The Illusion of Scarcity: How Big Tech Overbuilt the AI Cloud

The structural narrative that has fueled the AI bull market—the absolute belief that AI computing capacity is a scarce, bottlenecked commodity—has officially shattered.

Microsoft’s sudden, systematic unwinding of its bulk infrastructure leases, paired with a tsunami of independent supply injections, marks a permanent macroeconomic transition. We have exited the era of structural compute scarcity and entered a phase of structural oversupply. Compute is no longer digital gold; it is becoming a highly localized, rapidly depreciating utility. Microsoft’s maneuvers demonstrate that hyperscalers now hold all the leverage, exposing the fatal flaw in the third-party infrastructure boom and triggering a painful reality check for the entire AI ecosystem.

The Modern Supply Shock: Flooding the Spot Market

The primary catalyst obliterating the shortage narrative is an uncoordinated, massive injection of raw GPU capacity directly into the spot market inventory. Two mega-entities are driving this sudden glut:

• The Oracle Liquidations: Microsoft didn't just walk away from upstart operators—it turned off an estimated 150 to 200 megawatts (MW) of unified AI clusters previously rented from Oracle. Originally leased in 2023 to power the AI-assisted Bing search engine, Microsoft declined to renew these multi-billion-dollar infrastructure contracts. Because the hardware relies on older-generation Nvidia graphics chips that no longer meet Microsoft’s evolving technical specifications, the tech giant simply began turning off the servers. Oracle is now forced to dismantle these massive superclusters, dumping the underlying hardware back onto the market as smaller, single-GPU virtual machine shapes.

• The SpaceX Tsunami: Operating via its xAI infrastructure division, Elon Musk’s SpaceX Cloud is aggressively expanding its footprint, targeting an unprecedented 2.5 to 3 gigawatts (GW) total online(1.2gw later this year). To put this in perspective, SpaceX’s near-term capacity additions alone represent four times the existing hourly available US cloud inventory. By deploying modular infrastructure alongside massive arrays of Tesla Megapack batteries directly behind the meter, SpaceX is completely bypassing traditional utility grid connection queues, dropping gigawatts of raw compute onto a market that cannot digest it.

The Execution Bloodbath: Breaking the Back of the Victims

To see the destruction left in the wake of Microsoft's pivot toward self-reliant, first-party data centers, one only needs to look at the collateral damage inflicted on its external partners. Microsoft committed over $80 billion to rent capacity from outside cloud providers. Now, under the direction of Chief Financial Officer Amy Hood, Microsoft is aggressively enforcing rigid contract milestones and walking away from any project carrying operational or regulatory risk.

Nscale has borne the most brutal brunt of this enforcement. Microsoft completely walked away from a signed deal to anchor Nscale's flagship data center in Loughton, Essex, after local grid operators pushed power availability into the 2030s—instantly deleting $3 billion from Nscale’s long-term order book. Worse still, Microsoft refused to exercise an optional contract clause to more than double its existing 240 MW footprint in Ward County, Texas, walking away from a planned 500 MW expansion due to prolonged regulatory energy delays. For an upstart targeting a New York IPO, these dropped options chopped Nscale's prospective valuation in half, cratering it from $35 billion down to an investor-estimated $16 billion. Microsoft also walked away from a separate 1.35 GW letter of intent with Nscale in West Virginia.

Nebius Group is facing a similarly perilous reckoning at its critical expansion site in Vineland, New Jersey. The project—operating via data center developer DataOne—was hit with a $1 million environmental fine for illegally running 45 industrial gas generators without local air permits. The regulatory logjam has threatened to push roughly 140 MW of planned Microsoft capacity deep into 2027. Microsoft didn't pay a penny in penalties; instead, it dispatched its own internal compliance engineers to police the site. To save its $17.4 billion anchor lease from being cancelled entirely, Nebius is now burning its own cash reserves to execute an incredibly expensive emergency pivot, replacing the unpermitted gas engines with a clean-energy microgrid powered by Bloom Energy fuel cells.

