off prem

Starting small with $37m and maybe 50MW but reckons full-stack service plan can succeed

Indian tech services giant and retro software house HCL has decided to get into the AI datacenter business.

The company yesterday revealed its plan in an announcement [PDF] released alongside its Q1 results, which included news of three-percent year-over-year revenue growth to $3.65 billion and 20 percent growth in net income which reached $488 million.

CEO C. Vijayakumar also pointed to 62 percent year-over-year revenue growth for a segment HCL calls “Advanced AI” that encompasses building its own AI platforms.

The CEO said HCL’s strategy is to “Benefit disproportionately from the AI-native and AI-amplified opportunities” because they “together represent the fastest growing pool of enterprise spend.”

The company has therefore decided to get into the datacenter business and has found ₹3,500 crore ($36.5 million) to put toward facilities it says have “potential to scale to 50MW of capacity.”

That’s not a vast facility – just one of Meta’s datacenters will host 50GW of kit – but Vijayakumar said HCL can make it relevant by using its existing software to offer “full-stack” infrastructure.

“The biggest opportunity is not to rent AI, but to own the full stack,” the CEO said. “The datacenters that compute the models built to address client-specific needs.”

“This is a business which is shifting from physical infrastructure to higher value AI-ready solutions,” he added. “We will create full-stack offerings by combining our capabilities across AI datacenter design, DevOps, and cloud operations, as well as a software portfolio with our new datacenter business.”

HCL’s focus appears to be on Indian customers, as Vijayakumar said the datacenter investment will “position us as a key enabler of India’s sovereign AI ecosystem, expanding our presence in the fastest-growing market among largest economies with differentiated offerings around sovereign cloud, secure AI, and managed AI infrastructure.”

The CEO said HCL is already “in advanced discussions with clients to ensure we start with certain level of committed consumption from day one.”

The company didn’t say where it will build its bit barns, when they might come online, or how it will secure energy supply – an important consideration given we yesterday reported on an effort to locate a datacenter in renewable-energy-rich Bhutan to serve Indian customers.

Vijayakumar also revealed that HCL booked $2.4 billion of new business in the quarter, a record. The CEO pointed to one of those deals as an exemplar of HCL’s AI smarts, as it will see the services company work with an unnamed Fortune 250 semiconductor equipment OEM “to accelerate AI-driven transformation across its semiconductor engineering and manufacturing value stream.” To make that happen, HCL will deploy SAP, integrate it with existing systems, and establish “an enterprise backbone for a future-ready, scalable, AI-led digital supply chain.”

Another new deal, struck earlier this month and therefore not included in the $2.4 billion of new deals won in the quarter ended June 30, will see HCL work with an unidentified “Europe-headquartered Fortune Global 50 firm as a technology partner to accelerate AI-led transformation and management of their digital workplace and enterprise networks.”

Numerous reports in Indian media identified the new client as Mercedes Benz, and suggest the automotive giant has moved its business to HCL from Infosys, which announces its quarterly results next week. ®

Meta seems to be having a bit of an identity crisis. On Monday, the social networking singularity said it would spend $50 billion to expand its Hyperion datacenter project in Richland Parish, Louisiana, from 2.2 to 5 gigawatts.

The news comes less than a week after a report broke claiming that Meta was actively exploring options to offload its excess compute capacity to other AI labs.

So, which is it, Zuck? Did you invest too much or too little in AI?

The easy answer is that Meta overcommitted. Inspired by the early success of Llama, it made a huge bet on the AI gold rush. Offloading spare compute to the highest bidder is just a hedge in case its Superintelligence team turns out to be another pipe dream, like the Reality Labs Metaverse that utterly failed to spark enthusiasm for immersive environments accessible through Meta’s Quest cybergoggles.

The more pragmatic read is that Zuckerberg has woken up to the fact he’ll never be as cool as OpenAI boss Altman or Anthropic’s Amodei, and renting out spare compute is just the natural progression for any sufficiently large hyperscaler.

Dawn of the Meta cloud?

Meta’s business model is closer to Google’s than those operated by OpenAI and Anthropic.

Both Meta and Google offer various services which generate revenues by connecting users with advertisers. For Google it’s a search and entertainment empire. For Meta it’s enabling an endless feed of content generated by friends, family, influencers, and yes, bots. 

Both are immensely profitable, earning $132.2 billion and $60.5 billion in profits last year, respectively. That’s profit, not revenue.

