• Ukraine's autonomous interceptors stalk and destroy Russian Shaheds without human controls
• Ukraine compressed years of drone development into twelve battlefield months
• Brave1's interceptor automates 95% of the kill chain — the human only picks the target

Ukraine has cleared its first autonomous drone interceptor for battlefield deployment following combat testing completed recently in the Kharkiv region.

The system was developed under the Brave1 defence accelerator specifically to counter Shahed drones, which Russia increasingly launches in coordinated saturation attacks against Ukrainian cities and infrastructure.

Those attacks rely heavily upon volume and timing because large numbers of incoming drones can gradually overwhelm conventional air defence systems and human reaction speeds simultaneously.

Autonomous interceptors reach combat testing

Ukraine’s response now involves reducing how much of the interception process still depends upon direct human control during active battlefield engagements involving multiple aerial threats.

According to Ukrainian officials, the interceptor automates roughly 95% of the engagement sequence from launch through terminal destruction of the incoming drone.

Human operators still decide which drone should be engaged before the interceptor assumes responsibility for navigation, recognition, pursuit, and strike execution independently afterward.

That operational structure allows crews to supervise engagements instead of manually piloting interceptors through every stage of aerial combat under high-pressure battlefield conditions.

Ukrainian officials believe reducing operator workload could become increasingly important during large nighttime bombardments involving several incoming drones approaching defended airspace simultaneously.

The manufacturer reportedly moved from prototype development toward verified combat deployment within less than twelve months under continuing wartime operational pressures across Ukraine.

That unusually compressed timeline appears closely connected to Brave1’s institutional and financial backing, which reduced delays commonly associated with traditional peacetime procurement procedures.

Officials argue wartime conditions leave little opportunity for prolonged development schedules because interception delays increasingly determine whether drones successfully reach populated urban areas.

“We continue to systematically strengthen the defence of the sky,” the ministry stated while discussing interceptor systems already tested under active combat conditions recently.

Scaling ambitions meet an unverified record

Ukraine now says it is expanding production and deployment of these interceptors as part of broader efforts to increase domestic military drone manufacturing capacity nationwide.

Publicly available information regarding actual kill rates and long-term battlefield reliability remains extremely limited outside official Ukrainian statements.

Evaluating the system also becomes more difficult because Russia has continuously modified its Shahed drones throughout the conflict, using changing flight profiles and components.

Autonomous interception could become even more complicated once electronic interference, airborne decoys, civilian aircraft, and friendly drones begin sharing the same contested airspace simultaneously.

Because no independent technical assessment has yet been released publicly, the actual battlefield accuracy of the interceptor system remains difficult to verify externally.

The Kharkiv deployment nevertheless establishes an early proof of concept showing Ukraine’s growing interest in partially autonomous air defence systems during modern drone warfare.

Via Fedorov

Most portable power stations force you to choose among capacity, speed, and longevity. The Anker Solix F2000 is $749 (regularly $899) at Anker, and it doesn’t ask you to make that compromise.

With 2048Wh of LiFePO4 capacity, 2400W of continuous AC output, a 10-year rated lifespan, and a 0–80% recharge time of just 1.4 hours, it’s one of the most complete portable power stations you can buy at any price. And at $749 in Anker's own Prime Day sale, which guarantees the best price compared to other retailers, the value case is hard to argue against.

For home backup, charging devices and driving appliances during power outages or summer trips, this is a top pick. For more Anker products, we've rounded up the 12 best Anker Prime Day deals you can get right now.

Today's top portable power station deal

The core of the F2000 is its LiFePO4 battery chemistry, which underpins everything else the machine does well. LiFePO4 cells are chemically stable, fire-resistant, and certified for indoor use — unlike cheaper lithium-ion packs that can overheat or degrade significantly within a few hundred cycles.

The 2400W continuous output and 3600W surge ceiling via SurgePad mean this machine doesn’t just charge laptops and phones — it runs appliances. A 1500W space heater, a fridge, a microwave, power tools, a CPAP machine with heated humidifier, a coffee maker. You name it.

HyperFlash fast charging is the spec that changes how you think about using the machine. Most 2000Wh+ power stations take five to eight hours to recharge from a wall outlet — long enough that you plan your entire trip or outage around their recovery time. The F2000 goes from 0 to 80% in 1.4 hours, so you can drain it overnight during a blackout, plug it in at 7am, and have it substantially recharged before you head to work. For RV users, that means you can top it up at a campsite hook-up in the time it takes to have breakfast.

The port selection is among the most practical at this price. The dedicated TT-30 RV port is a genuine differentiator — it means F2000 owners with an RV don’t need an adapter, and it signals that Anker designed this machine for actual vehicle use rather than as an afterthought. The three 100W USB-C ports will simultaneously fast-charge three laptops or a laptop plus two tablets. The four AC outlets, two car outlets, and two USB-A ports round out a panel that handles every real-world device combination without requiring port juggling.

Solar charging supports up to 1000W input via an XT-60 connector, compatible with Anker’s own 200W and 400W panels as well as most third-party panels in the supported voltage range. This makes the F2000 a genuinely off-grid-capable system when paired with panels — useful for extended camping, van builds, and off-grid cabins. The Anker SOLIX app handles remote monitoring of battery level, charging time, and power usage over Wi-Fi and Bluetooth, with a clean interface that reviewers have found intuitive.

If you need serious, reliable backup power for a home, RV, or extended off-grid use, the Anker SOLIX F2000 is one of the strongest offers available in the 2000Wh class right now. Check out more options in our guide to the best portable power stations.

Also consider: More portable power station deals

• Shop all the portable power stations at Amazon

The AI industry may have found its next battle, and it’s not model training, GPU availability or data centers expansion. It’s actually how artificial intelligence is being framed in the first place.

For the past couple of years, AI coding assistants have been marketed as productivity boosters, promising to help across code generation, debugging and testing.

But the recent SpaceX acquisition of AI coding platform Cursor shows just how important AI coding systems can be in AI. While coding tools are generally considered from the developer’s point of view, software generation is increasingly being seen as a strategic capability for future autonomous systems.

Whether managing infrastructure, coordinating logistics or automating business processes, tomorrow’s coding agents will need to create, edit and maintain software instead of just helping developers to write code.

AI’s impact on code generation is trust

In other words, AI coding agents are less about being tools and more about being infrastructure layers through which organizations can build, operate and even govern their tech – similar to how we see cloud, networking and security today.

But in dehumanizing software generation, we’re uncovering the next challenge. Software development used to be limited by the speed at which humans can review, test and validate code before it goes into production, but it’s now all about trust and verification.

With AI now capable of generating code at lightning pace, questions around quality, security, compliance and governance are now being raised.

Acknowledging that code generation has been vastly democratized, Qodo CEO Itamar Friedman now believes what sets companies apart will be verification and governance. The winners will be the ones to deploy trustworthy software at scale, not the ones to launch it fastest.

To better understand what SpaceX’s Cursor acquisition means for the future of software engineering, how AI coding platforms are evolving from tools to infrastructure, and how we can tackle the next bottleneck that is trust, I spoke with Friedman.

• Does this acquisition signal that AI coding tools are becoming strategic infrastructure rather than just developer productivity tools?

Yes, and the price tag makes it explicit. You don’t just pay $60B for a productivity feature, you pay that price to have complete control over the coding mechanism your engineering teams run every day.

When the tool sits in the path of every change that ships, it is infrastructure, with the same reliability, security, and governance expectations as the rest of your stack.

• What does SpaceX's interest in Cursor say about the role AI coding tools will play in the next generation of AI-native companies?

AI-native companies want to own their means of production. SpaceX is not buying a chat window, it’s buying the ability to shape how code is generated across an organization that writes mission critical software.

What the next generation of tools is signaling is that the coding layer is too important to rent on someone else's roadmap. Instead, companies need to treat code generation, review, and governance as core infrastructure they direct, not as a third party relationship they hope holds up.