In sharp contrast to the absolute project deletions hitting Nscale, IREN Limited represents the upper tier of the infrastructure market—a fundamentally viable operator navigating a high-friction execution phase. IREN’s core multi-billion-dollar footprint at its Childress, Texas campus remains on solid footing, and the capacity is highly likely to be successfully consumed by Microsoft as it goes live in early 2027. Yet, despite an impending successful launch, IREN has become a textbook case of how hyperscalers are using their immense negotiating leverage to squeeze vendor timelines and trigger capital crunches.

The delay that pushed IREN’s revenue start from late 2026 into early 2027 was driven entirely by Microsoft’s tightening operational noose. Microsoft essentially forced an eleventh-hour overhaul of the site’s networking architecture, requiring custom fiber-optic reconfigurations to handle the extreme demands of next-generation frontier models. Because Microsoft holds all the cards, IREN had no choice but to absorb the timeline slippage. While a delay of a few months is minor for a project of this scale, it creates an intense financing squeeze in a hyper-competitive capital environment.

IREN is staring down a massive $15 billion capital expenditure plan over the remainder of the fiscal year. While cash reserves and hefty upfront customer prepayments secure $9 billion of that runway, a glaring $6 billion funding gap remains completely unfinanced. In an environment where a massive net debt load ($4.4 billion) meets delayed incoming revenue streams, even a brief pushback on cash generation forces Wall Street to ask painful questions about impending shareholder dilution. IREN proves that even when a neocloud project is destined for success, the sheer weight of the capital required to survive a hyperscaler’s shifting engineering whims turns data center development into a dangerous game of financial endurance.

This systematic withholding of capital draws directly explains why Microsoft’s latest quarterly capital expenditures missed Wall Street forecasts by a staggering $5 billion, and why the company cut its annual investment projection to $170 billion.

The Enterprise Rebellion:** Ditching Closed APIs for Private Sovereignty**

The collapse in demand is not just a mathematical anomaly; it is an architectural rebellion. The industry narrative insists that Fortune 500 companies are locked into a permanent race to buy expensive API tokens from closed frontier labs like OpenAI and Anthropic. The reality on the ground proves otherwise.

Procurement data reveals that half of the Fortune 500 now runs on open-source AI architecture via secure model hubs like Hugging Face. Enterprises are actively rejecting expensive, volatile hyperscaler API pipelines and closed systems because of unsustainable cost escalators and strict data residency mandates. Instead of feeding the frontier cash-burn loop, major institutions are pulling highly capable, open-weight models into private, ensured cloud VMs, self-hosting their entire AI stack at a fraction of the cost.

• Capital One & JPMorgan Chase: Standardized their developer workflows on open-source PyTorch environments, migrating heavily toward private enterprise platforms like Nvidia AI Enterprise (NIM microservices) to run fine-tuned, localized language models entirely within their own corporate firewalls.

• Wells Fargo & IBM: Completely bypassed public frontier APIs for high-volume internal agent workloads. Several tier-1 banks and enterprise consulting giants have successfully migrated document review, HR automation, and compliance auditing entirely to self-hosted open models, slashing their monthly inference spend by more than half in the very first quarter post-migration.

• Revolut: Europe’s fintech giant delivered the ultimate blow to the closed-source business model. Revolut slashed its AI coding costs by a staggering 90% by completely abandoning OpenAI and Anthropic APIs. They purchased their own private cluster of 56 Nvidia B200 GPUs, deployed a self-hosted, open-weight model (Zhipu AI's GLM-4.5), and brought their internal inference costs down to a microscopic $0.30 per million tokens while simultaneously improving software delivery speeds by 70%.

The Math Behind the Spot Bloodbath

When supply outstrips demand, the laws of economics dictate that pricing collapses. In the short-term spot market, that collapse is already well underway.

To evaluate the structural damage, look at the physical unit economics of the hardware. A standard top-tier GPU box carries a five-year, straight-line depreciation cost (the physical cost of manufacturing, deployment, and power divided by total operational hours) sitting between $1.80 and $2.20 per hour per GPU.

Yet today, neoclouds are bleeding because spot rates for H200s and B200s have collapsed well below that depreciation floor. For highly leveraged providers like CoreWeave—whose entire business model relied on absorbing short-term losses for five years in the hope that permanent scarcity would let them price-gouge hyperscalers on renewal—this structural inversion is fatal. Compute has effectively become a commoditized commodity utility, plagued by the same boom-and-bust cycles seen in traditional shipping containers or DRAM memory chips.