But both are now plowing over $100 billion a year into AI infrastructure to power large language and image and video generation models. As we learned from Meta’s recent earnings calls, the most commercially potent of those models get the right ads in front of the right eyeballs.

The open secret is Meta was already one of the most successful AI companies long before ChatGPT debuted. Except, it’s not large language models (LLMs) that make Meta money, at least not in the conventional sense. Instead, Meta’s most profitable AI models are the recommender systems that mine profiles for context and use it to infer your needs. Meta’s devs evolved those models considerably over the past few years, and their architectures now look a lot more like an LLM than the now-pedestrian neural networks on which Zuckerberg built his empire.

Google is in a similar situation. It’s investing heavily in AI to feed its fast-growing and profitable cloud business, even as advertising still pays most of the bills. But unlike Google, Meta hasn’t yet made the leap from hyperscaler to cloud provider.

Amazon, Google, Microsoft, even Oracle got there eventually, and it seems AI may be the catalyst that turns Meta into a cloud, too.

Recent reports suggest that Zuckerberg is warming to the idea. 

“I think that’s certainly a thing that we could do and that I think would make sense to consider,” he said in an interview with Bloomberg last week. “As a backstop, even if for whatever reason we don’t need all the compute ourselves or for any number of reasons, there’s a very large amount of demand that I think you could sell it long-term like AWS or Azure or Google Compute.” 

But while the demand may be there, Zuckerberg emphasized the compute capacity is not readily available.

But as Ben Thompson of Stratechery put it, cashing in on this compute may be more than a backup plan. In a post channeling an imaginary Zuckerberg, Thompson suggested that becoming a neocloud would force Meta to stop chasing pipe dreams and pet projects. His logic is that if Meta can’t make money with infrastructure it buys for AI ventures, the social networking giant can lease the orphaned hardware to the highest bidder.

The takeaway for investors — should Meta follow its fellow hyperscalers-turned-cloud-providers down this road — is that the profitability of its hardware investments would no longer be tied to its ability to commercialize them.

Seizing the means of production

If history tells us anything, scale matters. Building a cloud like Amazon Web Services (AWS) is next to impossible unless you’ve already figured out how to profit from those same resources.

Meta’s scale puts it in a position to acquire compute in volumes impossible for smaller players. Its ability to capitalize on infrastructure demand relies entirely on having something others want but can’t get anywhere else.

For what it’s worth, Zuckerberg wouldn’t be the first to come to this conclusion. Earlier this year Musk-owned xAI surprised many when it announced plans to rent out its Colossus supercluster in Memphis to rival model dev Anthropic.

The calculus here is the same. Making a profit off LLMs, like Grok, isn’t easy — just ask OpenAI — but selling the means of AI production to those that haven’t yet figured that out is enormously lucrative.

The logic appears to have gotten Zuck’s attention.

“The SpaceX model I think is quite interesting in terms of just making these short-term deals that are at a big premium,” Zuckerberg told Bloomberg. “So we get offers for all kinds of stuff like this and we’ll evaluate them and see what makes sense.”

Reports suggest Meta is seriously considering two strategies for commoditizing its compute assets. The first would be a usage-based compute platform similar to Amazon Web Services’ Bedrock.

The service would allow customers to run models and serve them through APIs — interfaces that abstract operational complexity. To be clear, Meta already offers API access to its homegrown models, at least the ones it didn’t pull after realizing the way they’d been implemented could be abused. So, from what we gather the difference would be allowing customers to run third party models as well.

The second scheme reportedly being explored would involve selling raw compute resources available to end customers — similar to CoreWeave or Lambda.

All the right ingredients 

Meta’s silicon strategy may help as well. One thing all the major cloud providers have in common is a growing catalogue of custom cloud silicon.

Once they’ve identified a core use case, Amazon, Google, and Microsoft all rolled their own silicon to maximize their margins. AWS Trainium, Microsoft Maia, and Google TPUs are all examples of AI accelerators originally built for internal workloads but later made available to the broader public.

Meta has been building its own AI chips for years. The first few Meta Training and Inference Accelerators (MTIA) were designed to speed up its recommender models. But new designs, developed in collaboration with Broadcom, are far better suited to running LLMs like Llama and Muse Spark, and whatever else its customers are willing to pay for access to.