• Why might companies building highly complex or mission-critical systems want greater control over the AI coding layer?

In a high-stakes environment, you never let the builder be its own reviewer. When a single system both writes code and judges whether it is correct, you get a biased feedback loop where the AI is grading its own homework.

Greater control means being able to insert independent verification, enforce your own standards, and produce an audit trail of every decision.

• How should enterprises think about evaluating AI coding tools after a major ownership change or acquisition?

Enterprises should evaluate for independence and continuity, not for features. They must ask who controls the roadmap now, whether your data stays in your environment, and whether the tool still works alongside the rest of your stack or quietly steers you into another owner's ecosystem.

During an acquisition, a tool that was neutral yesterday may be optimized for its new parent tomorrow. The durable choice is the layer you can keep running on your terms regardless of who owns the model underneath it.

• As AI generates more production code, what new challenges does that create for engineering teams?

As code volumes skyrocket, the bottleneck moves from writing code to verifying it. The hard part about software was never producing more lines, it is confirming the code does what it’s intended to do, holds to your architecture, and meets your standards at scale.

When AI writes thirty percent of your codebase and rising, line-by-line human review can’t keep up and trust becomes the scarce resource. The challenge is building verification that runs with near-perfect precision and recall across hundreds of quality dimensions, automatically, on every change.

• What will differentiate the winners in the AI coding market over the next few years: better code generation, better governance, or something else?

Generation is commoditizing. High-fidelity output already looks more and more similar across the major models, and that gap will keep closing.

Then, differentiation moves to the quality layer: governance, verification, and the ability to enforce standards with the precision an enterprise can actually trust.

Better governance is the moat: those enterprises that succeed will be the ones that make AI-gen code safe to ship, not the ones that generated it faster.

• What should developers and engineering leaders be watching for as the AI coding market continues to consolidate?

Watch out for vendor lock-in disguised as integration. As the market consolidates, the pressure will be to buy generation and verification from the same owner, which is exactly the conflict of interest that erodes trust.

Leaders should keep their verification layer independent from whoever writes the code, protect the portability of their data and standards, and avoid betting their SDLC on a single vendor's incentives.

Motherboard deals at this level of discount don’t come around that often. Right now, Amazon has the MSI PRO X870-P WiFi at $152 (was $170) for AMD Ryzen and the MSI B760 Gaming Plus WiFi at $133 (was $200) for Intel in the Prime Day sale.

Both are well-regarded ATX boards with built-in Wi-Fi, and both represent meaningful savings if you’re mid-build or planning an upgrade. And both punch above their discounted price.

The two boards target different platforms entirely — one is AMD AM5, the other Intel LGA 1700 — so this isn’t a true apples-to-apples comparison. Think of them as two separate opportunities: one for Ryzen 7000/8000/9000 builders, and one for 12th/13th/14th Gen Intel builders.

Today's top computer deals

More on the PRO X870-P WiFi — AMD AM5

At this price, the PRO X870-P Wi-Fi sits in an interesting spot: it’s the X870 chipset — AMD’s current-gen platform with full PCIe 5.0 support — at a price that previously would have bought you a mid-range B650 board. The X870 chipset brings wider PCIe 5.0 support and Wi-Fi 7 as a platform requirement (rather than an optional add-on), and AMD has committed to AM5 support through at least 2027, which means whatever Ryzen processor you pair with this board today will have a clear upgrade path for several years.

The connectivity package at this price is genuinely impressive. Wi-Fi 7 with its 320MHz channel width delivers noticeably lower latency and higher throughput than Wi-Fi 6E — the kind of upgrade you notice on a busy home network. The 5Gbps Ethernet port, USB4 at 40Gbps, and Thunderbolt 4 are the kind of specs that usually appear on more expensive boards. Three M.2 slots (including one Gen5) gives you flexibility for fast SSDs now and room to expand later.

MSI’s PRO series is positioned more toward productivity and professional use than the flashier gaming-branded boards, which means the design is restrained — minimal RGB, clean silver heatsinks, no aggressive aesthetics. Whether that’s a positive depends entirely on what you want your build to look like. Best Buy reviewers have praised it as straightforward to install and stable from the first boot, with multiple builders noting it’s working well with everything from Ryzen 7700X to the 9800X3D.

One honest note: the PRO X870-P is tuned more for stability and accessibility than aggressive overclocking. If you’re planning to push DDR5 kits to their absolute limits or run a flagship Ryzen 9 chip at sustained all-core maximum TDP, a higher-end X870E board would give you more VRM headroom. For the majority of builds — including enthusiast setups — it handles everything without issue.

More on the B760 Gaming Plus WiFi Gaming Motherboard

The B760 Gaming Plus WiFi is the more straightforward recommendation of the two — it’s a well-established board with a substantial real-world track record. The 840+ reviews on Amazon at 4.4/5 tell a consistent story: this board works reliably, installs without drama, and delivers good performance across a wide range of Intel builds.

Intel’s LGA 1700 platform supports 12th, 13th, and 14th Gen Core processors, which means there’s a wide choice of CPUs available at various price points — everything from a budget Core i3 to the Core i9-14900K. It’s worth noting that Intel has moved on to the LGA 1851 socket for its newest Arrow Lake generation, so this platform won’t support 15th Gen CPUs. That said, for builds centered on a 13th- or 14th-Gen processor, the B760 remains a very solid foundation.

Wi-Fi 6E built-in is a practical inclusion at this price — it saves you from buying a separate wireless adapter and keeps the build clean. 2.5 Gbps Ethernet provides fast wired connectivity for those who prefer a cable. Two M.2 Gen4 slots comfortably handle primary and secondary NVMe SSDs, and PCIe 4.0 x16 is more than adequate for any current GPU.

Reviewers have specifically praised the B760 Gaming Plus WiFi for its stability across sustained workloads — one reviewer noted it handled “AAA applications and multitasking effortlessly” over ten months of use without a single stability issue. The reinforced PCIe slots and robust VRM heatsinks contribute to a build quality that feels more substantial than budget Intel boards at a similar price.

Also consider: Pre-built PC deals

• See all motherboard deals at Amazon

• Lenovo replaces traditional applications with thousands of downloadable AI functions
• Lenovo AI Host Mini ships with access to over 8000 skills
• Lenovo's proprietary marketplace sits at the centre of the experience

Lenovo has launched a mini PC which replaces much of the traditional software model with downloadable artificial intelligence functions from its own marketplace.

The company says users can access more than 8000 AI Skills instead of relying exclusively on conventional desktop applications and services.

Unlike conventional software stores operated by third parties, those AI functions are distributed through Lenovo's own proprietary marketplace integrated into the Tianxi Claw platform.

A computer built around downloadable AI functions

The AI Host mini can be configured by scanning a QR code with a smartphone before enabling additional functions through the Skills marketplace.

After completing the configuration process, users can deploy AI tools through the Skills system.

Lenovo says the platform arrives with more than 20 preinstalled Skills while allowing access to thousands of additional capabilities afterwards.

Those functions can connect with QQ, WeChat, and Feishu, allowing the system to interact with existing communication platforms and workflows.

This system feels more like an app store for AI services, although whether users prefer buying AI skills over traditional software remains uncertain.

Also, multiple Claw instances can operate simultaneously while supporting access from several users sharing one device.

Compact hardware powers Lenovo's experiment

The Lenovo AI Host Mini supports developer mode and ships with Ubuntu, although support for Windows appears uncertain at this stage.

Under the hood, the device uses the Cixin P1 CD8180, a 6nm Arm processor featuring a 12-core CPU designed for computing tasks.

Developed by China's Cixin Technology, the processor reportedly delivers performance comparable to several older high-end Intel mobile chips.

The chip also incorporates a 10-core GPU based on Arm's Immortalis G720 architecture for handling graphics and visual workloads.