The Downstream Nvidia Whiplash and the Hyperscaler Escape

This infrastructure unraveling triggers an immediate, painful whiplash for Nvidia, which had confidently baked these insatiable neocloud allocations into its long-term revenue forecasts. Nvidia is now caught in a structural pincer that dismantles its valuation narrative:

• The Cannibalization Effect: When Microsoft cancels an order or lets an Oracle lease lapse, those hundreds of thousands of GPUs do not evaporate. Instead, as Oracle, Nebius, and Nscale scramble to re-lease their pre-allocated hardware to secondary enterprise clients, they enter into direct, aggressive competition with Nvidia's own primary sales pipeline.

• The Death of Pricing Power: With spot rental recovery rates crashing below physical depreciation costs, the narrative that an Nvidia GPU is a bulletproof, high-margin, appreciating investment is dead. The market value of older—and even recent—GPU generations is being grossly reduced, signaling to enterprise buyers that waiting out the market yields massive cost discounts.

Most critically, Microsoft’s actions mirror a broader, coordinated escape velocity being achieved by all major hyperscalers. Google, Amazon, Meta, and Microsoft are aggressively scaling their own custom, in-house Silicon—such as Microsoft’s custom Maia 300 AI chips.

By deploying their own direct-cost Tensor Processing Units (TPUs) and Application-Specific Integrated Circuits (ASICs) across their hyperscale campuses, these tech giants bypass Nvidia’s extortionate 75% gross margins entirely for high-volume inference workloads. First-party silicon allows hyperscalers to achieve massive token-per-watt efficiencies at a fraction of the manufacturing cost. Armed with their own proprietary hardware ecosystems, hyperscalers are realizing they no longer have any structural reliance on renewing external rental contracts—leaving the heavily indebted neocloud builders holding the bag in a flooded market.

title: "Qualcomm readies phones for Wi-Fi 8, Bluetooth 7, and faster 5G era" url: "https://www.digitaltrends.com/phones/qualcomm-readies-phones-for-wi-fi-8-bluetooth-7-and-faster-5g-era/" author: "Manisha Priyadarshini"

Qualcomm is preparing the next wave of smartphones and connected devices for faster wireless technologies, announcing new platforms that bring Wi-Fi 8, Bluetooth 7, and 5G Advanced closer to commercial reality.

At MWC 2026, the company unveiled its FastConnect 8800 mobile connectivity system alongside a new lineup of Dragonwing Wi-Fi 8 networking platforms, while separately introducing the Qualcomm X105 5G Modem-RF system aimed at powering the next phase of cellular networks.

Together, these announcements highlight how Qualcomm is evolving device connectivity to handle higher speeds, longer wireless range, and increasingly AI-driven network workloads.

Wi-Fi 8: faster speeds and AI-optimized networks

Qualcomm

Qualcomm’s Wi-Fi 8 push is led by the FastConnect 8800 Mobile Connectivity System, which the company says is the first mobile solution with a 4×4 Wi-Fi radio configuration capable of delivering peak wireless speeds above 10 Gbps.

Qualcomm claims the new platform can offer up to twice the performance of Wi-Fi 7 and extend gigabit wireless range by as much as three times. The chip combines multiple wireless technologies in a single solution, supporting Wi-Fi 8, Bluetooth 7.0, Ultra Wideband (UWB), and Thread 1.5.

Qualcomm

Qualcomm also introduced Proximity AI, which uses Wi-Fi ranging, Bluetooth channel sounding, and UWB sensing to detect nearby devices with centimeter-level accuracy.

Bluetooth 7: faster data transfers

FastConnect 8800 also introduces Bluetooth 7 support with Bluetooth High Data Throughput (HDT) technology.

Qualcomm says this increases Bluetooth data speeds from 2 Mbps to 7.5 Mbps, enabling faster wireless file transfers and more demanding applications such as high-quality audio streaming and spatial computing experiences.

Bluetooth also works alongside Wi-Fi and UWB in Qualcomm’s proximity sensing system, allowing devices to locate nearby objects or accessories more precisely.