More importantly, this mix of compute means that Meta can take advantage of the fact GPUs are extremely flexible to bring new products to market quickly. Then once they’ve proven performers, Meta could transition those workloads to its custom chips and offload spare GPU compute to its cloud customers.

Meta has all the ingredients, compute, scale, and capital necessary to become a major cloud provider.  ®

Rumor mill: Apple is overhauling its Mac chip roadmap to place artificial intelligence at the center of its hardware strategy. Instead of completing the M6 lineup with the usual Pro, Max, and Ultra variants, the company is reportedly moving directly to the M7 generation and planning significantly larger Neural Engine upgrades, according to people familiar with the plans.

This fall’s Macs are still expected to debut with a base M6 chip. Under Apple’s traditional release pattern, that chip would be followed by M6 Pro and M6 Max variants for higher-end laptops, along with an M6 Ultra for desktop-class machines.

This time, that sequence will reportedly stop after the entry-level chip. The company has already moved on to the M7 design, taping it out only about six months after the M6 reached the same stage. The compressed timeline highlights how urgently Apple wants its Macs to handle increasingly demanding AI workloads.

According to the internal roadmap, the first M7 Macs are scheduled to launch in the first half of 2027. Higher-end M7 Pro and M7 Max systems are expected later that year, while the M7 Ultra is targeted for 2028.

Apple has skipped an Ultra chip before – the M4 family did not include one – but abandoning every high-end M6 variant at once would mark a first. People briefed on the plans say Apple determined that the M7’s AI-focused upgrades were significant enough to justify skipping further M6 development.

At the center of those changes is the Neural Engine, Apple’s dedicated on-chip AI hardware that powers on-device generative models, accelerated inference, and Apple Intelligence features. Apple has refined the Neural Engine with every Mac chip generation since the M1 debuted in 2020, and the M4 represented one of the biggest improvements to date.

The company now wants the M7, particularly the M7 Ultra, to approach the level of performance developers expect from dedicated AI accelerators rather than traditional general-purpose desktop processors.

Memory support is a major part of that effort. The M7 Ultra is being designed to support up to 1.5 terabytes of unified memory. That is roughly double the capacity planned for the upcoming M5 Ultra server chip and matches the maximum RAM configuration available on Apple’s 2019 Intel-based Mac Pro.

With that much memory, significantly larger AI models can remain in memory, reducing bottlenecks and limiting the need to rely on external storage or cloud-based compute. Whether Apple ships Macs with the full 1.5TB configuration will depend on memory availability, as supply constraints and elevated prices remain concerns.

These desktop plans tie directly into Apple’s server strategy. The company is preparing a more powerful AI server based on the M5 Ultra, known internally as J246. Engineers are already working on a successor built around an M7 Ultra-derived server chip, with a launch window targeted for around 2029.

In other words, the same architecture expected to power Apple’s highest-end Macs could also underpin the next generation of servers running Apple Intelligence in the cloud.

Beyond the M7 family, Apple is developing an M8 generation with even more AI-focused silicon. The lineup reportedly includes a processor code-named Soko, targeted for 2028, along with other chips for high-end Macs under the Cardinal name.

The 2028 chips are planned for a 1.4-nanometer process, which should deliver another leap in efficiency. The shift comes as AI chips encounter increasing power and cooling constraints, pushing Apple to prioritize more transistors for neural processing units and memory bandwidth rather than simply expanding CPU and GPU cores.

All of this hardware development comes alongside a more mixed picture on the software side. Apple has struggled to deliver AI-powered services at the pace many expected. Apple Intelligence and the redesigned Siri have rolled out more slowly than planned, forcing the company to adjust expectations along the way.

Still, Apple’s hardware teams have spent more than a decade building the foundation for this moment, often through projects that never reached consumers.

The most notable example is the company’s abandoned self-driving car initiative. From the beginning, Apple targeted full Level 5 autonomy and invested heavily in machine learning and custom silicon capable of processing massive amounts of AI workloads in real time.

The car project never reached the market, but the chip development work directly contributed to the Neural Engine architecture that first appeared in the iPhone X in 2017 before expanding to Macs and other devices. That same hardware foundation now sits at the center of Apple’s AI strategy.