Alongside the graphics hardware sits a dedicated NPU rated at 30 TOPS, contributing to Lenovo's claimed total of 45 TOPS.

Supporting those components are 8GB of LPDDR5 memory operating at 6000MT/s, together with a 256GB SSD for local storage.

Despite carrying dedicated artificial intelligence hardware, the system occupies just 0.48L of space within its compact enclosure.

The chassis measures 100 x 100 x 48.65 mm, weighs approximately 0.373 kg, and can fit easily onto crowded desks.

For connectivity, it includes two USB 3.2 ports, two USB-C ports, HDMI 1.4, DisplayPort 1.4, and integrated 2.5GbE networking.

Lenovo has opened pre-orders in China at 2,999 RMB (about $444), but this device will remain limited to the Chinese market for now.

That said, whether users eventually prefer buying AI capabilities from proprietary marketplaces instead of installing conventional software remains to be seen.

Via Videocardz

If you’re after a pre-built desktop that doesn’t compromise on anything — CPU, GPU, storage, cooling, or aesthetics — the MSI Vision ZS (9NVP-2068US) is $2790 (was $3299) at Amazon for Prime Day.

• See all Prime Day deals at Amazon.com

The $510 saving is significant, but the more compelling argument is what you’re getting: AMD’s Ryzen 7 9800X3D paired with an RTX 5070 is arguably the best CPU-GPU pairing for an all-around high-performance desktop.

The 9800X3D is AMD’s flagship desktop processor and the current benchmark leader for CPU-intensive workloads. The RTX 5070’s 12GB of GDDR7 memory and DLSS 4 Multi Frame Generation support gives the GPU side of the equation equal footing.

Top Prime Day desktop PC deal

The Ryzen 7 9800X3D is the headline spec here, and it earns that billing. Built on AMD’s Zen 5 architecture and equipped with 96MB of stacked 3D V-Cache, it consistently tops multi-threaded benchmark charts and delivers the kind of sustained performance that creators, developers, and heavy multitaskers notice immediately.

The 3D V-Cache technology — a stack of additional cache sitting directly on the processor die — dramatically reduces the latency involved in fetching data from system memory, which is where the Zen 5 generation makes its most significant generational leap over Zen 4. At 4.7GHz base and 5.2GHz boost, it’s also among the most efficient high-core-count desktop chips available.

The RTX 5070’s 12GB of GDDR7 memory operates at a significantly higher bandwidth than the GDDR6X memory found in last-generation midrange cards, and the Blackwell architecture brings meaningful improvements in ray tracing performance and AI-accelerated workloads, alongside the expected DLSS 4 Multi Frame Generation support. For creators using DaVinci Resolve, Blender, or Adobe Premiere, the combination of the 9800X3D’s multi-threaded CPU performance and the RTX 5070’s GPU compute is a genuinely capable production workstation, not just a desktop built around a single use case.

The 360mm AIO liquid cooler is a meaningful inclusion for a chip with the power envelope of the 9800X3D. Many pre-builts at this tier cut costs by using a 240mm AIO or even a tower air cooler, which limits sustained all-core performance during extended workloads. The 360mm radiator keeps the 9800X3D at optimal temperatures without the fans having to ramp to uncomfortable noise levels — MSI’s specification confirms no thermal throttling under heavy loads, which matters for anyone running rendering jobs or compiling large codebases alongside everything else.

The Vision ZS chassis is one of the more considered aesthetic choices in MSI’s pre-built lineup. The MAG PANO 120R case uses machined mesh panels across the front to maximize airflow into the system, which complements the liquid cooling rather than working against it. MSI Mystic Light RGB runs throughout the build and is controllable via MSI Center software. The result is a machine that looks like a premium desktop without the more aggressive styling that some buyers want to avoid.

The 2TB NVMe SSD is the right storage allocation for a machine at this price. Modern AAA titles regularly exceed 100GB each, and 2TB gives you room for a full game library without immediate juggling. Wi-Fi 6E covers fast wireless on the less-congested 6GHz band, and USB Type-C on the front panel handles fast data transfer and peripheral connectivity without reaching around the back.

Overall, the MSI Vision ZS puts AMD’s current top-performing desktop processor alongside a class-leading GPU in a package that’s easy to recommend for buyers who want maximum performance without the time investment of a self-build.

Also consider

• See all computers and laptops at Amazon

The iPhone will go down in history as one of the most iconic technologies that any business has introduced to the world. But not everybody, including Microsoft's former CEO Steve Ballmer, thought it would be a hit.

Disrupting the market

Steve Jobs' iconic presentation revealing the iPhone in January 2007 had an air of magic around it, capturing the attention of plenty of commentators and technologists.

But there were also plenty queuing up to label it a fad that would never take off. Steve Ballmer, Microsoft's CEO at the time, was among them.

In an interview with USA Today, a grinning Ballmer laid into the new device, criticizing its expensive pricing and the lack of a physical keyboard. This is a feature that Microsoft supported with its Windows Mobile operating system that it licensed to third-party developers like Motorola, which itself was selling a $99 Motorola Q – something that Ballmer considered a much better business alternative.

The iPhone marches on

To be fair to Ballmer, he did say in the same interview that the iPhone could end up selling really well. But he certainly failed to catch onto its appeal not only as a popular device but one that businesses would embrace.

Businesses appreciate the intuitive mobile device management (MDM) system backing iPhone in the workplace and the interoperability between Apple devices is something many also find useful. That's not to say Android devices aren't suitable for businesses — but Apple indeed tops the rankings for the highest number of shipments among enterprises.

Ballmer has since conceded that he made the wrong call – and it just goes to show how tricky predicting the future could be.

Dell has knocked a massive $680 off its Dell 16 Plus Laptop now $750 (was $1430) as part of its laptop sale. But you'll need to be quick - this limited time deal ends in a few hours (seriously, Dell even has a countdown clock running).

• See all laptop deals at Dell

The 16 Plus is a true mid-range machine aimed at business and creative professionals who need real performance. It's thin, lightweight, and even boasts a full-sized keyboard for effortlessly ploughing through work.

The US configuration ships with an Intel Core Ultra 7 256V processor, 16GB LPDDR5X memory, and a 1TB SSD. The 16-inch 2.5K IPS display with ComfortView Plus is a highlight — it's a noticeably more capable screen than you'd typically get at this price. For UK readers, the closest configuration powered by the same chip is the Dell 16 Plus with 16GB RAM and 512GB SSD for £599 (was £1199).

Today's top laptop deal

In our Dell 16 Plus review, we said: "The Dell 16 Plus offers a solid mid-tier laptop experience throughout. Between great general computing performance for any college student or executive and long-lasting battery life, the base $799 options make a ton of sense. Add a thin and lightweight design, and there’s much to appreciate here."

However, it's worth noting that the webcam and speakers were disappointing, and I'd definitely recommend upgrading them with external options. We've added a few picks to our guide on Prime Day home office deals, like the Obsbot Meet and Anker Powerconf.

Overall, though, this Dell laptop proved to be a a quality Windows machine that's thin, lightweight, and seriously portable. We badged this one with a TechRadar Recommends award.

Also consider: More Dell laptop deals

• Shop all deals at Dell

AMD has seen its Instinct GPUs continue to get traction in an increasingly competitive marketplace, as it continues to take data center market share from new and existing players and ekes out wins with gaming-centric CPUs in the consumer market.

Its most recent acquisition of MEXT, an AI-centric startup that currently deploys software that allows users to treat NAND flash as DRAM at an operating system level.

AMD says Santa Clara-based MEXT a "pioneer in AI-driven memory optimization technology".

SSD storage to DRAM for data centers?

The idea that MEXT builds on is hardly a new one, but one that it seems to have refined considerably, making it an important acquisition at a time when hyperscalers continue to struggle with limited DRAM availability, even as an even worse SSD crisis looks to be on the horizon.

