belgian carrier Sobelair, and it was later bought by carriers from Vietnam, Spain, Kenya, and Russia, before freight airline Atlas acquired it in 2015 and converted it from passenger to freight operations. That’s a common fate for older planes, which passenger airlines sell at very low prices. Cargo airlines can therefore pick up well-maintained planes at very low prices, meaning they have lower borrowing or leasing costs. That cheap purchase price offsets the higher maintenance and operating costs that come with older planes. All commercial aircraft, however, share strict airworthiness requirements. This is the second major crash involving a 767 leased by Amazon, after the 2019 incident that saw a plane unexpectedly nose-dive. ®

Machine learning researchers have a new and expensive option for doing local AI. AMD just raided its spare parts bin and cobbled together a DGX Station rival boasting up to 576 GB of HBM3e memory that it’s calling the Threadripper Halo. Announced at IFA 2026, the system, which is set to launch next year, marks the first time that AMD’s Instinct accelerators have been offered in a workstation form factor. But much like the AI Halo we reviewed earlier this year, the Threadripper Halo isn’t packing anything we haven’t seen before. The liquid-cooled system is powered by a 96-core Threadripper PRO 9995WX, which you may recall made its debut in 2025. The CPU is backed by up to 2 TB of DDR5 memory and up to 2.6 TB of combined system memory. As you probably already guessed, AMD is positioning the box as a local AI workstation for researchers. From what we gather, the system can be had with up to four of the company’s PCIe-based Instinct GPUs — though the systems displayed at IFA only showed two installed. Launched back in May, the MI350P is essentially half of an MI350X that’s been crammed into a PCIe form factor. Each accelerator packs 144 GB of speedy HBM3e good for 4 TB/s of memory bandwidth, and up to 4.6 petaFLOPS of FP4 compute. But we’ll note that it's in its 600 W configuration. When deployed in the Threadripper Halo, we expect AMD will be running the cards in the more sedate 450 W config. Combined, the Threadripper Halo promises up to 576 GB of HBM3e and 16 TB/s of memory bandwidth, enabling the system to run models exceeding a trillion parameters in size (at four-bit precision) entirely in GPU memory. By offloading some of the model to system memory, the system should be able to run the largest open weights models, like Moonshot.AI’s 2.8 trillion-parameter Kimi K3. The ability to run large frontier-class models on your desk won’t come cheap and may not be available in every market, at least not without an electrical service upgrade. AMD hasn’t shared pricing, but as configured, we expect it to sell for somewhere between $100,000 and $150,000. A quad MI350P system is going to push the limits of a standard North American power outlet unless AMD either underclocks the cards from 600 W to 300 W or it mandates a 20 amp circuit. This might explain why the system AMD showed off at IFA only had two. At this price and power level, the system is well positioned to compete with Nvidia’s own high-end AI workstation, the DGX Station. Announced at GTC in 2025, the DGX Station features a 252 GB B300 GPU and a 72-core Grace CPU along with 496 GB of LPDDR5x memory, and it retails for around $100,000 if you can find one in stock. The House of Zen boasts that the system offers up to 3.4x the total system memory and more than twice the memory bandwidth of the DGX Station, which should give an edge in LLM inference, assuming tensor parallel operations don’t bottleneck on the CPU’s PCIe bus. AMD says the Threadripper Halo will be available starting next year. ®

