Google and Anthropic have finalized a cloud partnership worth tens of billions of dollars, granting Anthropic access to up to one million of Google's Tensor Processing Units and more than a gigawatt of compute power by 2026. CNBC reports: Industry estimates peg the cost of a 1-gigawatt data center at around $50 billion, with roughly $35 billion of that typically allocated to chips. While competitors tout even loftier projections -- OpenAI's 33-gigawatt "Stargate" chief among them -- Anthropic's move is a quiet power play rooted in execution, not spectacle. Founded by former OpenAI researchers, the company has deliberately adopted a slower, steadier ethos, one that is efficient, diversified, and laser-focused on the enterprise market.

A key to Anthropic's infrastructure strategy is its multi-cloud architecture. The company's Claude family of language models runs across Google's TPUs, Amazon's custom Trainium chips, and Nvidia's GPUs, with each platform assigned to specialized workloads like training, inference, and research. Google said the TPUs offer Anthropic "strong price-performance and efficiency." [...] Anthropic's ability to spread workloads across vendors lets it fine-tune for price, performance, and power constraints. According to a person familiar with the company's infrastructure strategy, every dollar of compute stretches further under this model than those locked into single-vendor architectures.

Microsoft is retiring Office Online Server on December 31, 2026, ending support and updates for organizations running browser-based Office apps on-premises. The Register reports: After this, there won't be any more security fixes, updates, or technical support from Microsoft. "This change is part of our ongoing commitment to modernizing productivity experiences and focusing on cloud-first solutions," the company said. Office Online Server provides browser-based versions of Word, Excel, PowerPoint, and OneNote for customers who want to keep things on-prem without having to roll out the full desktop applications. Microsoft's solution is to move to Microsoft 365, its decidedly off-premises version of its applications. The company said it is "focusing its browser-based Office app investments on Office for the Web to deliver secure, collaborative, and feature-rich experiences through Microsoft 365."

Other than migrating to another platform when the vendor pulls the plug, affected customers have few options. The announcement will also hit several customers running SharePoint Server SE or Exchange Server SE. While those products remain supported, Office Online Server integration will go away. The company suggested Microsoft 365 Apps for Enterprise and Office LTSC 2024 as alternatives for viewing and editing documents hosted on those servers.

Skype for Business customers will also lose some key features related to PowerPoint. Presenter notes and high-fidelity PowerPoint rendering will go away. In-meeting annotations, which allow meeting participants to write directly to slides without altering the original file, will no longer be available, and embedded video playback will run at lower fidelity. Features like whiteboards, polls, and app sharing shouldn't be affected. Microsoft's solution is a move to Teams, which the company says "offers modern meeting experiences."

Google is migrating all its internal workloads to run on both x86 and its custom Axion Arm chips, with major services like YouTube, Gmail, and BigQuery already running on both architectures. The Register reports: The search and ads giant documented its move in a preprint paper published last week, titled "Instruction Set Migration at Warehouse Scale," and in a Wednesday post that reveals YouTube, Gmail, and BigQuery already run on both x86 and its Axion Arm CPUs -- as do around 30,000 more applications. Both documents explain Google's migration process, which engineering fellow Parthasarathy Ranganathan and developer relations engineer Wolff Dobson said started with an assumption "that we would be spending time on architectural differences such as floating point drift, concurrency, intrinsics such as platform-specific operators, and performance." [...]

The post and paper detail work on 30,000 applications, a collection of code sufficiently large that Google pressed its existing automation tools into service -- and then built a new AI tool called "CogniPort" to do things its other tools could not. [...] Google found the agent succeeded about 30 percent of the time under certain conditions, and did best on test fixes, platform-specific conditionals, and data representation fixes. That's not an enormous success rate, but Google has at least another 70,000 packages to port.

The company's aim is to finish the job so its famed Borg cluster manager -- the basis of Kubernetes -- can allocate internal workloads in ways that efficiently utilize Arm servers. Doing so will likely save money, because Google claims its Axion-powered machines deliver up to 65 percent better price-performance than x86 instances, and can be 60 percent more energy-efficient. Those numbers, and the scale of Google's code migration project, suggest the web giant will need fewer x86 processors in years to come.