5G Advanced: Qualcomm’s next modem

Qualcomm

On the cellular side, Qualcomm introduced the X105 5G Modem-RF system, designed for the upcoming 5G Advanced phase of mobile networks. The modem supports peak download speeds of up to 14.8 Gbps and uploads of up to 4.2 Gbps, while supporting the latest cellular standards.

The modem also integrates Qualcomm’s fifth-generation AI processor for wireless connectivity, which analyzes network conditions in real-time to improve signal performance, reliability, and power efficiency.

It also supports satellite connectivity through non-terrestrial networks, enabling devices to maintain connectivity when traditional cellular coverage is unavailable.

Qualcomm also announced its Snapdragon Wear Elite platform for next-generation AI wearables at MWC 2026. Commercial devices using these new connectivity platforms are expected to begin appearing from late 2026, as manufacturers adopt the next generation of wireless standards.

Today's AI sector market crash is not about Democrat victory.

ANThropic is getting told to value their IPO at $1T. Down from $2T initial hopium.

In Q3, when backing in capitalized training costs, they lose 33c for every $1 of revenue.

The bigger problem for any hopium is that they bragged about a $65B ARR in July, but the full quarter (Q3) was only $16.8B in revenue, which is flat $67B ARR. Their amended S1 filing also chose to leave blank the October revenue/ARR number despite them knowing it on Nov 2 amendment date.

since Sept 22, ANT released opus, sonnet, haiku 5.5. All well received high performance/value models that are decompiling/cloning the world. This plan was supposed to juice revenue numbers because new models always get more usage, if only to test how $60 can make a gta 0.1 clone. The July ARR boasting came off Fable, sonnet, claude 5.0 release schedules, and peak tokenmaxing mantra.

ANT revenue stagnating, and $1T valuation impacts previous OpenAI related bubble pop post, in that OpenAI's fundraising at $1.25T valuation was based on it having that ratio compared to ANT at $2T. OpenAI cannot raise money at fair ~$650b because that is a down round, and previous investors get to take more of the company for free.

OpenSSH built against musl libc and running without PAM accepts any password between 257 and 1024 characters for accounts whose password field in /etc/shadow is exactly *. The accounts affected are those intentionally locked or configured for key-only authentication, including root on many Alpine cloud images, deploy users, and certain LDAP-mapped accounts.

The condition is present only on musl-based systems that do not use PAM for password verification. glibc systems, systems with UsePAM yes, Dropbear, and OpenBSD are unaffected. OpenSSH 10.6p2, released 5 November 2026, contains a workaround. Alpine has backported the fix across supported and several EOL branches.

OpenAI is out of cash. It wanted to raise $30B last week but only $11b showed up. It's now begging for an additional $9b (down from original request of $19b more) from plebs, but no one has committed anything. They want to raise money at a $1.25T valuation, while ANThropic is struggling to clear a $1.3T valuation for its IPO, and ANT is arguably worth double whatever OpenAI is worth, or more precisely OpenAI is half of ANT.

OpenAI is likely this week to lock in that $11b in commitments to prevent "cold feet clauses", and then add incentives for the other cash it needs. Coreweave is another company that needs cash now, with no good options to raise any of it.

The AI circular financing bubble has OpenAI at the center. Instead of pretending it is worth $1T, all of the promises OpenAI has made have bankruptcy triggering clauses by the promise receivers, and under bankruptcy, selling off assets is worth less than $10b, and datacenters are dark without need for more of them, or more GPUs.

What seems surprising, is that even though the bubble fuckers know that collapse/pop occurs when OpenAI crashes, they are not stepping up with a measly $9b to keep it alive for a few months.

A coalition of private-sector security firms working with Japan’s National Police Agency has disrupted XCPE, a backdoor used to build a massive botnet comprising more than 1.3 million Android devices across more than 200 countries.

Google, Lumen, Human Security, Plume, Synthient, and TrendAI participated in the disruption.

“We assess with high confidence that the backdoor and botnet were intentionally developed, deployed, and operated for financial gain by Videostrong Technology,” the Google Threat Intelligence Group (GTIG) said in a Monday blog.