With the M7 and M8 families, Apple is treating AI as the primary driver of its chip designs rather than simply an additional feature. AI workloads are now shaping which chips get developed, when they launch, and how their architectures are built.

from the the-last-gasps-of-antitrust-reform dept

I think I’ve extensively explained at this point why the $111 billion merger between Paramount/CBS and Warner Brothers is a gargantuan pile of shit that will indisputably harm labor, consumers, markets, creatives, and potentially even national security. It doesn’t matter the company names; every single major media merger of this type ends badly for everyone but the trust fund brunchlords at the top.

Not only that, every single merger involving this particular company (Time Warner, Warner Brothers) in the last quarter century has resulted in nothing but layoffs, price hikes, shittier product, and a lot of whimpering. And there are ample signs that the Paramount folks are even less competent than past suitors, including the AT&T executives, who quickly got too far out over their skis.

While the Trump DOJ has unsurprisingly rubber stamped Larry Ellison’s clumsy effort to dominate what’s left of U.S. corporate media, states keep hinting at the fact they’ll file a collective antitrust lawsuit.

That’s certainly the case in Oregon, where Attorney General Dan Rayfield is asking for a 60 day pause in deal finalization while his office investigates both the deal — and apparently the Trump cronyism that has helped enable it. Rayfield, for one, accuses Paramount of refusing to adequately respond to state AG requests for information about the deal’s impact:

“We’re not going to let Paramount Skydance play hide the ball so they can rush through their massive merger. Oregonians have a real stake in this deal – in our film industry, in our economy, in the choices they’ll have as consumers. Paramount had every opportunity to hand over records and answer a few basic questions. Instead, it is trying to run out the clock and evade scrutiny. We’re asking the court to make sure Oregonians get the answers they’re owed before this deal closes, not after.”

Rayfield says that Paramount has been particularly cagey when asked for data on its interactions with the Trump administration and Trump DOJ. Including details on a federal government influence campaign Paramount internally calls “project warrior”:

“Paramount has not complied. According to court papers, the company declined to accept service of the request, waited weeks to respond, and ultimately sent objections on the day its documents were due – objections the state dismisses as a baseless tactic to avoid turning over the records. Paramount has told Oregon it does not intend to close the deal before July 16 but has not agreed to hold off any longer while the state’s investigation continues.”

So while the $111 billion deal is abjectly terrible, it’s not quite yet a done deal yet. I’d suspect that a joint antitrust lawsuit featuring the handful of states that still care about such a thing will arrive sometime in the next month or two. While it might not succeed in scuttling the deal, it could extend the timeline in a way that could prove costly for Larry Ellison, David Ellison, and their debt-riddled proposal.

Larry is comically over-leveraged on AI, and should an AI bubble pop coincide in significant regulatory delays for his media domination bid, things could get very interesting. As an aside, I’d like you to notice how when the state lawsuits come, all the Republicans who spent last election season pretending they were suddenly interested in antitrust reform will be absolutely nowhere to be found.

Filed Under: antitrust reform, consolidation, Dan Rayfield, hollywood, larry ellison, media mergers, oregon, regulatory approval

X has made a “tweak” to its algorithm to boost the visibility of posts to users’ “mutuals” — the people they follow who follow them back, head of product, Nikita Bier, said Monday.

“We noticed this data was missing from the algo and it made your friends appear less in your replies. This resulted in the reply section feeling more like a battleground with people you don’t recognize.”

The change may not drastically revamp the site’s user experience, but may make X feel a little bit more like a community rather than a torrent of disparate voices shouting into the digital abyss.

Bier noted that the change would also “help clusters form around interests more easily, which many people have asked for.”

X has introduced a number of changes lately — many of which seem designed to make the site a bigger hub for creators. Earlier this year, the site changed how it compensates accounts in an effort to incentivize original content rather than mere aggregation, and, earlier this month, it also introduced a video editor designed to make it easier for users to work on the platform.

This tweak follows changes that Meta’s Threads has been making to its algorithm aimed at creating communities, largely as a differentiation from its main rival X. For instance, last month Threads rolled out a Your Algo feature, which lets users privately control what they see in their feed. It also reached 500 million monthly active users.

Singapore-based video-generation startup PixVerse said today that it has closed its Series C extension, with a total of $439 million raised in the round. The company told TechCrunch that, with the new tranche of funding, its valuation has crossed over $2 billion. With the cash, the company aims to expand its world model offering and reach customers across geographies.