Microsoft is upping the ante on Exchange Server security, warning that outdated on-premises systems will soon be blocked from sending email to Exchange Online. From the second week of September, Exchange Server 2016 and 2019 machines that send email to Exchange Online through an inbound connector of the OnPremises type must be patched to the final public update baseline released in October 2025, according to Microsoft. "This update level was released almost a year ago," Microsoft stated, "and all organizations should have updated to it." Microsoft has been advancing its "oldest acceptable version" for some time, but this change takes the baseline to the final public update for Exchange 2016 and 2019. Any subsequent increase would leave only customers enrolled in the Extended Security Update (ESU) program or those who have migrated to Exchange Server Subscription Edition able to meet it. Organizations may retain on-premises Exchange servers to support hybrid deployments, applications, or other systems that depend on them for email. According to Microsoft, throttling and blocking apply only to servers sending email to Exchange Online through an inbound OnPremises connector. Other delivery methods are unaffected, and the policy does not necessarily cover "all servers in your organization." The company added: "This might change in the future," which doesn't sound sinister at all. Administrators must balance the need to apply security patches promptly against the risk that an update introduces problems of its own – something Microsoft has, to put it delicately, caused in the past. Some administrators responding to the announcement were more than a little perturbed, although Microsoft has applied throttling and blocking to "significantly out of date" Exchange Server installations for several years. The announcement gives administrators advance notice of a change Microsoft says would have happened anyway. "With or without this announcement, the change would have been made (just as it was made in the past, without any announcement)," explained Microsoft Principal Project Manager Nino Bilic. "The bottom line is that admins really need to understand that updating Exchange should be done on time." ®

Premier League football club Tottenham Hotspur has ditched VMware as part of a wider datacenter and network refresh, saving itself more than 85 percent on licensing costs - though its Chief Technology Officer says the decision wasn't purely financial. Transforming Spurs into a winning force on the pitch remains a work in progress. Off the pitch, though, the club is nearing the end of a project to overhaul the tech inside its White Hart Lane stadium in north London. The ground holds almost 63,000 fans and runs more than 1,650 Wi-Fi access points, 840+ under seat access points, 700 Bluetooth beacons, 20,000 network access points with 1,849 IPTV screens and 519 CCTV cameras connected. A six aisle datacenter is also sited within the arena. Rob Pickering, CTO at Spurs, told The Register that on the server, storage and networking side, the club has switched entirely to the HPE GreenLake portfolio, “underpinned by OpsRamp and Morpheus (VM Essentials).” HPE will run the hybrid cloud managed service. In the datacenter, Spurs has moved to ProLiant Compute Gen12 servers and HPE Alletra storage MP. Connectivity is being upgraded to WiFi 7 and HPE Aruba Central with self-driving capabilities. Why migrate from VMware? “Savings is a big part of the reason,” said Pickering. "It’s a significant amount, north of 85 percent,” less than the club was forking out for licensing. “We moved away as we in put the infrastructure… over the last three months,” he added. “The issues with the Broadcom takeover of VMware are well known in terms of the commercial changes to their structure and the way that they engage with clients.” Since buying VMware in 2023, Broadcom has discontinued certain standalone products in favor of bundles that are often pricier regardless of whether customers use every components. Broadcom seems only focused on VMware's 10,000 to 30,000 largest customers, out of a base that numbered 350,000 before the takeover. Pickering framed virtualization itself as a commodity: “If it's not built into a broader AI operation stack, the value is even lower again. Having that all connected back into an operation stack that our team can work from - one pane of glass - is worth more than just the saving that come from moving vendor A to vendor B. We do see significant savings from that, but really it's all part of the broader story of how do we put AI ops at the heart of what we do from a technology perspective.” He said the tech team at Spurs is relatively small, so the "ability to use AI operations to proactively see faults and remediate them without the input of a person is really critical to us." Spurs declined to disclose what it had been paying VMware or which VMware products it used. Pickering did, however, describe HPE VM Essentials (VME) as a "completely capable replacement." The Aruba network upgrade is “running late” and should be completed within a month. Getting it right matters as up to about 30,000 fans hop onto the stadium’s Wi-Fi at any given point during a game. The ground also doubles as a concert venue and hosts NFL matches and rugby league games too. “The process of replacing all of those consistently in a non-impactful way - given the stadium operates really 365 days a year - is one of the hardest challenges of doing any technology project in a stadium. There's no real window of downtime in stadium.” Of course there's the physics too: "A stadium is a giant Faraday cage. It's glass and steel. It's very hard to get RF to work. We make sure we build things that ensure the RS signal can travel the distance it needs to. We also need a lot of density to do that… We're doing like 1,800 access points across the stadium in totality. So it's significant in terms of the physical engineering challenge to get it in here, but also the radio engineering challenge." With Spurs losing its first two Premier League Games despite spending £300 million on new players during the summer, the club’s technology upgrades may be delivering more reliable returns than its transfer business, at least for now. ®