MEXT's Predictive Memory is essentially a tiering engine that monitors which memory pages applications tend to access, treating regularly accessed sections as "hot" working sets kept in DRAM while offloading "cold" or less frequently accessed sections to SSDs.

This allows for a far lower performance offset than using all of one's flash memory as DRAM, with the latter being an order of magnitude faster for access, even as speed becomes a driving factor for newer chips that are increasingly memory-bound.

There is also an important economic factor at play here: DRAM is nearly 50 times more expensive than the corresponding NAND flash, making cost and scalability key considerations for most data centers looking to avoid an already expensive DRAM market that is slated to only get worse over time.

The move itself is not AMD's first foray into the storage segment, with its consumer-focused StoreMi offering essentially allowing a faster SSD to work as a cache, making up for slower drives on one's system by essentially creating a copy of files that regularly need to be loaded or accessed on the fastest possible storage solution.

Its lesser-known (and since abandoned) Radeon RAMdisk offering allows users to do the exact opposite of what MEXT is offering: creating a very fast virtual disk on existing system memory. even as enthusiasts have replicated the idea on AMD's ultra-fast 3D V-Cache tech.

AMD's purchase makes sense given how deeply embedded its hardware is expected to be in datacenters over the next decade, and one could argue that MEXT's team, which offers expertise in AI infrastructure and memory systems, could be a much more prized acquisition than the underlying technology it offers.

AI and chip talent have become increasingly difficult to lock in, with companies splurging to attract some of the biggest names in both segments, and MEXT's acquisition could help both AMD's short-term and long-term goals in the data center segment.

• ExtraHop’s Global Threat Landscape Report shows 49% of ransomware victims only detected attacks after data theft, up from 31% last year
• Average dwell time before detection is 2.5 weeks; attackers exploit encrypted channels, valid accounts, and alert fatigue to evade defenses
• Ransom payments fell from $3.6M to $2.8M, but payment frequency rose sharply, with 83% of surveyed victims paying in 2026 vs. 70% in 2025

Criminals are getting better at hiding within their victims’ infrastructure, lurking and stealing files without triggering any alarms whatsoever.

Earlier today, network detection and response experts ExtraHop released the “Global Threat Landscape Report”, based on a survey of more than 1,800 IT and security leaders worldwide. In it, it is said that roughly half (49%) of organizations that were struck by ransomware did not detect the threat until after the data was stolen.

This is up from 31% a year ago, ExtraHop stressed, showing the improvement criminals made within just 12 months.

Several factors

On average, cybercriminals have 2.5 weeks of quiet time before being spotted in ransomware incidents, the report stated. Furthermore, 14% of victims were unaware of an attack until receiving a ransom demand, which is also up from 6% a year ago.

“Prolonged dwell times often parallel a highly complex threat environment where critical alerts are obscured,” ExtraHop said in a press release shared with TechRadar Pro. The researchers uncovered several factors that led to delays in investigating critical alerts, including attackers using encrypted channels (41%), attacker activity mirroring legitimate workflows and processes (38%), using valid, high-privilege account permissions (34%), and alert fatigue (30%). Undermined baseline behavior also enabled anomalous actions to fly under the radar (27%).

The good news is that the average ransom payment dropped year-on-year, from $3.6 million down to $2.8 million. However, the bad news is that the payment frequency spiked. While in 2025 70% of respondents paid a ransom, this year 83% have done the same, at least among ExtraHop’s respondents.

When Chainalysis ran a similar survey recently, it said that in 2025 the number of successful ransomware attacks grew, while the number of payments remained relatively flat, meaning that in absolute numbers - there were fewer companies paying ransomware attackers.

• Nvidia’s closed-loop liquid cooling system virtually eliminates water waste
• Direct-to-chip cooling transfers heat more effectively than air
• It enables higher performance-per-watt and higher rack densities

Data centers aren’t without their fair share of criticism – energy-intensive compute raises temperatures and giant campuses consume considerable amounts of air and/or water to keep them running optimally.

Land scarcity and financial incentives have also been pushing new developments closer to high-risk areas, including drought-prone regions, ultimately leading to even higher cooling requirements.

But Nvidia knows this, and it knows traditional air cooling has pretty much reached its limits as AI hardware becomes denser and denser.

Closed-loop cooling virtually eliminates water waste

With cooling now a core part of AI infrastructure design, Nvidia’s latest liquid-cooled AI systems promise higher thresholds to reduce the burden, driving down water and energy consumption as a result.

By running coolant at higher temperatures – 45°C or 113°F, to be specific – it enabled simpler cooling systems and lowers electricity consumption. Nvidia’s concept uses 75% water, 25% glycol as a coolant, noting that it can run about 5-7°C higher than hot tubs.

Compared with traditional evaporative cooling towers, Nvidia’s latest proposition involves a closed-loop system where coolant continuously circulates through servers to remove excess heat from chips. The warm coolant then cycles through external dry coolers leaving virtually no water evaporation.

The company boasted that cooling-related water consumption can be reduced by as much as 100% in suitable climates subject to the occasional extreme day, with cooling towers totally eliminated.

“The Nvidia DSX reference design for AI factories has zero water consumption – we have eliminated massive amounts of power usage and pretty much all water usage,” Data Center Cooling and Infrastructure Director Ali Heydari said.

The system’s efficiency primarily comes from direct-to-chip liquid cooling, where liquid flows directly through cold plates attached to CPUs and GPUs. This captures and expels the heat from exactly where it’s produced.

Not only is this more effective than cooling entire rooms, but liquid also promises to transfer heat thousands of times more effectively than air.

Major improvements across water consumption, energy efficiency and Power Usage Effectiveness (PUE) all help on the sustainability front, but there’s another layer to the benefits.

Nvidia says it can increase rack density and performance

Nvidia acknowledged that chip power consumption and rack densities continue to rise, so by implementing liquid cooling data center companies can add more GPUs per rack, use higher rack power and ultimately pack in larger AI clusters within the same building footprint.

The company explained that its Rubin systems now fit inside two racks, instead of six, marking a major space saving.

Simultaneously, air cooling has become ineffective. “Once the watts per chip crossed a certain level, liquid cooling became mandatory,” Motivair’s CEO Richard Whitmore said.

Independent testing totally separate from Nvidia’s latest announcement shows that its H100 systems delivered around 17% higher performance when water cooled, compared with air cooling. Under sustained AI workloads, GPU temperatures fluctuated between 41-50°C when water cooled, and 54-72°C when air cooled.

Besides improving immediate and sustained performance, greater thermal efficiency could also boost longevity.

The new, higher-temperature closed-loop water cooling model is set to be used in upcoming Rubin deployments this year.

• Amazon is hiring 11,000 new workers, mostly new graduates and interns
• Garman says they're more impressionable and willing to adopt AI
• Willingness to learn could be more important than existing skills

Amazon CEO Matt Garman has confirmed the company's plans to hire 11,000 interns and new grads this year despite an ongoing internal push for AI tools.

The news comes around six months after the company warned that 16,000 workers would be losing their jobs, with Garman's cloud business hit hard.

With the latest news, the company looks to be refocusing human resources instead of getting rid of human workers altogether as business priorities change and new opportunities open up.

Amazon to hire 11,000 workers in the same year it laid off 16,000

Speaking with Casey Newton of Platformer in a YouTube interview, Garman explained that white-collar jobs are changing as a result of AI, but they're not being eliminated. He compared AI to Excel spreadsheets, which dramatically increased productivity for accounting and finance work.

"If you look at what your job was two years ago, and what your job is going to be in two years, it's going to be vastly different," he said.

Garman also noted a shift in where humans bring the most value – writing code itself is becoming less valuable, but engineers are still important reviewing AI-generated code, understanding business requirements and designing systems with the AI-generated code.

As for why the company is hiring junior workers specifically, Garman acknowledged that they're among the cheapest labor to hire, but they also learn company culture faster and can embrace AI more positively than older generations.