The company closed its initial Series C round in March, led by CDH Investments. While it didn’t disclose the funding amount, Bloomberg reported it to be in the range of $300 million. PixVerse said that investors in the extension round include Alibaba, Lollapalooza Capital, Ivy Capital, Grand Mount Capital, Eastern Bell Capital, Mirae Asset, BlueFocus, and CloudAlpha, joining returning investors iGlobe Partners and OCBC’s Lion X Ventures.

The company was founded by Wang Changhu and Jaden Xie in 2023. Changhu previously worked at ByteDance on computer vision, and Xie was an executive director at investment firm Lighthouse Capital.

PixVerse offers multiple models, including a V-Series video model for consumer and API use, a C-Series video model for professional film and commercial workflows, and an R-Series of world models for game development and world building, which was released earlier this year.

Through its tool, users can generate videos in up to 4k resolution with audio baked in. The startup said that its consumer product has over 150 million registered users and over 15 million monthly active users. The company declined to specify how many of them are paying users but it offers a competitive rate of $4.80 per minute of generation for image-to-video.

Xie believes that despite the huge opportunity for video generation to succeed, only a few companies are making progress in the market.

“OpenAI exited the business when they shut down Sora 2. Other companies like Meta and Tencent are not able to create high-quality video models. So there are only a few companies that can meet the quality bar,” he told TechCrunch.

He said that there is equal opportunity in the consumer and enterprise markets as users are creating videos for fun and also consuming short video content made with AI, while enterprises are using video generation for creative, learning, and marketing use cases.

However, saying that the startup’s model produces a “high-quality” output is hardly a unique qualifier. Xie mentioned that its core strength lies in labeling.

“We think the key difference is not in data, but how you label it, because data is available everywhere. My co-founder worked at ByteDance, where he built core visual understanding technology behind TikTok using AI. Using this tech, TikTok was able to label data accurately and build a strong recommendation algorithm. This experience comes in handy when building a video-generation platform,” Xie said.

The company has big ambitions this year. It wants to expand its enterprise outreach across the globe. The startup already has a deal with its investor Alibaba to deploy the video-generation features.

In terms of product rollout, it plans to launch a new V-Series model for video generation and release a new version of its world model this year. It has 150 employees across offices in Singapore, Beijing, and Shanghai. With the new funding, PixVerse aims to hire more researchers and people in go-to-market function.

Despite its confidence in its own models and products, the video market is heating up. There are players like ByteDance with its Seedance model, former Tencent AI head Dr. Wei Liu’s Video Rebirth, and Kling AI from Asia. In the West, there are competitors like Midjourney, Runway, and Luma. Multiple companies, including Yann LeCun’s and Fei-Fei Li’s startups, are building world models.

The hardest problem in AI is no longer the chip but the megawatt.

For much of the past three years, the global AI race has focused on semiconductors, with governments competing for advanced chips, technology outfits scrambling to secure GPUs, and investors pouring billions into ever larger datacenters. Yet the binding constraint has shifted from compute to the power required to run it.

For anyone trying to energize a new AI cluster today, the bottleneck is rarely silicon; it is grid access, interconnection delays, and aging infrastructure. That was the central message from Envision founder and CEO Lei Zhang at VivaTech in Paris this June, where he argued that AI amounts to an energy revolution as much as a computing one.

The steam engine transformed the industrial age by converting coal into motion, and the GPU now transforms the AI age by converting electricity into intelligence. History offers another lesson: James Watt changed industry through the efficient use of energy rather than by producing more steam. AI faces the same problem today, because the binding constraint has shifted from how many chips can be built to how they can be powered.

The real risk: AI competing with society for energy

The numbers behind the argument are stark. Goldman puts US datacenter power demand at 31 GW in 2025, rising to 66 GW by 2027, while assuming only about 72 percent of scheduled facilities arrive on time because electricity, not construction, is what typically slips. The IEA estimates that datacenters consumed roughly 1.5 percent of world electricity in 2024, a share rising to 3 percent by 2030 as AI-specific demand triples.

The structural mismatch sits at the heart of the problem: AI models iterate every six months and chips refresh annually, while power grids have changed little in decades. Rack densities that sat at 5 kW are climbing toward 200 kW, and the IEA notes that AI server power density rose elevenfold between 2020 and 2025, with a further fourfold rise expected by 2027, straining the supply chains for power electronics and transformers that keep a cluster stable. The growing gap raises broader questions about where the energy will come from and who will bear the cost.