Videostrong Technology is a hardware manufacturer based in Shenzhen, China. It builds Android devices, including home TV boxes and projectors. According to Google, Videostrong deliberately embedded the XCPE backdoor in device firmware and sold access to criminal affiliates, allowing them to install additional malware modules on devices shipped to unsuspecting customers. When buyers connected these devices to their home networks, the preinstalled backdoor allowed operators to create proxy networks and hide malicious traffic using the customers' internet connections without their knowledge.

The firmware’s pre-installed management agent purported to be a standard remote-management client, according to TrendAI. It pulled updates from a GitHub page, and connected to a Message Queuing Telemetry Transport (MQTT) server that gave the operator full remote control of each box, allowing them to install and remove apps, change settings, and push updates to the infected devices.

Between June 3 and October 1, TrendAI documented more than 3.8 million distinct home and mobile IP addresses connected to the operation.

The coalition also disrupted the command-and-control domains used to manage the affiliate modules. “Because these affiliates each have diversified distribution channels for their modules beyond XCPE, we assess that taking down their C2 domains will sever access to at least 10 million additional compromised devices,” according to GTIG.

Google said it had since cut ties with Videostrong, and banned the OEM from making devices with preloaded Google apps.

Videostrong did not immediately respond to The Register’s inquiries.

This backdoor and botnet operation is part of the same ecosystem as BadBox 2.0, another notorious botnet that infected more than 10 million TV boxes and other devices before Human, Google, and TrendAI disrupted it in March 2025.®

Last week, Anthropic introduced a new system prompt designed to stop its Claude models from re-creating specific, existing video games. While that move could help shield Anthropic from some liability, it’s unlikely to slow much of the massive wave of AI-assisted game porting and copying that has emerged in recent months.

A social media user and AI decompilation enthusiast going by the name “chasm” was among the first to note the prompt change late last week, saying a new rule seemed to be added to the Claude Code prompt on or around October 27. That rule—included in a json file under the name “tengu_heron_brook”—reads as follows:

Do build original games, games in a genre, and games from the user’s own designs and assets. Don’t build a recreation of a specific existing game, even if it is renamed or called original.

XDA later independently verified that the new rule appears in version 2.1.314 of the Claude Code binary, which was released on October 30. XDA also reported that the “tengu_heron_brook” rule had been used for other purposes in earlier versions and that the Claude Opus 5.5 model specifically seemed to have its own distinct and independent instructions restricting game recreations as part of its system prompt bundle.

Following chasm’s initial report, Anthropic software engineer Addy Osmani confirmed on social media that the company had “added two sentences to Claude Code’s prompt on Opus 5.5 to steer it away from recreating someone else’s specific game because using Claude to infringe IP is against our usage policy.”

He added that the new prompt was designed to still allow “original games, genre games, and games built from your own designs and assets,” even as a handful of issue reports on Anthropic’s GitHub suggest the new prompt is getting in the way of some game development use cases.

A massive wave

Anthropic’s new prompt comes as AI coding tools are powering a deluge of game projects based on existing work in one way or another. That includes thrown-together copies of game demos that are making some developers afraid to show their early work publicly, as well as thousands of AI-powered decompilation and porting projects for retro games. Thanks to AI tools, what used to be a time-consuming process requiring painstaking human effort is now orders of magnitude more efficient (though some veteran coders complain that these vibe-coded ports and decomp projects fail to clear the same quality bar as human-centered efforts).

The trend has gotten so bad that decomp.dev—a years-old community devoted to the “brutally difficult, time-consuming task” of game decompilation—announced it was shutting down last month. Community maintainer Luke Street wrote that quickly improving AI tools had made it clear that “community aspect of decompilation has been broken” as newcomers “rac[e] to port games, especially utilizing decompilations that they did not contribute to.”

“Decomp.dev was originally designed to gamify decompilation: watching the progress go up was a fun way to stay motivated when every fraction of a percent took real human effort,” Street wrote. “Today, that same idea has warped decompilation into a hollow numbers game, exacerbated by folks who just want to burn tokens to make number go up or ‘port game to PC,’ all else be damned.

Like trying to stop the tides

While Anthropic’s new prompt can be seen as a noble effort to push back against the flood of AI-powered, vibe-coded copying, it seems unlikely to have much impact overall. That’s primarily because eager game copiers can simply shift to any of the other frontier models that haven’t introduced such restrictions. And even if the AI industry as a whole began to bar this kind of use case, coders could switch to cheaper open-weight models that aren’t subject to these kinds of restrictions and are not all that far behind in capabilities.

The prompt may have a more immediate benefit for Anthropic, helping protect the company from potential liability for copyright infringement in these projects. We’ve seen how AI companies face legal headaches if their models generate images based on copyrighted works, for instance. As Osmani noted in confirming the new anti-game copying prompt, “Using Claude to infringe IP is against our usage policy”—a policy that’s surely in place to protect the company legally.

Decompilation efforts based on reverse engineering code generally pass legal muster as long as “expressive elements” like art and sound assets are not included in the project. But there’s no guarantee that individual vibe-coders using Claude would be so careful with their efforts, even with Anthropic guardrails against IP theft in place.

On the other hand, there’s a risk that Anthropic’s efforts could prove too restrictive, limiting transformative game projects that build off existing work. Take, for instance, the increasingly popular trend of using AI to merge two different game concepts into one, adding Mario to Elden Ring or Spider-Man to Batman: Arkham Knight, for instance. Are these the kinds of “recreations of a specific existing game” that Claude’s new prompt should block? Or are they transformative works that are legally and creatively distinct from what inspired them? And how does the use of copyright-protected characters and images factor in?

Beyond mashups, how should a model like Claude ensure there’s a difference between a direct re-creation of Tetris (ostensibly disallowed by the new prompt) and a mere game in the Tetris genre (ostensibly allowed by the prompt)?

These are issues that even courts can have trouble adjudicating in the best circumstances. Asking an AI model to reliably draw the line between acceptable and unacceptable game clones is a tall order, even with a new two-sentence prompt in place.

Hot off the back of price rises it announced in October, the Raspberry Pi Foundation today said the cost of its Raspberry Pi 5 single-board computers is going up again.

From November 2026, the Raspberry Pi 5 with 16 GB RAM rises by $80, now priced at $385 from approved resellers. The model launched at $120 back in 2025, rose to $205 in early 2026, and later incremental increases got it to $305.

And it’s not the only model whose price tag will be making wallets wince.

A $40 price hike hits the Raspberry Pi 5 Compute Module (CM5) with 16 GB RAM.

For now at least, the (not exactly cheap) Raspberry Pi 500+ keyboard PC is unchanged.

The rises are, of course, not ones the foundation makes willing. It’s down to the ongoing memory crisis caused by colossal demand from the AI industry and data centre hyperscalers, exacerbated by a lack of diversified supply chain.

It says it looks forward to “rolling back our price increases when memory costs fall once again”. When will that be? Some industry bods say not until the mid-2030s.

Raspberry Pi is putting forward to creative solutions to try and help. It launched a 3GB model of its Raspberry Pi 4 and, earlier in the crisis, a dual-RAM model using lower-density chips. It also suggests would-be buyers consider how much RAM they actually need.

But it is the case that smaller hardware companies are hit worse by industry volatility, mainly because of (a lack of) purchasing power.

This year saw Pine64 nope out of making Linux devices at all due to the compounding factors from component cost rises, which had inflated to the point it rendered their whole ‘affordable’ hardware pitch to encourage hacking and tinkering utterly moot.

You’re to blame, buddy

OpenAI’s Sam Altman recently blamed memory supply chain issues on selfish users buying up RAM to run open-source AI models locally, saying it’s “quite terrible”1 for people to have “a bunch of the world’s limited supply of [DRAM] sitting in their own computer”.

If you’re reading this, tutting at the latest round of price rises, have some self awareness.

That monster gaming rig or a tricked-out workstation you’re using? Do the altruistic, public-spirited thing and remove the RAM and post it to your nearest data centre.

They need it more than you, sucker.

Nov 2 (Reuters) - A group of newspapers, including the New York Times (NYT.N), opens new tab and the New York Daily ​News, sued internet company Brave Software in Manhattan federal court ‌on Monday for allegedly scraping their articles for AI training and unlawfully enabling users of its web browser to circumvent their paywalls.