The CEO previously said that replacing junior employees with AI was "one of the dumbest things I've ever heard."

Without hiring graduates today, companies risk creating a long-term skills gap in the future because they have nobody to train into senior and experienced engineers, he added.

He implied that recruitment is becoming more about a willingness to learn, not the skills that have already been mastered, but the interview ultimately confirmed the continued value of human workers.

Operations tech leader Samsara has unveiled a host of new hardware and tools aimed at making fleet management more intelligent than ever, as well as boosting driver efficiency and safety.

At its Samsara Beyond 2026 event in Las Vegas, the company revealed its new 360 camera, designed specifically for heavy equipment and operations usage, making it easy to navigate around crowded sites and factories.

It also revealed an expansion to its dash cam platform which will now include two-way voice capabilities, allowing dispatchers or even AI agents to communicate easily with workers in the field.

Smarter cameras

Samsara says its new device is the first 360-degree camera built for operated equipment, and looks to address one of the most annoying pain points for workers everywhere.

On crowded sites, depots or factories, navigating huge and heavy equipment or delivery trucks can often be incredibly tricky, leading to possible safety risks and dangers for workers.

However the new camera can now provide a full 360-degree view from a single mount point, including the ability to pan and zoom even in harsh weather conditions, giving insight to any risks in real-time, and allowing safety managers to track incidents in detail.

Samsara also revealed new AI multicam capabilities which can provide a worker with a birds eye view, looking top-down on their immediate situation, allowing them to spot any tight situations, or avoid collisions with people or equipment, vital for vehicles like school buses, garbage trucks, yellow iron, and box trucks.

Elsewhere, new Rear Collision Warning and Vehicle in Blind Spot Detection tools will provide dynamic audio and visual alerts when a driver is reversing or changing lanes, in real time.

And to give an extra oversight to dispatchers and bosses, the company has added two-way voice communication through its dash cam, meaning AI and human co-workers can stay in contact whenever needed.

Say a driver has forgotten a vital piece of kit for a job site, or needs to detour to another depot to pick up additional hardware - the dispatcher can now call them directly, without the driver being distracted by engaging with a smartphone, or worrying about poor reception in remote areas. The system can also deliver AI alerts such as changing weather conditions, a lower speed limit, or parking restrictions at a site.

“By combining the power to see everything with the automation to act on it, we are shifting into the next gear on safety,” said Johan Land, Chief Product Officer at Samsara.

“The 360 Camera brings first-to-market visibility to operated equipment, AI Multicam gives road fleet drivers sharper awareness of what surrounds them, and two-way voice means the AI can respond the moment a question arises. Millions of frontline workers show up every day to keep our world running, and we are fully committed to helping get every one of them home safely.”

Getting AI-ready while building your data infrastructure is like learning to drive a manual transmission on the wrong side of the road.

It’s complicated and requires potentially dangerous multitasking.

Organizations with immature data-handling processes that are adopting AI are trying to solve multiple technology problems at once, and risk stalling out.

Unsurprisingly, 48% of enterprises cited data-related issues as their top challenge to AI adoption in NVIDIA's 2026 State of AI report.

Most enterprise AI programs don't fail because of the model or solution selected. They fail because underlying data is fragmented, inconsistent and poorly governed.

Get Your Data Foundation in Order

Enterprise data is messy in layers. It’s scattered across many systems, making it hard to pull together into a coherent picture. Even when you can consolidate it, you often will run into granularity or identifier mismatches. One application may store account numbers as plain digits, while another adds “ACCT” as a prefix. That small inconsistency creates an extra reconciliation step every time you join those data sets.

Data governance compounds the problem. Without a system intentionally designed to control who accesses data, where it moves and what protections are in place, gaps emerge fast. PII exposure is the most obvious risk: an email address that ends up in the wrong hands can trigger a serious breach. Raw, unstructured data also yields mediocre AI outputs and is more expensive to process.

Clean, structured data yields better results at lower cost. A third gap, explainability, is quickly becoming a legal requirement. Many countries and several U.S. states now require organizations to demonstrate how AI-driven decisions were reached. Cut corners on the data foundation and you may not be able to show that chain of reasoning.

At that point, you’re either in compliance violation territory or your model is producing outputs you can’t defend.

Three Steps to Get Your Data AI-Ready

Define governance before you deploy. Classify your data: what is it, where did it come from and who can touch it. Separate the roles of technical decision-making and compliance oversight. Keeping those responsibilities with different people prevents a compromising situation where the same person sets the rules and monitors compliance.

Run cross-functional AI governance as a standing function. Assign a representative from every department and meet monthly to discuss what teams are working on, what concerns have surfaced and what support they need from one another.

Approach larger AI-readiness initiatives like any other business project: assign a project manager, designate an executive owner, set a weekly cadence, build a task list and work through it.

Collect behavioral data even before you need it. The outcomes you get from AI vary enormously depending on how skilled the operator is, ranging from using it as an expensive search engine to developing autonomous workflows. Without visibility, you might be pouring money into AI licenses and getting Google-level output in return.

You don’t know who needs training, whether they have the right tool in front of them or what outcomes they’re achieving. The risk is that you make the wrong strategic call as a result—abandoning a rollout, for example, when the real fix was better training or a different tool.

Further considerations

Here’s another layer to consider. When an experienced worker completes a task, with AI assistance, they leave more skilled than when they started. The output and the learning happen together. That's what behavioral data should demonstrate over time – not just task completion, but upward skill trajectories.

When someone at the beginning of the learning curve accepts whatever AI produces without critically engaging with it, you get the output but not the growth. Behavioral data is how you catch that gap early, before it becomes a long-term cost you can't unwind.

Stay curious and look for the easy wins. Focus your data readiness efforts on the workflows where work actually happens, and prioritize tools that let you get at that data.

A recent example illustrates the payoff. A product manager ran an AI-powered analysis of quarterly bug patterns using data from the department’s most commonly used tools. The results were unexpected. One team carried a disproportionate share of incoming tickets, most of them requests for manual workarounds to a missing product feature.

While other teams split their time roughly 75% on new work and 25% on incoming bugs, that team was closer to 50-50. By not building a single feature, the organization was effectively operating 1.5 people below capacity.

The entire analysis took about 45 minutes. None of it would have been possible without data that was organized, tagged by team, connected to individual contributors, accessible via existing AI connectors and protected by role-based access controls.

The organizations that get the most from AI are the ones that empower their people to ask "I wonder if there's something here" — and have data to diagnose in an afternoon. That only happens when the foundation is already in place.

Make your data safe with the best backup software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Recent court developments in Germany have put an important issue into the robotics spotlight.

A German court in Hamburg has issued a preliminary injunction against Elite Robots Germany in a copyright infringement case involving copying of Universal Robots software.

As a result, the company is not allowed to offer or distribute the products covered by the decision in Germany while the case continues.

At first, this may sound like technical legal news only relevant for the German market.

But it highlights broader questions that matter to every company investing in automation – especially when choosing a collaborative robot that will operate close to people and become part of daily production.

Here are five reasons why.

1. Copying creates real risk for customers

When protected robot software or design is copied without permission, the impact extends well beyond the supplier and exposes all parties in the value chain to significant legal risk.

It can affect end-customers directly as using an infringing product for commercial purposes, such as a robot with infringing software in a production line, can itself constitute a legal violation.

This not only creates a risk of court-ordered remedies, including preliminary or permanent injunctions requiring the immediate shutdown and removal of the affected robots, but also exposes customers to costly and disruptive litigation and potential business interruption.

Automation systems are long‑term investments meant to run for years. Legal uncertainty at supplier level can turn into a real business risk on the factory floor.

2. “Similar” does not mean safe

Collaborative industrial robots are often described as safe, but safety is not automatic. It depends on how a robot is designed, tested, and used in real applications.

A robot that looks or behaves like another system does not share its safety profile. Safety comes from reliable hardware, validated software, certified functions, clear limits, and proper documentation. These cannot be copied by appearance alone.

Superficial similarity creates a dangerous false sense of security, which may result in serious physical injury to operators and bystanders.

3. Lower price can mean higher cost later

The purchase price of a robot is easy to compare. The long‑term cost is not.

If your robot vendor ends up in a legal battle, besides the question of even being able to use it legally, you also face uncertainty about product availability, software updates and service support.

Unexpected downtime, lack of updates or compliance challenges can quickly outweigh any initial savings. This has never been more relevant as modern robots are software‑driven machines. Motion control, force limits, diagnostics, and safety logic all depend on software.

If customers do not know where the software comes from, who owns it, or how it is maintained, they introduce uncertainty into production. Original, well‑understood software is essential for reliable and predictable operation over time.

In automation, shortcuts often appear affordable at first but expensive later. And as with all things in life: if it seems too good to be true, it probably is.

4. Buying copycat tech shapes the future of automation

Every automation investment sends a signal about what the market rewards. Choosing original, lawfully developed technology encourages long‑term engineering, robust safety practices, and continued product improvement. Choosing copycat technology does the opposite: it normalizes shortcuts, weakens incentives to invest in research and compliance, and shifts competition away from quality and reliability.

Over time, widespread tolerance of intellectual property infringement affects the entire robotics ecosystem, from suppliers and integrators to suppliers and regulators. It increases uncertainty and ultimately makes it harder for manufacturers to rely on stable platforms that will be supported and improved for years to come.

Protecting original technology is not about limiting choice or slowing competition. It is about ensuring that competition is based on real innovation, verified safety, and accountability – and that customers can invest in automation with confidence, knowing the technology they rely on is built to last.

5. Trust in the original

In summary, choosing a robot is not only about specifications and price. It is about trust.

Buyers need confidence that a robot is legally sound, properly certified, and supported by people who truly understand the technology. Trust comes from transparency, responsibility, and deep technical knowledge – not from claims or visual similarity.

Automation is becoming increasingly central to modern manufacturing and as it does, questions of originality, safety, and integrity become part of responsible decision‑making.

So, no matter if you’re purchasing your first robots or expanding your fleet, before asking what a robot can do, it is worth asking a simple question:

Do you trust where it comes from?

I tried 70+ best AI tools.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Although the technology powering them has evolved in performance and efficiency, the core principle of data centers hasn’t changed much in the past two decades.

Every new generation of infrastructure has been designed to shorten the distance to users, reducing latency and increasing reliability. However AI is putting serious strain on that model, and unprecedented demand for data centers is making companies think outside the box.

Terrestrial projects now face challenges from all directions, including land availability, cooling, power and water consumption, and local opposition.

Step in orbital data centers. Some pioneers are now looking to rocket data centers up into space, and the model keeps shifting further from concept and closer to reality.

While enormous technical hurdles remain, advocates argue orbital data centers could eventually complement terrestrial campuses by shifting some of the most energy-intensive AI workloads away from land-based constraints.

Networking is what will define good orbital compute

Some of the benefits include virtually unlimited space to expand, uninterrupted access to green solar energy and cooler environments that don’t require so much cooling.

But while sending compute into space tackles some of the problems we face today, it reverses one of the biggest fixes we’ve spent years improving – latency.

In today’s instant economy, a data center has no use if it can’t exchange information quickly and reliably. Terrestrial networks lean heavily on fiber networks and physical connections, but orbital data centers would need to improve wireless transmission across hundreds of miles.

As work on this progresses, Internet exchange operator DE-CIX CEO Ivo Ivanov says orbital compute should be viewed as another layer in our increasingly distributed digital ecosystem – not an outright replacement for terrestrial compute.

To better understand whether orbital data centers are just another ambitious experiment or whether they could be the next evolution of digital infrastructure, I spoke with Ivanov about the industry’s biggest challenges.

• You noted that SpaceX's IPO has brought the idea of AI data centers in space into the mainstream. Setting the hype aside, how close is this to real, and what's actually driving the biggest companies in the world to look at space?

The idea of placing data centers in orbit is the kind of thing that immediately captures people's imagination. It belongs in the same category of technological ambition as reusable rockets, commercial spaceflight, or landing on the Moon.

For decades it felt like something that belonged in science fiction. Today, for the first time, it's being discussed as a serious infrastructure proposition. But I don’t think the real story here is about space – it’s about AI, and how it’s altering our global trajectory.

Over the last few years, we've seen an extraordinary increase in demand for compute capacity. AI training clusters are growing larger, power requirements are rising, and in many regions the availability of energy, land, and cooling has become a genuine constraint.

These constraints are what drive innovation, and that’s what we’re seeing play out here.

In theory, space offers access to continuous solar energy and room to scale without the physical limitations we experience here on Earth. Having said that, we're still at a very early stage.

I don't think we're talking about replacing terrestrial infrastructure anytime soon. What we're seeing today is the beginning of a serious exploration into how orbital resources could complement the digital infrastructure ecosystem that already exists.

There is still work to be done, particularly around networking – putting an AI data center into orbit would be a milestone achievement, but it won’t count for much if we have to truck hard drives into space to do any training. For now, all anybody can say with confidence is that it's becoming part of a broader conversation about how society will support the next generation of AI workloads.

• What problems could orbital data centers potentially solve that terrestrial infrastructure struggles with today?

I think the broader answer is that orbital infrastructure has the potential to ease some of the physical constraints that have shaped and limited digital infrastructure for decades. We can’t simply keep building more data centers – they need access to land, power, cooling, transport links, and of course, connectivity.

As AI continues to grow, some of those resources will become harder to secure, especially at the pace the industry wants to move.

That's why the industry is starting to look at non-terrestrial alternatives. Suddenly, you're talking about access to potentially unlimited solar energy, plenty of space, and the opportunity to scale at a pace that matches our ambitions for AI.

But putting data centers into orbit, as impressive as it would be, is only the first step. Google’s Project Suncatcher is already exploring how power might be harvested from the sun, while the European Space Agency’s OFELIAS project is exploring how optical feeder links between Earth and orbit can be optimized to reduce latency and disruption. There are lots of moving parts.

• From a networking perspective, what challenges will emerge once data centers move into orbit?

The moment you put a data center into orbit, networking becomes the make-or-break factor. A data center on Earth can usually rely on a dense ecosystem of fiber routes, Internet Exchanges, cloud on-ramps, and interconnection facilities. In orbit, that ecosystem doesn't exist yet.

Every workload, every application, and every AI model depends on data moving between different locations, and suddenly those journeys become much more complex.

A lot of the discussion I’m hearing focuses on bandwidth or speed, but I think predictability is the more interesting challenge. AI systems need data to arrive quickly, but they also need it to arrive consistently.

Optical and laser-based communications have enormous potential, but they also introduce new variables like cloud cover, atmospheric turbulence, satellite handovers, and changing orbital positions that can all affect how data moves between Earth and space.

That's why I believe the real challenge isn't simply connecting an orbital data center to the ground, but creating an interconnection layer that makes orbital, terrestrial, cloud, and edge infrastructure behave as though they're part of the same ecosystem.

That's a much bigger networking challenge than people realize.

• Elon Musk said orbital infrastructure is "only a few milliseconds away." How much could those milliseconds matter, and is latency the biggest challenge?

Those milliseconds absolutely matter. AI systems are fundamentally data-driven, and every additional moment spent moving information between users, applications, models, and infrastructure affects performance and productivity.

That's one of the reasons the industry is investing so heavily in edge computing – the closer you can bring compute resources to the data, applications, and users they serve, the better the experience tends to be.

Certainly, low-Earth orbit is only milliseconds away, but for latency-sensitive use cases, the 20-40 ms it takes to reach the stratosphere will provide a significant challenge for some AI inference use cases. So not all applications can possibly be served from hundreds of kilometers above the ground.

Having said that, performance isn't just a question of latency. A network that consistently delivers twenty-five milliseconds of latency is often more valuable for many enterprise use-cases than one that fluctuates between twenty and a hundred.

Physics determines how quickly data can travel, but factors such as atmospheric conditions, satellite handovers, routing decisions, and network architecture all influence how predictable that experience is.

• If a data center is in orbit and the AI workloads it serves are on Earth, what actually has to happen for those two worlds to behave as one network?

The way I like to think about it is that users shouldn't have to think about where a workload is running. Whether an application is being served from a terrestrial data center, an edge location, a cloud region, or one day an orbital platform, the experience should feel seamless.

Achieving that is much harder than it sounds, because data has to move continuously between different environments, and those environments need to behave as though they're part of the same network, even when they're separated by hundreds or thousands of miles and, increasingly, by the boundary between Earth and space.

That's one of the reasons projects like the ESA's OFELIAS initiative, which I mentioned earlier, are so important. As part of that initiative, DE-CIX is working alongside partners including the German Aerospace Center to explore how optical feeder links between satellites and the ground can become more stable, efficient, and predictable.

Rather than simply building more connections, we need to create an infrastructure foundation that allows terrestrial, satellite, and orbital resources to work together as a single ecosystem.

We've spent decades building that kind of seamless interconnection on Earth, and now we need to take what we’ve learned and combine it with new technologies to bring non-terrestrial infrastructure into the picture.

• Where does this leave terrestrial data centers and interconnection? Will orbital compute compete with what exists on the ground, or depend on it?

I don't see this as a competition at all. Digital infrastructure has a long history of adding new layers rather than replacing existing ones. Cloud computing didn't eliminate enterprise data centers, edge computing hasn't eliminated centralized clouds, and satellite connectivity hasn't replaced terrestrial networks.

Instead, each technology tends to find the role it's best suited to, and the overall ecosystem becomes more capable as a result.

I think orbital compute will follow a similar path. Some AI workloads may benefit from being processed closer to abundant energy sources in orbit, while others will remain in terrestrial edge data centers because they require ultra-low latency inference, regulatory compliance, or proximity to users and applications.

It’s not about choosing one over the other, but creating an architecture where workloads can be placed wherever they make the most sense and where data can move seamlessly between them.

• What should the industry be doing now to prepare networks for orbital compute?

I think the first step is recognizing that orbital compute is more than a compute challenge. It's a networking one. The industry has spent decades building highly interconnected ecosystems on Earth, where data can move efficiently between enterprises, cloud providers, networks, content platforms, and users.

If compute begins expanding into orbit, we'll need to ensure the same level of interconnection exists between terrestrial, satellite, and orbital environments. Advances in satellite communications, laser-based networking, and distributed infrastructure are creating new possibilities that didn't exist a decade ago.

What I'd like to see now is greater collaboration between network operators, cloud providers, infrastructure companies, and space organizations.

The technologies themselves are advancing rapidly, but the challenge is ensuring they evolve as part of a coherent ecosystem rather than a collection of isolated systems.

• Looking ten years ahead, what does a fully interconnected ecosystem spanning terrestrial, edge, cloud, satellite, and orbital infrastructure actually look like? What technologies will make it work?

The most interesting thing about the future is that people probably won't think about infrastructure at all. Today, we spend a lot of time talking about whether something runs in a cloud region, an edge location, a data center, or on a particular network.

Ten years from now, if we succeed with orbital infrastructure, workloads will simply move to wherever they can be processed most efficiently, and data will flow across terrestrial, satellite, and orbital environments as naturally as it moves across continents today.

When we look back at the great infrastructure achievements of the past, whether it's railways, electricity grids, undersea cables, or the Internet itself, their success wasn't measured by the technology, but by how completely they disappeared into everyday life.

People stopped thinking about the infrastructure and started focusing on what it enabled. I think we'll see something similar here. The future of digital infrastructure will be defined by our ability to connect every layer of the ecosystem into something that feels seamless.

That's when we'll know we've succeeded.

• AI infrastructure spending has broken the global memory supply chain
• DDR2 prices are set to surge 60% in Q2 2026 alone
• Some manufacturers are redesigning DDR3 systems around DDR2 to secure components

The AI boom is creating unexpected consequences across the memory industry, including renewed demand for technologies introduced more than two decades ago.

As manufacturers prioritize advanced products serving AI infrastructure, older memory generations are becoming increasingly difficult to source in meaningful quantities.

That shift is now pushing DDR2 prices sharply higher despite the technology first appearing around 25 years ago.

Legacy memory demand rises as newer DRAM becomes scarce

DRAM shortages have pushed memory prices sharply higher, with DDR4 contract pricing reportedly surging by as much as 2200% at some point.

Industry executives such as Nothing CEO Carl Pei and Framework have warned that supply constraints and elevated pricing are unlikely to ease in the near term.

According to TrendForce, these structural changes in the DRAM market are forcing buyers to move backward through successive memory generations, and limited mature-node supply is pushing some brands to switch from newer memory to older ones that are available in larger volumes.

TrendForce estimates that DDR2 contract prices will increase by roughly 55% to 60% during the second quarter of 2026.

Prices are then expected to climb another 35% to 40% in Q3, creating one of the strongest pricing surges seen in the legacy memory segment for years.

The situation stems partly from decisions made by the world's largest DRAM suppliers, which continue directing manufacturing resources toward advanced memory technologies.

Growing demand for HBM and server DRAM, both closely linked to expanding AI infrastructure investments, has reduced wafer allocations available for older consumer-focused products.

As availability tightens, companies purchasing DDR4 memory have increasingly turned to Taiwanese suppliers for additional capacity.

Demand has therefore spread across several generations of products, creating pressure that now extends into DDR3 and DDR2 markets.

Some manufacturers have reportedly replaced DDR4 designs with DDR3 solutions, while certain DDR3-based systems are being redesigned around DDR2 memory to improve component availability.

Suppliers gain leverage as production shifts toward higher-margin products

The imbalance between demand and available output has strengthened the negotiating position of some memory manufacturers.

With supply unable to keep pace, vendors have gained greater flexibility to raise contract prices while concentrating on products that generate stronger returns.

Winbond, one of the key suppliers of DDR2, is reducing the production of older standards and redirecting the capacity toward DDR3, DDR4, and LPDDR4 products.

Such a transition is expected to place additional strain on DDR2 availability during the coming quarters.

As production contracts, buyers dependent on legacy platforms could face increasing competition for a shrinking pool of components.

Manufacturers like Elite Semiconductor Memory Technology (ESMT), however, are increasing their focus on DDR2 production within their existing wafer allocation.

The company hopes to capture the surging demand in an effort to improve profitability while helping address part of the supply shortfall from Winbond's withdrawal.

Whether these supply conditions represent a temporary market distortion or the beginning of a prolonged shortage remains uncertain.

What is becoming clear is that AI-driven demand for advanced memory is now affecting even very old product generations, creating ripple effects that few expected to reach DDR2.

• GEEKOM GeekBook M16 launches at $799 with MacBook-style design
• Core Ultra 9 power pushes sustained 55W performance in a thin chassis
• A 99.9Wh battery claims up to 24 hours of usage on paper

GEEKOM's GeekBook M16 is not trying to hide what it is inspired by — the chassis, the wallpaper, the ultra-thin bezels, and the centered webcam all speak directly to Apple's M3 MacBook Pro design.

The 16-inch laptop is built around Intel's Core Ultra 9 185H processor running 16 cores, 22 threads, and a 5.1GHz turbo boost, with GEEKOM's proprietary Dynamic Tuning Technology keeping it at a sustained 55W under load.

At $799, this notebook is an aggressive entry into a market that premium laptop makers have long treated as their own territory.

Hardware hosts a mix of strengths and tradeoffs

GEEKOM’s GeekBook M16 uses a 16-inch display with a 1920 x 1200 resolution in a 16:10 aspect ratio, reaching a 90% screen-to-body ratio, which helps give the panel a more modern and immersive look on a desk.

However, the 60Hz refresh rate stands out as a clear limitation, especially for users accustomed to smoother high-refresh displays.

The business laptop offers two storage configurations, including a PCIe Gen 4 2280 slot supporting up to 4TB and a PCIe Gen 4 2230 slot supporting up to 2TB.

The system comes with 16GB of LPDDR5 memory and Intel Arc integrated graphics for general compute and light creative workloads.

For connectivity, it includes a 40Gbps USB4 port, a USB-C 3.2 Gen 2 port, two USB-A 3.2 ports, and an HDMI 2.0 output.

Wireless support is handled through Wi-Fi 6E and Bluetooth 5.4, while a 2MP 1080P webcam includes a physical privacy switch for added control.

On the security side, the laptop integrates TPM 2.0 and a fingerprint reader in the power button for system and biometric protection.

The device ships with Windows 11 Pro pre-installed, alongside confirmed Linux compatibility for users who prefer alternative operating systems.

Power and endurance in one chassis

The battery is where GEEKOM makes its most attention-grabbing claim, and the numbers are genuinely difficult to ignore at this price point.

The GeekBook M16 carries a 99.9Wh cell — the absolute maximum battery capacity that commercial airlines permit passengers to bring on flights.

GEEKOM's intelligent power management technology claims up to 24 hours of Wi-Fi usage and approximately 20 hours of local video playback from a single charge.

A 100W GaN USB-C fast charger ships in the box, with the company claiming a 10-minute charge restores around 3 hours of productive runtime.

Thermal management falls to the IceBlade 3.0 cooling system, which uses dual fans and dual heat pipes working together efficiently.

Compared to its predecessor, the system increases air pressure by 50% and total airflow by 28%, keeping temperatures stable during extended and intensive workloads.

The CNC-machined aerospace-grade aluminium chassis measures just 0.66 inches thin and weighs a reasonably portable 1.73kg across its full 355 x 250mm footprint.

Whether real-world battery and performance results match GEEKOM's claims is something only rigorous independent testing will ultimately confirm.

Preorders for the GeekBook M16 are now live on GEEKOM’s store with a $100 discount code (GKM16), bringing the price down to $799, while shipping is scheduled for late June.

• As AI data centers consume an ever-increasing amount of RAM chips, researchers are looking for cheaper, more scalable alternatives
• imec research seen by many to be the answer, even as engineering challenges continue to exist
• Two published studies using ferroelectric materials show capacitors essentially able to replicate modern DRAM functionality

It is no secret that AI has affected much of the world, with many welcoming the advent of a "smart" digital assistant, with others using it to speed up the more mundane tasks that previously required oversight.

On the flip side of the equation, some finding their jobs redundant, others being forced to upskill or pivot to different industries just to eke out a living in 2026.

There is however, one area which has been significantly impacted where the effect might have been understated so far: the PC hardware industry.

DRAM for data centers first

The reason the PC hardware industry (and, to a lesser degree, the smartphone and game console industries), or any industry that uses fast RAM and NAND flash for its core functionality, is affected is overwhelming demand in data centers.

With billions of dollars in buildouts and procurement budgets, the average consumer is unlikely to compete for memory with a data center being built by hyperscalers and with good reason: there is simply too much money at stake here.

The problem for data centers, however, is a different one: despite consuming nearly 70% of all memory produced in 2026, resulting in the worst memory supply crunch in the last 15 years, things are projected to only get worse, even as demand continues to grow unabated.

With memory giant Micron telling consumers to wait till 2028 before they can expect any relief in their segment, and SSDs heading down the same path, thanks to a similar situation when it comes to high-density NAND flash-based storage, AI data centers are also looking for a memory breakthrough that could allow for cheaper, faster memory in the near future to better handle their needs.

European chip research lab imec seems to be working on an answer, however. At the 2026 IEEE/JSAP symposium on VLSI Technology & Circuits, the event showcased two major advances that could enable new ferroelectric memory to become mainstream over the next decade, as it attempts to address the AI-centric memory shortage with a solution that better caters to data center consumers.

At the heart of the current breakthroughs is a ferroelectric capacitor that operates at low voltage, enabling a large number of write cycles and holding charge well, making it a potentially viable replacement for traditional DRAM-based memory.

The second is a transistor that they have managed to stack vertically, enabling a denser NAND-flash-style storage design while adding a back-gate modification to fix its memory-erasure issues.

FeRAM is not a new concept; first conceived in 1952, it holds massive promise, even as most researchers concur. It had, until recently, limited interest, but a world spending billions on AI data centers, with memory and storage increasingly scarce resources, has brought it back to the forefront, even if most of its gains still keep it limited to lab conditions.

“This work shows how imec’s multidisciplinary expertise, from materials science to advanced 3D integration, enables us to tackle some of the most pressing challenges in memory technology," noted program director at imec, Maarten Rosmeulen. "We are exploring multiple paths toward the memory solutions that will be required to sustain the rapid growth of AI and data-intensive applications."

It must be noted that imec does not operate in a vacuum; It shares its research with hundreds of industry partners, including chip designer Nvidia, photolithography systems manufacturer ASML, and fab giant TSMC.

Other industrial partners include Intel, Samsung, Micron, Qualcomm, AMD, and Apple, indicating that its research could eventually be integrated into future memory offerings by multiple interested entities, including other hyperscalers.

Should imec's research solve the modern datacenter conundrum by offering a cheaper, denser memory module using different materials, it could very well spark a new AI-centric war for high-end storage and memory.

But for now, it seems to require a considerable amount of time before it is production-ready, with imec admitting it was "addressing remaining challenges" while reminding interested parties that the entire breakthrough was still a proof of concept and still in the research stage.

The CEO of Samsara has hailed the benefits of AI in helping modernize and boost efficiency and productivity in the operations industry.

Speaking at the opening keynote of Samsara Beyond 2026 conference in Las Vegas, Sanjit Biswas laid out how the company is, “working together to build the future of operations”.

“These are industries that you can't read about in a book, or just visit a website…you have to put on your boots and go onsite,” he noted.

“Automate away the grind”

Biswas noted how Samsara’s customers are all facing the same challenges, namely the rising cost of fuel, caused by the war in the Middle East, put also the rising cost of doing business across the board, with rising maintenance costs also biting hard.

But technology was also proving a challenge in these more “hands-on” industries, he noted, with too many reports and disparate systems in conflict, leading to “data overload” as too many apps and distractions actually just annoy and frustrate workers.

To combat this, Biswas set out his idea of “automate away the grind” - using technology to give workers time back to focus on the really important tasks.

“We're entering a whole new age - the age of intelligence," he said, noting how Samsara wants to be, “taking operations to the next level”.

“We can collect tons and tones of data in the field,” he added, “and now we can use AI to find insights, and figure out what’s really going on…AI can now formulate plans, figure out tools, and really help you run your operation to take action on your behalf.”

Biswas laid out a host of useful examples where AI could benefit frontline workers, such as automatically adjusting vehicle settings during extreme weather situations, monitoring usage of tools and equipment to help predict when maintenance is needed, and giving drivers more information on delivery situations/environments such as a new drop-off location.

To back this up, Samsara announced its new Agent Studio, specifically designed for companies in the physical operations sector, which lets customers utilize a range of pre-configured agents or build their own from scratch.

The company says that going forward, tasks such as managing paperwork, communicating with drivers, and working with vendors can now be automated with agents in minutes, freeing staff from hours of manual work each week

“Reducing risk for the frontline, increasing the efficiency of the operation, and really helping you digitally transform with this technology - that’s what we’re excited about today," Biswas concluded.