Around the world, communities are asking whether AI infrastructure should draw on electricity that households, factories, hospitals, and public services also depend upon, with familiar concerns surfacing about consumer bills, manufacturer access to limited grid capacity, and the burden that ever-larger models place on public infrastructure.

Those questions have moved beyond the purely technical into the societal, because the future of AI cannot rest on a model in which humanity competes with AI for power.

Mission Gobi: Let AI follow energy

Envision’s answer, Mission Gobi, unveiled at VivaTech, aims to develop 5 GW of green AI computing capacity across deserts and arid regions by 2030. For decades energy followed computing, and Mission Gobi reverses that logic on the premise that in the AI era, computing may need to follow energy.

The logic is grounded in geography, because deserts offer some of the world’s richest solar and wind resources alongside vast expanses of low-cost land, with the additional advantage of little competing residential or industrial demand. Rather than drawing power from homes, factories, and public services, Mission Gobi seeks to build entirely new renewable energy systems dedicated to AI, expanding the available supply instead of asking society to share a fixed pie.

The philosophy reduces to a single idea: compute should chase power, not the other way around.

The economics matter because electricity determines whether a facility is viable, with power consistently accounting for the single largest operating cost at a datacenter and some estimates placing it at as much as 60 percent of the operational budget.

Building energy-native AI infrastructure

Envision splits the system into three layers: an intelligent operating hub, Physical AI powered by its Tianji Weather Foundation Model and Dubhe Energy Foundation Model, and advanced power infrastructure. Together they integrate generation, storage, grid, power electronics, computing, and large-scale AI models into a unified architecture.

The challenge lies in coordinating renewable power rather than merely generating it, because AI facilities require stable, high-quality electricity while solar and wind output fluctuate continuously. Envision argues that large-scale predictive models can help balance generation, storage, and demand in real time.

The concept has already moved beyond theory. In Chifeng, Inner Mongolia, Envision runs a 2 GW system on 100 percent renewable energy, coordinating wind, solar, storage, hydrogen, and compute in real time, while a gigawatt-scale AI and computing campus in Ulanqab is being developed as a demonstration of what energy-native computing infrastructure could look like.

A 5 GW pledge is ambitious, but the underlying read is sound: retrofitting decades-old city grids for gigawatt AI loads is a difficult undertaking, and purpose-built renewable compute, sited where power is cheapest, offers a credible alternative.

SpaceX looks up, Mission Gobi looks out

Envision is not alone in recognizing energy as AI’s defining constraint. Elon Musk’s SpaceX has explored concepts for orbital datacenters powered by uninterrupted solar energy in space, and the vision rests on the same recognition: the future bottleneck of AI may lie in energy rather than silicon. Both approaches seek to place computing where energy is most abundant.

The two visions diverge in geography, with one reaching upward beyond Earth’s atmosphere and the other outward toward deserts and Gobi regions, though both start from the same premise: AI should not compete with humanity for power.

A new blueprint for AI infrastructure

If the industrial age was built around coal and the electrical age around power grids, the AI age may be built around energy abundance. The success of future AI infrastructure will not be measured by GPU counts and model sizes alone. It will also depend on whether the industry creates new energy supply, eases pressure on communities, and enables technological progress without reducing others’ access to power.

Whether deserts become the preferred destination for future computing remains to be seen. What is becoming clear is that the next phase of the AI race will be defined not only by who builds the most powerful models, but by who can build the energy systems capable of sustaining them.

The path forward runs through creating new energy supply rather than reallocating existing capacity away from households, factories, and public services.

Contributed by Envision.

The Jscrambler client-side web security company disclosed that a threat actor published a malicious version of its npm package that has been downloaded almost 1,500 times.

The malicious Jscrambler package spanned releases 8.14, 8.16, 8.17, and 8.20 and included information-stealing malware that executed during the ‘preinstall’ hook.

“Today, we identified the unauthorized publication of a malicious version of our jscrambler npm package, which is used with our Code Integrity product,” Jscrambler says in a warning on Saturday.

“This incident was limited to that package and did not affect any other Jscrambler products, including Webpage Integrity,” the company said.

Although Jscrambler reacted quickly, the malicious package lasted for two hours before the developer deprecated it and released the safe version 8.22.

The affected package was a dependency for four other Jscrambler packages, which the vendor has also deprecated and replaced with new versions.