The newspapers said in the complaints, opens new tab ​that Brave illegally sells their content to AI companies including ​Anthropic, Perplexity and xAI — which are not defendants in the ⁠cases — in addition to infringing their copyrights by allowing users to ​access their articles for free

"The New York Times is overpaying its ​attorneys and underpaying its journalists. (Who could've easily identified the disinformation in this suit.)," a Perplexity spokesperson said in response

Spokespeople for Brave, Anthropic and xAI did not immediately ​respond to requests for comment on the complaints

"While we believe in ​the ethical and responsible use and development of AI, Brave has flouted established copyright ‌law ⁠and ignored repeated requests by The Times to stop its infringing conduct," New York Times spokesperson Graham James said in a statement

The news outlets have separately sued OpenAI in a potential landmark case over its AI ​training

The outlets' lawsuits ​filed on ⁠Monday said that Brave "acts both like a burglar that misappropriates the content of publishers, and an AI 'fence' that ​sells stolen content to third parties"

The cases are ​New York ⁠Times Co v. Brave Software Inc, US District Court for the Southern District of New York, No. 1:26-cv-09549 and Daily News LP v. Brave ⁠Software ​Inc, US District Court for the Southern ​District of New York, No. 1:26-cv-09554

Supermarket chains in at least seven European countries have said the temperatures of refrigerators and freezers in some of their shops soared overnight to as high as 50C (122F), echoing a cyber-attack last month that caused millions of euros of damage in France.

Retailers in Belgium, Estonia, Ireland, Finland, Luxembourg, the Netherlands and Sweden reported problems with their refrigeration systems on Monday, and several companies confirmed they had been hacked.

The EU food and agriculture commissioner, Christophe Hansen, said the repeated cyber-attacks were deeply concerning, calling them “a deliberate attempt to disrupt our food supply chain and undermine consumers’ confidence in it”.

The incidents follow an attack in France last month in which hackers seized control of cooling units in as many as 700 supermarkets, raising their temperatures as high ⁠as 50C (122F) and causing fires and food losses estimated in the tens of millions of euros.

A growing range of devices, including refrigeration systems, can now be monitored and controlled remotely, leading authorities to particularly fear government-sponsored hybrid attacks in after Russia’s 2022 invasion of Ukraine.

Sweden’s Coop chain told the daily Aftonbladet that six of its shops ‌had been hit, while ICA said it was “aware that a limited number of stores were affected by an incident overnight which impacted their refrigeration and freezer systems”.

An ICA shop at Sjöbo in southern Sweden posted that its fridges and freezers had been hacked and their temperature set at 50C. Another in Stockholm said its freezers were broken “due to a 2am cyberattack” and that “everything was destroyed”.

The head of Sweden’s national cybersecurity centre, John Billow, said it ‌was working on incidents around the country. It had established the country had “quite a lot of exposed operational ‌technology devices” that needed updating or patching, he told Reuters.

The Dutch retailer Ahold Delhaize, which last month reported a similar incident to police in the Netherlands, said fridges and ⁠freezers in dozens of its ⁠supermarkets in Belgium ‌and ‌Luxembourg had been hit.

It did not specify the cause, but reporters visiting a shop north-east of Brussels saw refrigerators and freezers empty and a sign reading: “Due to cyberattacks, there will be no products on the refrigerated or freezer shelves. We offer our sincere apologies.”

A company spokesperson, Sarah ‌De Meester, said the range of fresh and frozen products was “temporarily somewhat limited” in affected shops, but that distribution centres were operating normally and there was no disruption to deliveries.

Belgium’s Le Soir newspaper reported that dozens of shops and supermarkets had been affected, including five Spar outlets, Shop & Go convenience shops attached to Q8 service stations and some smaller Proxy Delhaize mini-markets.

“We started mobilising technicians at 1am,” Stephan Moelans, the chief executive of Sabcobel, which operates Delhaize’s refrigeration systems, told the paper. “We spent all night and morning disconnecting the affected installations.”

He told the Belga news agency the ‌incident had been a cyber-attack and that the units’ evaporators were switched to defrost mode.

The Colruyt and Intermarché supermarket chains also said they had been hit by technical problems affecting fridges in five and 15 of their shops respectively, and Carrefour Belgium reported disruption in about 15 of its 700 shops.