Early Monday, an Amazon Web Services outage disrupted banks, games, and Peloton classes. Eight Sleep customers faced a different problem. Their internet-enabled mattresses malfunctioned. People woke to beds locked in upright positions, excessive heat, flashing lights, and unexpected alarms. Matteo Franceschetti, the company's chief executive, apologized and said engineers were building an outage-proof mode. By Monday evening, all devices functioned again, though some experienced data processing delays. The mattresses adjust temperature between 55 and 110 degrees and elevate bodies into different positions. They activate soundscapes and vibrational alarms. The advanced models cost over $5,000. A yearly subscription of $199 to $399 is required for temperature controls.

An anonymous reader quotes a report from Ars Technica: On Monday afternoon, Amazon confirmed that an outage affecting Amazon Web Services' cloud hosting, which had impacted millions across the Internet, had been resolved. Considered the worst outage since last year's CrowdStrike chaos, Amazon's outage caused "global turmoil," Reuters reported. AWS is the world's largest cloud provider and, therefore, the "backbone of much of the Internet," ZDNet noted. Ultimately, more than 28 AWS services were disrupted, causing perhaps billions in damages, one analyst estimated for CNN.

[...] Amazon's problems originated at a US site that is its "oldest and largest for web services" and often "the default region for many AWS services," Reuters noted. The same site has experienced two outages before in 2020 and 2021, but while the tech giant had confirmed that those prior issues had been "fully mitigated," apparently the fixes did not ensure stability into 2025. ZDNet noted that Amazon's first sign of the outage was "increased error rates and latency across numerous key services" tied to its cloud database technology. Although "engineers later identified a Domain Name System (DNS) resolution problem" as the root of these issues and quickly fixed it, "other AWS services began to fail in its wake, leaving the platform still impaired" as more than two dozen AWS services shut down. At the peak of the outage on Monday, Down Detector tracked more than 8 million reports globally from users panicked by the outage, ZDNet reported. Ken Birman, a computer science professor at Cornell University, told Reuters that "software developers need to build better fault tolerance."

"When people cut costs and cut corners to try to get an application up, and then forget that they skipped that last step and didn't really protect against an outage, those companies are the ones who really ought to be scrutinized later."

Alibaba Cloud claims its new Aegaeon GPU pooling system cuts Nvidia GPU use by 82%, letting 213 H20 accelerators handle workloads that previously required 1,192. The advancements have been detailed in a paper (PDF) at the 2025 ACM Symposium on Operating Systems (SOSP) in Seoul. Tom's Hardware reports: Unlike training-time breakthroughs that chase model quality or speed, Aegaeon is an inference-time scheduler designed to maximize GPU utilization across many models with bursty or unpredictable demand. Instead of pinning one accelerator to one model, Aegaeon virtualizes GPU access at the token level, allowing it to schedule tiny slices of work across a shared pool. This means one H20 could serve several different models simultaneously, with system-wide "goodput" -- a measure of effective output -- rising by as much as nine times compared to older serverless systems.

The system was tested in production over several months, according to the paper, which lists authors from both Peking University and Alibaba's infrastructure division, including CTO Jingren Zhou. During that window, the number of GPUs needed to support dozens of different LLMs -- ranging in size up to 72 billion parameters -- fell from 1,192 to just 213. While the paper does not break down which models contributed most to the savings, reporting by the South China Morning Post says the tests were conducted using Nvidia's H20, one of the few accelerators still legally available to Chinese buyers under current U.S. export controls.

Foreign hackers breached the National Nuclear Security Administration's Kansas City National Security Campus (KCNSC) by exploiting unpatched Microsoft SharePoint vulnerabilities. The intrusion happened in August and is possibly linked to either Chinese state actors or Russian cybercriminals. CSO Online notes that "roughly 80% of the non-nuclear parts in the nation's nuclear stockpile originate from KCNSC," making it "one of the most sensitive facilities in the federal weapons complex." From the report: The breach targeted a plant that produces the vast majority of critical non-nuclear components for US nuclear weapons under the NNSA, a semi-autonomous agency within the Department of Energy (DOE) that oversees the design, production, and maintenance of the nation's nuclear weapons. Honeywell Federal Manufacturing & Technologies (FM&T) manages the Kansas City campus under contract to the NNSA. [...] The attackers exploited two recently disclosed Microsoft SharePoint vulnerabilities -- CVE-2025-53770, a spoofing flaw, and CVE-2025-49704, a remote code execution (RCE) bug -- both affecting on-premises servers. Microsoft issued fixes for the vulnerabilities on July 19.

On July 22, the NNSA confirmed it was one of the organizations hit by attacks enabled by the SharePoint flaws. "On Friday, July 18th, the exploitation of a Microsoft SharePoint zero-day vulnerability began affecting the Department of Energy," a DOE spokesperson said. However, the DOE contended at the time, "The department was minimally impacted due to its widespread use of the Microsoft M365 cloud and very capable cybersecurity systems. A very small number of systems were impacted. All impacted systems are being restored." By early August, federal responders, including personnel from the NSA, were on-site at the Kansas City facility, the source tells CSO.

AWS experienced a three-hour outage early Monday morning that disrupted thousands of websites and applications across the globe. The cloud computing provider reported DNS problems with DynamoDB in its US-EAST-1 region in northern Virginia starting at 12:11 a.m. Pacific time. Over 4 million users reported issues, according to Downdetector. Snapchat saw reports spike from more than 22,000 to around 4,000 as systems recovered. Roblox dropped from over 12,600 complaints to fewer than 500. Reddit and the financial platform Chime remained affected longer. Perplexity, Coinbase and Robinhood attributed their platform disruptions directly to AWS.

Gaming platforms including Fortnite, Clash Royale and Clash of Clans went offline. Signal confirmed the messaging app was down. In Britain, Lloyd Bank, Bank of Scotland, Vodafone, BT, and the HMRC website faced problems. United Airlines reported disrupted access to its app and website overnight. Some internal systems were temporarily affected. Delta experienced a small number of minor flight delays. By 3:35 a.m. Pacific time, AWS said the issue had been fully mitigated. Most service operations were succeeding normally though some requests faced throttling during final resolution. AWS holds roughly one-third of the cloud infrastructure market ahead of Microsoft and Google.

On Cloudflare's blog, a senior research engineer shares a plan for "improving the trustworthiness of JavaScript on the web."

"It is as true today as it was in 2011 that Javascript cryptography is Considered Harmful." The main problem is code distribution. Consider an end-to-end-encrypted messaging web application. The application generates cryptographic keys in the client's browser that lets users view and send end-to-end encrypted messages to each other. If the application is compromised, what would stop the malicious actor from simply modifying their Javascript to exfiltrate messages? It is interesting to note that smartphone apps don't have this issue. This is because app stores do a lot of heavy lifting to provide security for the app ecosystem. Specifically, they provide integrity, ensuring that apps being delivered are not tampered with, consistency, ensuring all users get the same app, and transparency, ensuring that the record of versions of an app is truthful and publicly visible.

It would be nice if we could get these properties for our end-to-end encrypted web application, and the web as a whole, without requiring a single central authority like an app store. Further, such a system would benefit all in-browser uses of cryptography, not just end-to-end-encrypted apps. For example, many web-based confidential LLMs, cryptocurrency wallets, and voting systems use in-browser Javascript cryptography for the last step of their verification chains. In this post, we will provide an early look at such a system, called Web Application Integrity, Consistency, and Transparency (WAICT) that we have helped author. WAICT is a W3C-backed effort among browser vendors, cloud providers, and encrypted communication developers to bring stronger security guarantees to the entire web... We hope to build even wider consensus on the solution design in the near future....

We would like to have a way of enforcing integrity on an entire site, i.e., every asset under a domain. For this, WAICT defines an integrity manifest, a configuration file that websites can provide to clients. One important item in the manifest is the asset hashes dictionary, mapping a hash belonging to an asset that the browser might load from that domain, to the path of that asset.
The blog post points out that the WEBCAT protocol (created by the Freedom of Press Foundation) "allows site owners to announce the identities of the developers that have signed the site's integrity manifest, i.e., have signed all the code and other assets that the site is serving to the user... We've made WAICT extensible enough to fit WEBCAT inside and benefit from the transparency components." The proposal also envisions a service storing metadata for transparency-enabled sites on the web (along with "witnesses" who verify the prefix tree holding the hashes for domain manifests).

"We are still very early in the standardization process," with hopes to soon "begin standardizing the integrity manifest format. And then after that we can start standardizing all the other features. We intend to work on this specification hand-in-hand with browsers and the IETF, and we hope to have some exciting betas soon. In the meantime, you can follow along with our transparency specification draft,/A>, check out the open problems, and share your ideas."

An anonymous reader shares a report: Cloud-computing firm Salesforce was hit with a proposed class action lawsuit by two authors who alleged the company used thousands of books without permission to train its AI software. Novelists Molly Tanzer and Jennifer Gilmore said in the complaint that Salesforce infringed copyrights by using their work to train its xGen AI models to process language.

Backblaze has been tracking hard disk drive failures in its datacenter since 2013. The backup and cloud storage company's latest analysis of approximately 317,230 drives shows that peak failure rates have dropped dramatically and shifted much later in a drive's lifespan. Where the company once saw failure rates of 13.73% at around three years in 2013 and 14.24% at seven years and nine months in 2021, the current data shows a peak of just 4.25% at 10 years and three months.

This represents the first time the company has observed the highest failure rate occurring at the far end of the drive curve rather than earlier in its operational life, it said. The drives maintained relatively consistent failure rates through most of their use before spiking sharply near the end. The improvement amounts to roughly one-third of the previous peak failure rates.

Physicists using SLAC's X-ray free-electron laser discovered two new laser phenomena that allowed them to generate the shortest, highest-energy X-ray pulses ever recorded (60-100 attoseconds). These breakthroughs could let scientists observe electron motion and chemical bond formation in real time. Physicists Uwe Bergmann and Thomas Linker write in an article for The Conversation: In this new study we used X-rays, which have 100 million times shorter wavelengths than microwaves and 100 million times more energy. This meant the resulting new X-ray laser pulses were split into different X-ray wavelengths corresponding to Rabi frequencies in the extreme ultraviolet region. Ultraviolet light has a frequency 100 million times higher than radio waves. This Rabi cycling effect allowed us to generate the shortest high-energy X-ray pulses to date, clocking in at 60-100 attoseconds.

While the pulses that X-ray free-electron lasers currently generate allow researchers to observe atomic bonds forming, rearranging and breaking, they are not fast enough to look inside the electron cloud that generates such bonds. Using these new attosecond X-ray laser pulses could allow scientists to study the fastest processes in materials at the atomic-length scale and to discern different elements.

In the future, we also hope to use much shorter X-ray free-electron laser pulses to better generate these attosecond X-ray pulses. We are even hoping to generate pulses below 60 attoseconds by using heavier materials with shorter lifespans, such as tungsten or hafnium. These new X-ray pulses are fast enough to eventually enable scientists to answer questions such as how exactly an electron cloud moves around and what a chemical bond actually is. The findings have been published in the journal Nature.

China now generates well over twice as much electricity as the United States. The country's economy has become substantially larger than America's in real terms, measured at purchasing power parity, economist Paul Krugman wrote this week. The Trump administration has moved aggressively against renewable energy development. It rolled back Biden's tax incentives for renewables through the One Big Beautiful Bill. The administration is attempting to stop a nearly completed offshore wind farm that could power hundreds of thousands of homes. It canceled $7 billion in grants for residential solar panels. A solar energy project that would have powered almost 2 million homes was killed. The administration canceled $8 billion in clean energy grants, mostly in Democratic states, and is reportedly planning to cancel tens of billions more. Energy Secretary Chris Wright said solar power is unreliable because "you have to have power when the sun goes behind a cloud and when the sun sets, which it does almost every night."

California has already integrated substantial solar power into its grid through battery storage technology. Republican support for higher education has collapsed over the past decade, according to polling data. The administration has also targeted vaccines and research in multiple areas. Krugman argues that by 2028 America will have fallen so far behind China that it is unlikely to catch up.

An anonymous reader quotes a report from BleepingComputer: U.S. cybersecurity company F5 disclosed that nation-state hackers breached its systems and stole undisclosed BIG-IP security vulnerabilities and source code. The company states that it first became aware of the breach on August 9, 2025, with its investigations revealing that the attackers had gained long-term access to its system, including the company's BIG-IP product development environment and engineering knowledge management platform.

F5 is a Fortune 500 tech giant specializing in cybersecurity, cloud management, and application delivery networking (ADN) applications. The company has 23,000 customers in 170 countries, and 48 of the Fortune 50 entities use its products. BIG-IP is the firm's flagship product used for application delivery and traffic management by many large enterprises worldwide. [...]

F5 is still reviewing which customers had their configuration or implementation details stolen and will contact them with guidance. To help customers secure their F5 environments against risks stemming from the breach, the company released updates for BIG-IP, F5OS, BIG-IP Next for Kubernetes, BIG-IQ, and APM clients. Despite any evidence "of undisclosed critical or remote code execution vulnerabilities," the company urges customers to prioritize installing the new BIG-IP software updates.

An anonymous reader shares a report: After acquiring GitHub in 2018, Microsoft mostly let the developer platform run autonomously. But in recent months, that's changed. With GitHub CEO Thomas Dohmke leaving the company this August, and GitHub being folded more deeply into Microsoft's organizational structure, GitHub lost that independence. Now, according to internal GitHub documents The New Stack has seen, the next step of this deeper integration into the Microsoft structure is moving all of GitHub's infrastructure to Azure, even at the cost of delaying work on new features.

[...] While GitHub had previously started work on migrating parts of its service to Azure, our understanding is that these migrations have been halting and sometimes failed. There are some projects, like its data residency initiative (internally referred to as Project Proxima) that will allow GitHub's enterprise users to store all of their code in Europe, that already solely use Azure's local cloud regions.

An anonymous reader quotes a report from Ars Technica: Bose will brick key features of its SoundTouch Wi-Fi speakers and soundbars soon. On Thursday, Bose informed customers that as of February 18, 2026, it will stop supporting the devices, and the devices' cloud-based features, including the companion app, will stop working. The SoundTouch app enabled numerous capabilities, including integrating music services, like Spotify and TuneIn, and the ability to program multiple speakers in different rooms to play the same audio simultaneously.

Bose has also said that some saved presets won't work and that users won't be able to change saved presets once the app is gone. Additionally, Bose will stop providing security updates for SoundTouch devices. The Framingham, Massachusetts-headquartered company noted to customers that the speakers will continue being able to play audio from a device connected via AUX or HDMI. Wireless playback will still work over Bluetooth; however, Bluetooth is known to introduce more latency than Wi-Fi connections. Affected customers can trade in their SoundTouch product for a credit worth up to $200.

In its notice sent to customers this week, Bose provided minimal explanation for end-of-life-ing its pricey SoundTouch speakers, saying: "Bose SoundTouch systems were introduced into the market in 2013. Technology has evolved since then, and we're no longer able to sustain the development and support of the cloud infrastructure that powers this older generation of products. We remain committed to creating new listening experiences for our customers built on modern technologies."

An anonymous reader quotes a report from CSO Online: On Sept. 17, security vendor SonicWall announced that cybercriminals had stolen backup files configured for cloud backup. At the time, the company claimed the incident was limited to "less than five percent" of its customers. Now, the firewall provider has admitted that "all customers" using the MySonicWall cloud backup feature were affected. According to the company, the stolen files contain encrypted credentials and configuration data. "[W]hile encryption remains in place, possession of these files could increase the risk of targeted attacks," SonicWall warns in its press release.

Security specialist Arctic Wolf also warns of the consequences of the incident. "Firewall configuration files store sensitive information that can be leveraged by threat actors to exploit and gain access to an organization's network," explains Stefan Hostetler, threat intelligence researcher at Arctic Wolf. "These files can provide threat actors with critical information such as user, group, and domain settings, DNS and log settings, and certificates," he adds. Arctic Wolf has previously observed threat actors, including nation-state and ransomware groups, exfiltrating firewall configuration files to use for future attacks. SonicWall urges all customers and partners to regularly check their devices for updates. Admins can find additional information here.

smooth wombat writes: At one point in technology history, floppy disks reigned supreme. Files, pictures, games, everything was put on a floppy disk. But technology doesn't stand still and as time went on disks were replaced by CDs, DVDs, thumb drives, and now cloud storage. Despite these changes, floppy disks are still found in long forgotten corners of businesses or stuffed in boxs in the attic. What is on these disks is anyone's guess, but Cambridge University Library is racing against time to preserve the data. However, lack of hardware and software to read the disks, if they're readable at all, poses unique challenges.

Some of the world's most treasured documents can be found deep in the archives of Cambridge University Library. There are letters from Sir Isaac Newton, notebooks belonging to Charles Darwin, rare Islamic texts and the Nash Papyrus -- fragments of a sheet from 200BC containing the Ten Commandments written in Hebrew.

These rare, and often unique, manuscripts are safely stored in climate-controlled environments while staff tenderly care for them to prevent the delicate pages from crumbling and ink from flaking away.

But when the library received 113 boxes of papers and mementoes from the office of physicist Stephen Hawking, it found itself with an unusual challenge. Tucked alongside the letters, photographs and thousands of pages relating to Hawking's work on theoretical physics, were items now not commonly seen in modern offices -- floppy disks.

They were the result of Hawking's early adoption of the personal computer, which he was able to use despite having a form of motor neurone disease known as amyotrophic lateral sclerosis, thanks to modifications and software. Locked inside these disks could be all kinds of forgotten information or previously unknown insights into the scientists' life. The archivists' minds boggled.

These disks are now part of a project at Cambridge University Library to rescue hidden knowledge trapped on floppy disks. The Future Nostalgia project reflects a larger trend in the information flooding into archives and libraries around the world.

American companies have begun cutting middle management positions at rates not seen in years. Google eliminated 35% of managers overseeing teams of fewer than three in August. Fiverr announced in September it would shed managers to focus on AI. Amazon trimmed its management ranks throughout the year and cut positions at its cloud-computing division in July. Meta's Mark Zuckerberg has complained about managers managing managers since 2023.

Phrases relating to reducing management layers appeared 98 times on earnings calls of companies in the S&P global index this year, twice the frequency of all of 2022. The cuts stem partly from an uncertain economic environment and President Donald Trump's tariff regime, Economist writes. The pandemic created the conditions for the current retrenchment. Companies furloughed staff during Covid-19 and then hired rapidly to meet demand for e-commerce and digital services. They promoted employees to management positions to retain talent even when those managers supervised only one or two subordinates. Between 2019 and 2024, five of the ten fastest-growing job categories were management roles. Since November 2022, listed American companies have cut middle-management positions by around 3% on average.

An anonymous reader quotes a report from 404 Media: Sora 2, Open AI's new AI video generator, puts a visual watermark on every video it generates. But the little cartoon-eyed cloud logo meant to help people distinguish between reality and AI-generated bullshit is easy to remove and there are half a dozen websites that will help anyone do it in a few minutes. A simple search for "sora watermark" on any social media site will return links to places where a user can upload a Sora 2 video and remove the watermark. 404 Media tested three of these websites, and they all seamlessly removed the watermark from the video in a matter of seconds.

Hany Farid, a UC Berkeley professor and an expert on digitally manipulated images, said he's not shocked at how fast people were able to remove watermarks from Sora 2 videos. "It was predictable," he said. "Sora isn't the first AI model to add visible watermarks and this isn't the first time that within hours of these models being released, someone released code or a service to remove these watermarks." [...] According to Farid, Open AI is decent at employing strategies like watermarks, content credentials, and semantic guardrails to manage malicious use. But it doesn't matter. "It is just a matter of time before someone else releases a model without these safeguards," he said.

Both [Rachel Tobac, CEO of SocialProof Security] and Farid said that the ease at which people can remove watermarks from AI-generated content wasn't a reason to stop using watermarks. "Using a watermark is the bare minimum for an organization attempting to minimize the harm that their AI video and audio tools create," Tobac said, but she thinks the companies need to go further. "We will need to see a broad partnership between AI and Social Media companies to build in detection for scams/harmful content and AI labeling not only on the AI generation side, but also on the upload side for social media platforms. Social Media companies will also need to build large teams to manage the likely influx of AI generated social media video and audio content to detect and limit the reach for scammy and harmful content." "I'd like to know what OpenAI is doing to respond to how people are finding ways around their safeguards," Farid said. "Will they adapt and strengthen their guardrails? Will they ban users from their platforms? If they are not aggressive here, then this is going to end badly for us all."