Statistical data from Node Package Manager (npm) shows that the malicious package was downloaded 1,479 times during the two-hour window.

Jscrambler is a commercial platform for protecting web and mobile JavaScript applications from reverse engineering and tampering.

Its npm package has 17,000 weekly downloads and enables app developers to upload their JavaScript to Jscrambler’s service to protect the code from alteration. This helps defend against real-time modifications like injecting malicious code.

Application-security company Socket detected the compromise and analyzed the unauthorized Jscrambler release. The researchers say that the package included an infostealer that targeted multiple types of sensitive data:

• Source code and project files
• Developer credentials and secrets (Git, SSH, environment variables, CI/CD tokens)
• Cloud credentials and secret managers (AWS, Azure, GCP, Kubernetes)
• AI coding tools and MCP configurations (Claude, Cursor, Windsurf, VS Code, Zed)
• Cryptocurrency wallets and seed phrases (MetaMask, Phantom, Coinbase, Exodus, Trust Wallet)
• Browser data (cookies, saved credentials)
• Messaging and collaboration apps (Slack, Discord, Telegram)

Socket reports that the malware used strong per-string obfuscation via the ChaCha20-Poly1305 encryption algorithm, which made it difficult to reverse-engineer the code.

According to Jscrambler, the compromise was possible due to compromised npm publishing credentials, which the company has revoked.

Following the incident, additional security controls have been implemented for the publishing pipeline.

Developers who have used the malicious npm packages should treat their environments as compromised, rotate all secrets, and restore from safe backups.

Jscrambler recommends that customers make sure that they are using the latest version of the product.

Security teams log 54% of successful attacks and alert on just 14%. The rest move through your environment unseen.

The Picus whitepaper shows how breach and attack simulation tests your SIEM and EDR rules so threats stop slipping by detection.

Get the whitepaper

Criterion has announced The Complete Kubrick, a 30-disc 4K UHD plus Blu-ray collector’s set arriving on October 20, 2026, with a listed price of $599.95. That is not casual movie-night money. That is “I may need to explain this charge to another adult” money. But if any filmmaker was going to justify this kind of physical media overkill, Stanley Kubrick is on the very short list.

Kubrick has always been one of those directors I return to when I want to be reminded that cinema can be cold, furious, absurd, beautiful, cruel, and technically obsessive all at once. Paths of Glory, Dr. Strangelove, or: How I Learned to Stop Worrying and Love the Bomb, and Full Metal Jacket are not films I merely admire on the shelves in my film collection. I have watched them dozens of times. They shaped how I think about war films, political satire, military authority, moral cowardice, and the specific terror of men in rooms making decisions that ruin other people’s lives.

Paths of Glory remains one of the most devastating antiwar films ever made because it does not need battlefield sprawl to make its point. Dr. Strangelove is still horrifying because it is funny in exactly the wrong way. Full Metal Jacket never lets you get comfortable with its structure, its violence, or its view of how institutions break people before sending them somewhere worse.

That is the appeal of this set. Kubrick is not a casual background viewing director, unless you are the kind of person who folds laundry to A Clockwork Orange, in which case we should probably alert someone.

His films are built for repeat viewing because the details keep changing on you. The framing. The sound. The silence. The faces. The rooms. The way a joke in Dr. Strangelove starts out funny and then turns into a mushroom cloud. Presentation quality is not some luxury add-on with Kubrick. It is part of the experience, right up there with existential dread, institutional cruelty, bad men in sealed rooms, and the uncomfortable suspicion that somebody may ask you to surrender bodily fluids if you stop posting cheerful messages about how wonderful it is that New York City is being run by a Communist.

Related Reviews:

What You Get in The Complete Kubrick

Criterion says the set brings together Kubrick’s thirteen features and three shorts, all restored in 4K, with their original soundtracks alongside restored and remastered 5.1 mixes. The package also includes more than 25 hours of interviews, documentaries, and behind-the-scenes materials.

Criterion’s listed feature lineup includes Killer’s Kiss, The Killing, Paths of Glory, Spartacus, Lolita, Dr. Strangelove, or: How I Learned to Stop Worrying and Love the Bomb, 2001: A Space Odyssey, A Clockwork Orange, Barry Lyndon, The Shining, Full Metal Jacket, and Eyes Wide Shut. Criterion’s special-features listing also specifies that Fear and Desire is included among the thirteen features restored in 4K.

The short films include Kubrick’s early documentary work, and outside coverage of the set lists Day of the Fight, Flying Padre, and The Seafarers, with Day of the Fight included in both its original and RKO versions.

The major technical and collector details are:

• Release date: October 20, 2026
• Format: 4K UHD plus Blu-ray
• Disc count: 30 discs
• SRP: $599.95 USD
• Films included: 13 features and 3 shorts
• Restoration: all features and shorts restored in 4K
• Audio: original soundtracks plus restored and remastered 5.1 mixes
• Supplements: more than 25 hours of interviews, documentaries, and behind-the-scenes materials
• Packaging: deluxe box inspired by Kubrick’s archive, with rare photographs, artwork, and annotated documents
• Packaging design: Drusilla Adeline/Sister Hyde Design

Criterion has not yet published a detailed per-film technical breakdown for every disc, so buyers should not assume that every title has the same HDR format, audio configuration, or supplement loadout beyond what Criterion has explicitly listed. The confirmed umbrella detail is that the films are restored in 4K, with original soundtracks and restored/remastered 5.1 mixes included in the set.

The Supplements Are a Big Part of the Price

The headline number is the $599.95 price, but the extras are where Criterion is trying to justify the scale of the release.

The set includes Kubrick’s international version of The Shining, a new 4K restoration of Vivian Kubrick’s behind-the-scenes documentary Making “The Shining”, newly recorded commentary tracks with filmmaker Lee Unkrich and author Michael Benson, rare films from Graphic Films and computer-animation pioneer John Whitney that inspired the special effects in 2001: A Space Odyssey, unseen Lolita screen tests with James Mason and Sue Lyon, and rare Full Metal Jacket behind-the-scenes footage.

Criterion also lists a newly recorded conversation with novelist Jonathan Lethem and film historian Kevin Wynter on Kubrick and authorship, plus an essay by author and critic Nathaniel Rich.

That is not filler. For Kubrick, the context matters. These films have been examined, argued over, imitated, worshipped, misunderstood, parodied, and dissected for decades. A serious box set needs more than transfers. It needs production history, alternate perspectives, archival material, and enough scholarly weight to make this feel like the early Chanukah present I will absolutely justify buying for myself in December instead of the new winter boots my children probably need.

Why the Price Might Make Sense

$599.95 is a lot of money. There is no point pretending otherwise. But the math is not completely insane once you break it down.

At the listed Criterion price, the set works out to about $20 per disc across 30 discs, or roughly $38 per included film or short if you divide the price across the 16 works. If you count only the 13 features, it comes to about $46 per feature.

That does not make it cheap. It does make it less ridiculous than the sticker shock suggests, especially if you are someone who would eventually buy most of these titles individually on 4K anyway.

The real question is whether you want Kubrick as a unified collection. If you only care about 2001: A Space Odyssey, The Shining, and A Clockwork Orange, this is probably not the smartest buy. But if you care about the full arc from the early independent work through Eyes Wide Shut, the value proposition changes and you get to see Nicole Kidman when she still had those curly locks and before she made those annoying AMC Movie trailers.

That is why a complete set has a stronger case than another loose pile of individual UHDs.

Who Should Buy It

This is for Kubrick collectors, Criterion completists, film students with irresponsible credit cards, and anyone who wants a single archival-style edition of the director’s entire feature output.

It is also for home theater owners who understand that Kubrick’s films are not merely “content.” They are image, sound, rhythm, geometry, silence, and discomfort.

Be honest with yourself: if you are seriously considering this set, you are not buying movies. You are reporting for inspection. Full Metal Jacket people know the drill: R. Lee Ermey is screaming, Vincent D’Onofrio is unraveling in real time, and somewhere in the room a $599.95 $479.96 Criterion preorder is sitting there like the world’s most expensive jelly doughnut.

The Bottom Line

Criterion’s The Complete Kubrick is expensive, but it is not a lazy cash grab. The set includes 30 discs, 13 features, 3 shorts, 4K restorations, original soundtracks, restored and remastered 5.1 mixes, more than 25 hours of extras, rare archival material, new commentary tracks, and deluxe archive-inspired packaging.

That is the right kind of excessive.

For casual viewers, this is overkill. For serious Kubrick collectors, it may be one of the biggest physical media releases of 2026. Start saving.

Where to Pre-order

Available October 20, 2026

Related Reading: