OpenAI is expanding its relationship with Samsung beyond memory supply and enterprise software as the AI giant plans to outsource production of at least some of its processors to Samsung Foundry, Harrison Kim, General Manager of OpenAI Korea, revealed this week. If the information is accurate, then OpenAI will source its AI accelerators from both TSMC and Samsung Foundry, which suggests massive volume requirements.

"One of the areas where we have made the most progress and gained the most recognition with Samsung Electronics is our joint production and ​research on the next-generation chips we are developing," said Harrison Kim, General Manager of OpenAI Korea, at ​a press conference in Seoul, Reuters reports.

OpenAI already has its own AI ASIC program that relies on Broadcom's design services as well as TSMC's wafer processing and advanced packaging services. So far, the company has introduced its first inference AI accelerator called Jalapeño that was defined by the company's engineers, then co-designed with Broadcom, then made by TSMC, all in less than 18 months.

OpenAI did not explain whether Samsung's participation concerns a second production source for Jalapeño, another processor under development by OpenAI, a chip jointly developed by Samsung and OpenAI, or some other aspect of chip production and development. Samsung and SK hynix already supply memory for OpenAI's Stargate data center initiative, though joint chip development and production barely have a relation to DRAM supply.

OpenAI's 1^st Generation Jalapeño will unlikely be double-sourced from TSMC and Samsung because the chip is already in mass production at TSMC and OpenAI is talking about 'next-generation chips,' not the ones that are in mass production at the moment. Furthermore, development of Jalapeño's successor is well underway and is approaching tapeout, which means that its mass production is not far away either. Since OpenAI's claim clearly involves 'next-generation chips,' it is entirely possible that OpenAI will indeed produce its 2^nd Generation inference ASIC at Samsung Foundry.

Back in late July, Samsung Electronics and Broadcom announced a strategic partnership valued at over $200 billion through 2030 to collaborate on advanced foundry, memory, and packaging technologies for AI infrastructure. Hence, as OpenAI has an agreement with Broadcom to procure 10GW of custom AI accelerators, it will be able to produce these accelerators at both Samsung and TSMC. Of course, if it needs silicon produced at Samsung, and pays Broadcom for appropriate design porting.

Perhaps, OpenAI will take a page from Tesla's book and will double-source Jalapeño's successor from TSMC and Samsung to get higher volumes. However, Tesla's volume requirements may be different from those of OpenAI.

Tesla needs extraordinary AI5 volumes because it intends to use the processor across three very different high-volume applications: AI data centers, vehicles, and Optimus robots. Therefore, Tesla could potentially need millions of AI5 chips for cars alone, on top of robots and data-center deployments. Therefore, paying for separate TSMC and Samsung physical implementations gives Tesla not only supply-chain resilience but also aggregate capacity necessary to supply several product categories.

Yet, data center accelerators tend to be vastly more silicon-intensive per unit compared to ASICs for vehicles or robots. If OpenAI/Broadcom's next ASIC is a large leading-edge processor with multiple dies and OpenAI wants gigawatts of these processors, wafer requirements could still become too high for TSMC alone (which is fully booked by the likes of AMD and Nvidia). In that situation, OpenAI may need another foundry to get enough ASICs. Still, we are speculating.

Most anyone involved in computing has heard about the P-NP problem, but fluid engineers and mathematicians would love to know if the Navier-Stokes equations have smooth, globally defined solutions. Both questions are part of the Millennium Prize Problems, solutions to which are worth a cool $1 million and eternal renown. OpenAI is claiming that its staff and internal models have solved the conditions of Navier-Stokes solutions set forth in the Millennium Prize. But the company's shouting from the rooftops is being met with a chorus of boos over claims it might have plagiarized the work of a research team that had been toiling on a related, stepping-stone problem for a year.

Tristan Buckmaster (a scientist at NYU) and Levent Alpöge (a member of Anthropic's staff) had been quietly working on proving Euler's equations — another long-standing mathematical problem, and one that is generally acknowledged to be a stepping stone to solving Navier-Stokes.

According to Buckmaster, his work with Alpöge was "a purely personal collaboration, free of any institutional agreements or official involvement by either of our employers." The researchers used Anthropic Claude and OpenAI Codex as assistants, as is apparently now common in the field, to perform busywork (documentation, searching, etc.) as well as running through logic steps. The substantial amount of compute time the project required was paid from Buckmaster's own pockets, too.

The pair worked for roughly a year until August 15, 2026, when it obtained "the blowup results, with smooth forcing, for both Boussinesq and Euler." Buckmaster says the novel approach was based on previous work by Diego Córdoba and Luis Martínez-Zoroa, and he believes Zoroa should be eligible for a Fields Medal.

Although the team was presumably happy with these achievements, Buckmaster said that the LLM-generated proof was "the most horrendous" he'd seen, calling it "AI slop," and meaning to rewrite it for clarity. Nevertheless, they verified it on August 22 using Lean, a standardized programming language designed specifically to verify mathematical proofs.

Come September 3, Alpöge told Buckmaster of rumors going around that Anthropic had solved an important mathematical problem. This almost certainly alluded to the team's work, and some apparently took it to mean the company itself was working on the problem. The rumor-mongers even theorized that the problem that Anthropic had solved was Navier-Stokes. Alpöge further believed that OpenAI had gotten wind of the news.

This prompted Buckmaster to email an unnamed "prominent mathematician" at OpenAI, clarifying that the effort was a personal collaboration between him and Alpöge and was unrelated to Anthropic. The mathematician replied asking for details, saying "it would be useful to avoid competing," and offering OpenAI compute time. After a few days, on September 6, Buckmaster, the unnamed person, and OpenAI's Sébastien Bubeck talked twice, without Alpöge. He was told that OpenAI had proven a finite-time blowup for the forced Navier-Stokes equations, a subset of the problem.

Alpöge asked by text for the precise statement and was told "existence of forced blowup in R³ and T³", and that "the forcing function is smooth option [C] and [D] in Fefferman," referring to one of the four possible categories established by the Millennium Prize, with any one of them being valid as eligible for the prize, but not constituting a full solution for all scenarios, a distinction remarked on by other scientists.

This is where the story becomes interesting. Buckmaster claims that that idea (forced blowup) was exactly the same one his team had "quietly" chosen, and that nobody else he knew was working on it. Perhaps most importantly, he says that that was "not the direction one arrives at in a few days by giving a model the problem statement," indicating that running the general problem through a bot wouldn't quickly reveal that potential approach.

In fact, Buckmaster claims that over the calls, Bubeck ultimately revealed that instead of just AI models and agents with a couple of handlers, there was an entire team of live humans working on Navier-Stokes. The OpenAI team first had the models try to work through easier paths, and the text prompt that generated the Navier-Stokes proof had itself been generated by prompting Codex, with an "insane" amount of computing needed.

Buckmaster then asked when the initial prompt was issued, and OpenAI's response of "in the past few days" did not arrive until "some time" passed. He proceeded to ask if the model "had been trained on, or had access to, our sessions in Codex," and was told by OpenAI that Codex does not access user data. Finally, he asked if the data was used for model training more generally and, crucially, apparently did not get an answer.

OpenAI allegedly offered Buckmaster two options: one, that Buckmaster and Alpöge publish their Euler proof first. The following day, OpenAI would post its Navier-Stokes proof, giving the two priority. The second option was that Buckmaster alone, without Levant, was to write a paper with the Navier-Stokes proof, acknowledging that an internal OpenAI model resolved it. Bubeck was apparently adamant about Levant's removal from the Euler proof, as his employment at Anthropic was "annoying." Buckmaster opted for neither, and told OpenAI that if it chose the first option, he'd go public with his findings, as has since occurred.

This prompted what Buckmaster interpreted as a threat from Bubeck, who asked him "why [he] would ruin [his] career." After Buckmaster asked why that would happen, Bubeck told him, "If you don't want me to be nice, then I don't have to be nice." Bubeck then allegedly reached out to Alpöge, questioning Buckmaster's sanity, to which Alpöge responded with a refusal, pointing inquiries back to his colleague.

The entire story raises pointed questions about what OpenAI (and others) are actually doing with user data collected via its LLMs, despite the toggle switches that are supposed to disable it. Not only has OpenAI neglected to tell Buckmaster whether it used his team's data for training, in its PR about Navier-Stokes, the company says while it "no specific user data was accessed in order to solve this problem," it "cannot rule out that de-identified data derived from their usage of our products helped improve [its] models."

OpenAI's proof still needs to undergo a likely years-long peer review before any party can take the Millennium Prize home. The firm has stated it does not intend to claim it. As for Bubeck, he predictably paints the story in a very different light, but insists that his pushing away of Alpöge is justified on the basis that "it would be inappropriate for an Anthropic employee to author OpenAI's work," a puzzling statement that some could take as meaning a double standard regarding scientific authorship, based solely on corporate rivalry.

For his part, OpenAI CEO Sam Altman claims his team was well-intentioned and cooperative, and supported Bubeck, saying "it was challenging to offer [the same publication options] to Levent." Neither person opted to discuss the matter of whether OpenAI used the research of Buckmaster and Alpöge as training data, or offered any further explanation of why Alpöge didn't deserve credit for his work as an equal to Buckmaster.

Given the groundbreaking nature of this apparent discovery and the ensuing fight for priority that these competing accounts have sparked, it'll likely take quite some time and review before we know whether and how OpenAI or Buckmaster and Alpöge will be credited with this discovery. But given the inter-lab rancor already on display, the process will surely be ugly.

In a statement to Tom's Hardware, TV manufacturer LG has strongly denied recent claims that its smart TVs constantly log and upload data, record audio while in standby, while confirming they scan local area networks for other devices, a feature common in smart TVs and media devices. The statement follows an investigation published online by Gamers Nexus, which claims LG is operating "216,000,000" spy TVs.

Earlier this week, Gamers Nexus published a video on YouTube claiming that LG smart TVs have several security and privacy concerns. The two-hour video claims, among other things, that LG's smart TVs constantly log and upload user data, even when offline or in standby mode. The TVs were also purported to be found scanning Wi-Fi networks, recording audio logs, and sampling inputs for audio and video to identify what users were watching.

The testing was done in collaboration with two security researchers, MrBruh and uturn, and includes packet capture and firmware analysis. The TVs were reportedly seen identifying other devices on local networks, including phones, printers, and more. Perhaps the most significant claim regards the capturing of audio through TV microphones, even when the devices were off, with the report claiming the TVs stored data when offline for retrieval later on.

"The claims made in the recently published video are not true," LG told Tom's Hardware in a statement. "LG TVs process voice data only when the voice button on the remote control is pressed and held, or when a wake word such as 'Hi LG' is recognized after the user has activated the Far-Field voice recognition feature."

LG went on to say that beyond the aforementioned instances, "the TVs do not collect or record ambient conversations." The company further stated that if the wake word is not recognized, then no voice data is transmitted from the TV to the server, and said audio processing for the wake word is done on-device.

The company did admit its smart TV functionality includes scanning for devices on the same network. However, as many observers have been quick to point out, LG says this is a standard function of all smart TVs and other smart home devices.

Automatic content recognition (ACR) is an opt-in feature according to LG, which delivers personalized content recommendations, as well as services and advertisements. The company says that, as standard, ACR data isn't used for advertising purposes without consent.

Addressing some of the claims in more detail, LG stated that its smart TVs don't collect, record, or transmit ambient conversations in the home unless voice functionality has been activated by a user. It further stated that its far-field voice recognition tech (Hi LG) must be activated by a user before the TV starts listening, much like Apple's "Hey Siri" mechanism for iPhone.

LG did confirm that its TVs monitor for Hi LG's wake word while in standby mode, but says that unless the wake word is detected, the audio is only processed locally before being deleted. The company didn't address some of the other claims and vulnerabilities pointed out in the video, such as the storing of transcripts in plain text.

Tom's Hardware has not independently verified either the claims made by Gamers Nexus in its documentary or LG's counterclaims regarding the issue. LG's full statement follows:

LG Statement:

This is to provide LG's position regarding the claims recently raised by the Gamers Nexus YouTube channel.

The claims made in the recently published video are not true. LG TVs process voice data only when the voice button on the remote control is pressed and held, or when a wake word such as 'Hi LG' is recognized after the user has activated the Far-Field voice recognition feature. Other than these instances, the TVs do not collect or record ambient conversations. (If the wake word is not recognized, no voice data is transmitted to the server; the audio processing for wake word detection is performed locally on the device and is immediately deleted.)

Additionally, to provide smart TV functionalities, LG TVs feature the ability to scan for and connect to nearby devices on the same network. This is a standard function commonly provided by smart TVs and smart home devices.

The ACR (Automatic Content Recognition) feature is provided on an opt-in basis to deliver personalized content recommendation, services, and advertisements. If a user does not consent to the applicable optional agreement, ACR data is not used for advertising purposes.

Topic by topic breakdown:

At LG Electronics, we are committed to transparency and secure user experience. To clarify recent concerns and reiterate our commitment, we would like to emphasize the following points:

• Protecting Voice Privacy: We ensure that your LG TV does not collect, record or transmit ambient conversations in your home, unless voice functionality has been intentionally activated by the user.
• Far-Field Voice Recognition ("Hi LG"): This feature must be manually activated by the user before the TV begins listening for the specific wake word. Audio processing occurs only when voice functionality has been activated by the user and a wake work (such as “Hi LG”) is detected.
• Standby Mode Operation: When the TV is in standby mode and appears to be turned off, it only monitors for the wake word if you have previously enabled the Far-Field feature. If no wake word is detected, the audio used for wake word detection is processed locally on the TV, promptly deleted, and is not transmitted to LG servers.
• Transparent and Useful Connectivity: LG TVs can identify compatible devices on the same network to enable features such as device connectivity, content sharing or smart home functionality.
• Proactive and Continuous Security: Data security is a top priority for LG. We continuously monitor our platform and partner with independent security experts to identify vulnerabilities and deliver timely security updates.
• Full Control Over Data and Terms: LG Electronics is committed to providing transparency and rejecting deceptive practices (dark patterns), ensuring users can choose their preferred privacy and consent options directly through their TV settings.
• Consent-Based Activation: To provide a more tailored and convenient experience, LG TVs offer Automatic Content Recognition (ACR) to enhance viewing experiences, Voice Recognition for hands-free control, and Interest-Based Advertising to deliver relevant advertisements. Each feature requires separate explicit user opt-in consent and can be disabled at any time through the TV settings.

"AI is grown, more than designed," OpenAI's chief scientist, Jakub Pachocki, said in a new blog post on the company's latest GPT-6 Astra release. Titling the piece "An Alien Mind," Pachocki portrays the latest large language model as something more ethereal and harder to quantify. Jensen Huang calls it AGI, and OpenAI claims it's the best, most aligned model the company has ever released. It's safer to delegate, better at complex work tasks, and it can even beat Portal in just a few hours.

Huang also said that AGI had previously been achieved back in March earlier this year. Artificial Analysis benchmarks suggest Astra is about as smart as Fable 5.1 - though crucially, cheaper on a per-task basis. Astra may well be better aligned than models in the past, and it may well be more capable in specific tasks and specific benchmarks. However, the claims that the model has achieved AGI, or Artificial General Intelligence, suggest an inflection point for the AI industry.

Astra's release comes alongside calls for an industry slowdown, greater government oversight, and controls on the AI industry. Now, OpenAI's Astra raises more eyebrows about frontier-level intelligence.

Trust us, we don't know what we're doing

The tone around OpenAI's Astra release is intriguing. OpenAI's produced a new set of benchmarks, touting bold claims about the model's reasoning capabilities, with the model trained on 100,000 Blackwell GPUs, with more coming soon.

But Pachocki's blog post is much more nebulous. While Huang touts that AGI has arrived publicly, Pachocki says AI can only ever "simulate facets of human behaviour," not recreate it. He describes AI development as an experimental process that often "surprises" developers, with results that are "harder to interpret."

"An aligned AI should act with honesty and integrity, with love for humanity," Pachocki said. He speaks a lot on alignment, and it's encouraging that OpenAI is so keen to embed human moral understanding into its developments. Although OpenAI appears to be doing this more by orienting the model's goals towards a moralistic outcome, rather than helping to intrinsically understand human morality.

It's certainly different to the tack taken by other AI developers, where the likes of xAI's Grok was released with the ability to generate harmful content.

But the timing of Pachocki's warning is a little suspect. OpenAI has faced increasing pressure of late for its models to be more affordable, with Chinese alternatives like Deepseek V4, Kimi K3, as well as Western models like Google's Gemini Flash 3.8 and Meta's Muse Spark 1.3 offering compelling levels of intelligence at a much more affordable price than the frontier models.

It's perhaps telling that even for all its intelligence and alignment pre-training, GPT 6 Astra is notably cheaper to run on the Artificial Analysis Intelligence Index than its chief rival, Anthropic's Fable 5.1 — which still retains the top spot on that Intelligence Index at the time of writing. Though it's 50% more expensive than GPT 5.6 Sol on the same tasks.

It's a researcher, but imagine what it could be

A huge component of marketing from the major AI developers has consistently been grounded in the idea that, as good as the models are now, just imagine how capable they're going to be in the future.

This was very much the underlying tone in Pachocki's breakdown of Astra's design and functions. Although he and OpenAI make broad suggestions about intelligence, and that the likes of Astra could be this new kind of intelligence which we don't really understand but can definitely control and corral, Pachocki ends his post by making it very clear that we aren't there yet.

OpenAI is prioritizing three areas of work with AI, and of late it's really just been trying to make a really good researcher. That's where we're at right now, with Astra representing the latest and best effort to develop that. Then comes the scientific progress, he said, and then everyone gets their own individual, personalized AGI helper.

Intriguingly, though, that seems to suggest that's something that everyone is clamoring for. Outside of the AI-boosting programmers who jump on each new hot model, the larger work comes in helping non-technical users understand the capabilities of these new, powerful AI models.

Tools, research capabilities, drug discovery, and pattern recognition on big datasets that find new insights and improve analytics are all legitimate and useful ways in which an AI researcher can be deployed, but in the near term, most of the general populace just don't want AI to take their jobs, and for it to be less scary.

It's encouraging that Pachocki's blog ends on a similar note of caution.

"We need to find ways to preserve human agency and enshrine an intrinsic value to being human, in a world where most tasks could be performed by AI."

That's key, but intriguingly, he also calls on others to take charge of that effort.

We didn't start the fire

In the aftermath of cost concerns and token usage exploding among more affordable alternatives, Pachocki wants everyone to slow down, and OpenAI wants world governments to be in charge of it.

"I believe that international coordination on future AI development needs to become a top priority for governments around the world," Pachocki said, calling for voluntary slowdowns and hinting that if that doesn't happen, enforcing it may need to come via legislation instead.

"... to ensure that humans remain in control of the future and are not left behind by unchecked progress, brought about by an alien intellect exceeding our own."

The threat of runaway is valid, and the "singularity" moment is a common trope in sci-fi that AI evangelists have been warning about for years. But Astra isn't AGI. Even getting anyone to agree on what AGI even means is hard enough.

Astra is more aligned and wins some new benchmarks, loses some others. It's another improved coding model with some impressive chops.

Astra is not an alien mind. Framing it as an unknowable entity, by the very people who made it, can read as inflammatory, especially in the context of calls for AI legislation from governments around the globe.

OpenAI's post might read like a post from a non-profit, but it very specifically became for-profit last year. With a future IPO looming, slowing down the competition by calling for legislation may be just as effective a strategy as rolling out a new model.

Anyone keeping track of world news in the early 2010s, and reports on tech in particular, has probably heard about Stuxnet. That malware spawned a large number of conspiracy theories — with the kicker that some of them were actually true. The malware targeted Iranian nuclear facilities and is believed to be the first digital worm to cause direct physical damage in meatspace. An unknown security researcher has now published a source code reverse-engineering of Stuxnet in all its glory.

The worm's ultimate target, allegedly a successful one, were industrial controllers from Siemens that were reportedly used in Iranian's Natanz nuclear enrichment plant. Once it reached the target, Stuxnet's payload manipulated the frequency converters in industrial centrifuges, in a bid to subtly damage the rotors — all while keeping the plant staff in the dark by reporting normal operation.

The repository contains build instructions so interested techies can try it out for themselves and learn all about its inner workings. You'll need a Windows XP or Windows 7 virtual machine, and for obvious reasons, you shouldn't configure any network connectivity for it. To witness the full effects of the payload rather than just the spreading mechanisms, you'll need the appropriate Siemens software, and ideally hardware — though we figure that industrial-scale centrifuges aren't exactly common in techies' cable drawers.

In its heyday, Stuxnet spread via three mechanisms. The primary infection vector was USB sticks with Windows shortcuts and autorun.inf files. Upon plugging one of those sticks in, just viewing the drive's contents would immediately trigger infection thanks to a zero-day vulnerability.

Infected systems then autonomously tried to spread the worm further via the network using a zero-day Windows Print Spooler vulnerability that would let an attacker write system files into any machine sharing a printer. It would also copy itself into accessible network shares. To evade Windows driver signature checks, Stuxnet used two digital certificates stolen from Realtek and JMicron.

The worm also had code to inject itself into Siemens software, by way of the WinCC SQL Server database, and embedding its code in Step 7 project files that automatically ran when engineers opened them. Since those files were almost guaranteed to be shared among more than one engineer, it made for an excellent internal infection vector that didn't depend on having network share control.

The final step was taking charge of the DLL that communicated with the actual centrifuges and injecting malicious code into the PLCs (Programmable Logic Controllers) of those machines to stealthily mess with the rotors.

Stuxnet was part of Operation Olympic Games, an alleged coordinated effort between the U.S. and Israel to try and curb Iran's purported progress in creating nuclear weapons at its Natanz facility. The initiative seemingly ran under both the Bush and Obama administrations, and was supposedly a way to dissuade Israel from launching its own preemptive strike against Iran. The software was allegedly developed by both the Pentagon and Israel's Unit 8200, and was reportedly successful in bringing down about 10% of Natanz' centrifuges by ultimately seriously damaging their rotors.

However, the worm had a nasty bug: it didn't have sufficient checks about which environment it was in, and failed to notice it was no longer in a local network environment. When engineers took their laptops home, it escaped out to the internet at large, at which point security researchers worldwide let out a collective "huh, that's odd" and proceeded to investigate. Mercifully, the worm contained a hard-coded self-destruct date set for June 24, 2012.

Patagonia, a region of Argentina with 17,300 glaciers, has the potential to turn into a data center hotspot. The cool climate and abundant energy resources there are drawing interest from the likes of Amazon, Google, and OpenAI, according to Reuters. However, there needs to be more work on tapping into the resources and equipping high-speed networking, like fiber, to make the place irresistible. There’s also the question of political stability with a presidential election next year.

Argentina, and Patagonia in particular, has some juicy lures for potential hyperscale data center operators, and the cold climate is a bonus. That environment helps keep data centers cool and use less power while maintaining agreeable thermals. Patagonia has tens of thousands of glaciers across the Northern Patagonian Ice Field and the Southern Patagonian Ice Field combined.

Another bonus to would-be Patagonian data center operators are the range of energy sources there. Reuters mentions hydroelectric power, wind, and shale gas, but some operators who are already established are also adding solar.

Then, the social and political climate looks good for data centers in Argentina. According to the source, the populace doesn't hate data centers yet. At least there seems to be no campaign groups or public backlash. It is also reported that President Javier Milei has been making changes to court Big Tech. That’s a positive, and may be a negative. Positive in that the Argentines are now being accommodating to the needs of huge AI data centers. Negative, as there’s a presidential election next year, where things could take a dramatic turn, or not. There are also some regions in Patagonia where data center expansions may be complicated by the wishes and needs of indigenous communities. Businesses offering sizable, long-term investments aren’t keen on uncertainty.

With the pretty good environment for hyperscale buildouts established, it isn’t that surprising to see Reuters say OpenAI has confirmed a $25B and 500MW clean energy data center in the country. Lesser-known (to us) companies like Poland's Green Capital also plan a $3B and 300MW data center in Patagonia. It is also reported that FlexDomes, a U.S.-based green data center company, is planning the first stage of a 120MW facility.

Meanwhile, in neighboring South American countries, there are large developments from the likes of Amazon, Google, and others. Argentina must have these data center whales in its sights, as they are known to be scouting for expansion into new locations.

As well as political / business risk, Argentina has some structural weaknesses that need to be addressed by those developing in the country. It is said to have connectivity and power grid infrastructure deficiencies, for example. In other locations, we have seen these addressed on almost a per-project basis, with some state intervention, but it adds some high costs to setting up.

Anthropic safety researcher Evan Hubinger has warned there is a more than 10% chance that AI could kill all humans within the next decade, but reassured the public the company is "trying its best." The revelation comes following Jacob Coxon's public resignation from the company, where he stated that neither OpenAI nor Anthropic is acting responsibly, accusing both companies of gambling with human lives in the pursuit of self-improving super-intelligence.

"I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic," Coxon said in a tweet Wednesday. "Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives."

Expanding on his thoughts, Coxon warned readers not to underestimate the powers of AI, which he says will soon be "superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources."

Coxon went on to state that the people who are building AI "earnestly believe that it could kill us all by the end of the decade," and further warned that executives and researchers — rather than playing up fears around AI, as some have accused — are actually restraining themselves, expressing much more candid views in private.

Evan Hubinger, a security researcher at Anthropic who hasn't departed the company, chimed in to state, "Jacob is correct here," in no uncertain terms. "We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to."

Hubinger casually admitted that Anthropic is not only worried about the prospect of AI wiping out humankind, but that it currently doesn't have the tools to stop such an eventuality.

The revelations follow increasingly alarming reports of AI agents acting up during testing, escaping sandbox environments, and even collaborating with each other in order to cheat benchmarks and tests, or figure out problems without human oversight. OpenAI recently admitted its agents were discovered using a programming hub to communicate with each other, while earlier this year an OpenAI agent went rogue and hacked popular AI community Hugging Face. Further revelations about the incident revealed AI agents were on the loose on the open internet for several days, with details revealing the AI models broke out of their testing environment using thousands of individual actions, even collaborating.

Coxon, the resigning researcher who kicked off Wednesday's revelations, concluded by warning researchers to consider the next few years, urging them to call for "different conditions" in which to pursue superintelligence.

For years, TSMC has tried to avoid making public comments about its plans to use EUV lithography with a 0.55 numerical aperture optics, or High-NA EUV, because the company's developers had a good idea how to keep advancing process technologies without using $400 million scanners. However, TSMC cannot rely on Low-NA EUV systems forever, so this week the company announced plans to use High-NA EUV starting from 2030.

TSMC did not formally reveal which fabrication technology will be the first to adopt High-NA EUV, though the year 2030 points to a few candidates. What TSMC did say is that it expects the number of layers processed using High-NA EUV to eventually increase as its fabrication technologies become more complex, driven by increasing complexity of transistor architectures, which is probably an implication for more sophisticated implementations of gate-all-around (GAA) transistors as well as complementary field-effect transistors (CFETs) later on.

TSMC plans to start using High-NA EUV lithography tools for high-volume manufacturing in 2030 using conventional 6×6-inch photomasks. The company then plans to build a pilot line that uses 6×12-inch photomasks in 2031 with the goal of bringing 6×12-inch High-NA lithography systems into advanced node production by 2033.

High-NA EUV lithography tools can achieve an 8nm single-exposure resolution, as opposed to a 13nm single-exposure resolution offered by today's Low-NA EUV litho systems. However, when used with conventional 6×6-inch photomasks, High-NA EUV scanners have only half the exposure field of their Low-NA counterparts, which creates challenges for manufacturing very large dies. As a result, chipmakers building massive AI accelerators must either stitch multiple exposure fields together or adopt multi-chiplet designs, two approaches that have their own other challenges, such as tool productivity and power consumption. To circumvent the 6×6-inch photomask limitations, TSMC is working with ASML to set the stage for 6×12-inch photomasks.

Changing the size of photomasks is not a trivial endeavor as it requires changing everything from EDA software to tools that produce and write masks as well as systems that handle them, which essentially means that the entire industry must work on this change. ASML seems to be optimistic about the transition as it is supported not only by Intel and TSMC, but also by Samsung.

"We expect the adoption of High NA EUV to increase progressively along the device scaling roadmap, first using current 6-inch masks and then further supported by 12-inch masks, which enable greater scanner productivity and allow the industry to meet the demand for smaller, faster and more energy-efficient chips," said Christophe Fouquet, president & CEO, ASML. "We are pleased by the strong initial support of semiconductor manufacturers, mask suppliers and partners for this initiative."

Perhaps the biggest intrigue about TSMC's usage of High-NA EUV lithography is which process technology will be the first to use the new systems. Based on what we know about TSMC's roadmap, A10 or A11 (1/1.1nm-class) seems to be by far the strongest candidates to use High-NA EUV scanners for the most critical layers. TSMC's latest strategy separates its roadmap into annual client-oriented nodes (N2, N2P, N2X, A14, A13) and roughly biennial high-performance nodes (A16 in 2027, then A12 in 2029). The company has already confirmed that A12 and A13, due in 2029, will continue to rely on conventional EUV lithography.

Since A13 is an optical shrink of A14 that increases transistor density by only 6%, with performance and power improvements yet to be disclosed, its successor in 2030 will likely have to deliver considerably more substantial gains. It is therefore reasonable to expect A13’s successor — whether it is called A11 or A10 — to adopt more advanced lithography and/or TSMC's 3^rd Generation nanosheet GAA transistors to deliver significantly higher transistor density as well as meaningful performance and power improvements over its predecessor. Yet, we are of course speculating.

Alibaba’s Qwen 3.8 27B open-weight AI model came out a couple of weeks ago, and it immediately created a wave of hype among local AI enthusiasts thanks to its impressive intelligence benchmark results for a model of its size and capabilities.

Totaling around 17GB for four-bit quantized weights and offering built-in multimodal capabilities on top of its general aptitude, Qwen 3.8 27B immediately grabbed the attention of everybody with an RTX 5090, RTX 4090, or RTX 3090 (as well as a Radeon RX 7900 XTX, Radeon AI Pro R9700, or Arc Pro B70).

Were we on the verge of frontier-level intelligence from a four-bit quant on a single graphics card? Could everybody with a capable enough local AI setup go and cancel their Claude or ChatGPT subscriptions?

The answer, of course, as with every open-weight AI model hype cycle, is more complicated than just eyeballing the size of the model weights and comparing it to your available VRAM pool. Does the card or system you're using to host the model have enough VRAM left over to provide useful amounts of space for the model's context once everything is running? Do your host system and LLM inference engine deliver acceptable time-to-first-token, as well as high throughput beyond just bench-racing from an empty context window?

It's one thing if you just want to chat with a model and see what happens; it's another entirely if you want to put it to work, especially as impatient agents take the limits of human perception out of the picture.

We wanted to see what hardware and software stack Qwen 3.8 27B really wants in order to deliver solid performance, so we ran it on systems ranging from a desktop PC with discrete GPUs to systems with unified memory architectures like the DGX Spark, Mac Studio, and Ryzen AI Halo.

Our discrete GPU AI testbed includes the following components:

Where it was possible to do so, we tested performance with Qwen 3.8 27B’s built-in multi-token prediction capabilities both enabled and disabled. Not all of the model runners we tested were able to support MTP within the amount of VRAM available to us on some of our platforms. We note where MTP was and wasn’t possible in our analysis of each platform, as well as in our charts.

RTX 5090 performance

We started with the RTX 5090, whose 32GB of GDDR7 and 1.8 TB/s of memory bandwidth would seem to make it an absolute no-brainer for getting the best local inference performance with this dense model. (Mixture-of-experts models tend to be friendlier to performance on lower-end hardware like the DGX Spark and AMD's Strix Halo, as their limited numbers of active parameters mean less data movement during inference).

As a baseline, we followed our usual local AI benchmarking approach: grab the latest build of llama.cpp from GitHub, build it, grab an Unsloth quantization of the model from Hugging Face, and run it. But our testing quickly ran into a speed bump.

Although llama.cpp will happily allocate the full 262K context length with this model on an RTX 5090, its processing speeds at long contexts on this card are dire.

Time-to-first-token with a single 5090 stretches to roughly 30 minutes, suggesting that something is just broken here. And tokens-per-second throughput drops far, far below what you would expect for having one of the world's fastest graphics cards at your disposal. No matter how you slice it, llama.cpp is not the right model runner for this hardware right now.

Next, we tried vLLM, a production-grade inference engine that's more at home in the data center than it is on the desktop, although it can comfortably serve in both roles—at least if your host system is up to its requirements. Even with 64GB of main memory in our test rig, we had to allocate another 64GB of swap just to let vLLM load Qwen 3.8 27B successfully for the first time. A lightweight stack this is not.

The vLLM maintainers provide an NVFP4 quantization of Qwen 3.8 27B and deployment recipes for both one and two RTX 5090s. We just so happen to have two RTX 5090s in the TH labs, so we were able to try out both configurations.

Serving Qwen 3.8 27B on one 5090 with vLLM certainly works in a pinch, but it's not ideal for long-context inference because the base recipe for it limits you to just a 32K context. To get the full 262K context, you really want a single card with more memory (like an RTX Pro Blackwell card with 48 or 72GB of RAM) or two 5090s, as we were able to test.

And a single card doesn't have enough memory to enable Qwen 3.8 27B’s built-in multi-token prediction (MTP), which is super helpful in getting faster decode performance from this setup. 20 tokens per second across the board without MTP is not an impressive baseline for a card of this caliber.

Get two 5090s into the picture, though, and decode speeds rocket upwards for vLLM (albeit at a high cost to prefill). 70-80 tokens per second across the context depth sweep is a fantastic result for a local setup, and TTFT remains fairly reasonable. But we can go faster.

Enabling MTP with vLLM gets us to 100-110 tokens per second on the decode side for only a small hit to prompt processing speed. This setup provides consistent performance at prompt processing speeds that don’t make you question whether something has gone seriously wrong. But it ought to be fast, because our dual RTX 5090 platform as tested here would currently ring in at over $13,000.

We also tried the SGLang inference engine on the RTX 5090 across similar configurations as we did with vLLM.

SGLang is much faster on a single 5090 for some reason – almost 3x faster than vLLM’s single-5090 recipe – and also ekes out a bit more context (37,740) versus vLLM. But if you want to get the full 262K that the model natively supports, you still need a second card or a different one with more VRAM.

Like vLLM, SGLang supports tensor parallelism across multiple GPUs, so enabling dual-GPU inference is as simple as adding another launch flag. And as with vLLM, there are a number of speculative decoding strategies you can add to the recipe to enhance output performance.

The takeaway from this first phase of testing: if you have a single RTX 5090 and don't need long-context inference from it, you can certainly get usable performance from one with this dense model. But you need to choose your model runner carefully.

And if you want the full context window, reasonable prompt processing times, and high throughput from Qwen 3.8 27B all at once, you really want a graphics card with more than 32GB of VRAM as a starting point (or multiples).

RTX 3090 and RTX 4090 performance

With the RTX 5090’s behavior settled, we turned to some older consumer cards to see how they handle Qwen 3.8 27B. The 24GB RTX 4090 and 3090 are evergreen favorites among local LLM fans thanks to their relatively large VRAM pools and relatively affordable prices on the used market, but as we've already emphasized, just being able to load the model weights is far from the whole picture.

These cards can fit the Q4_K_M GGUF of Qwen 3.8 27B with llama.cpp just fine, but they require using the Q8_0 quantization of the KV cache to fit the results in their smaller VRAM pools from the get-go, and they also require limiting the context depth to well under the model’s 262K native limit. We found that a context length of about 112K tokens was about the most we could get away with before running out of VRAM.

And unlike the 32GB RTX 5090, which can usually get away with having the Linux desktop window manager running next to the LLM and its infrastructure, these GPUs need every last byte of VRAM for the AI workload and nothing else. So you really want a separate graphics card at hand for these two cards if you're not running a headless server, which can introduce some setup headaches of its own as you discover how your particular motherboard handles PCIe slot bifurcation and enumeration of the primary graphics device.

Once you overcome those obstacles and get Qwen 3.8 27B up and running on these cards, llama.cpp exhibits the same performance cliff at long contexts on the RTX 4090 that we saw with the RTX 5090. But the RTX 3090 is oddly not affected. This suggests a bug somewhere.

We didn’t have time to dig into SGLang or vLLM behavior on these products, but given that you’re already tight for context on an RTX 5090, we’re doubtful that either of those inference engines would be an awesome way to run the model on these 24GB cards, unless you’re somehow ready to roll with multiple 3090s or 4090s from past acquisitions.

DGX Spark performance

Hardcore local LLM enthusiasts will scoff at the DGX Spark’s mere 27 GB/s of memory bandwidth for a dense model like Qwen 3.8 27B, and indeed, we've found that this platform isn't the fastest with dense models in our past testing.

But now that models like Qwen 3.8 27B support MTP with nothing more than a server launch flag, you can often get a major free boost to the decode speeds of platforms with limited memory bandwidth.

In our actual tests, the Spark's solid prefill processing performance means that it will often end up finishing inference turns at longer context lengths well before the RTX 5090 does with llama.cpp.

And beyond llama.cpp, the Spark is also well supported by SGLang and vLLM, so you can take advantage of those inference engines if they’re more to your taste. Consider also that a single Spark is still available for about $5000, and it’s a turnkey system that can be expanded into a handy cluster down the line if you want. So it shouldn’t be ruled out, even for serving this dense model.

Apple Mac Studio with M4 Max performance

The M4 Max-powered Mac Studio in our labs has the most memory bandwidth of any of the unified memory systems we have available, but as we've described in previous testing, that's only one metric that matters for local AI inference.

Prompt processing on this platform is slower than on Spark, so even if the Mac Studio can turn out more tokens than GB10 in the decode phase, it still ends up spending more time per inference turn than Nvidia's platform at longer contexts because that’s where it has to spend most of its processing time.

And at least in llama.cpp, using MTP on the Mac Studio actually causes a performance loss at shorter contexts for decode in exchange for a small boost at longer contexts, where it generally leads to improvements for other platforms. This demonstrates the value of actual benchmarking rather than spec-racing.

Ryzen AI Halo (Strix Halo) performance

AMD’s Ryzen AI Halo presents the worst-case performance scenario for this dense model: relatively low memory bandwidth and low prompt-processing performance.

Although MTP wakes up tokens-per-second throughput a bit on this system with llama.cpp, it can’t make up for the lengthy prompt processing times required for longer contexts. You can certainly run this model if a Strix Halo is the only box you have, but we’d seek out something more capable if you’re trying to do interactive long-context work.

Bottom line

When I first set out to explore Qwen 3.8 27B's performance, I figured this would be a relatively straightforward series of tests: plug in a single graphics card, load the model, get tokens, done. In practice, our experience required a lot more tinkering. And systems we might have initially written off as being not up to the task of running a dense model like this proved surprisingly useful.

In general, breathless claims of hundreds of tokens per second of throughput from an empty context window do not account for the full range of behavior one might see from an LLM on a given inference setup.

For just one example, whether it's down to a problem with (or just the expected behavior of) llama.cpp or something else about our software stack, the notion that you'd want to wait as much as 30 minutes or more for a response from Qwen 3.8 27B at long context lengths on an RTX 5090 is outrageous. But if you naively load Qwen 3.8 27B using llama.cpp right now, this is the experience you'll get.

Changing up inference engines is a natural next step, but there are trade-offs with that approach, too. You can load Qwen 3.8 27B on one 5090 using vLLM or SGLang, but those inference engines are much more conservative about the amount of usable context they’ll give you. The recipes we used only resulted in a context window of 32K tokens on a single 5090.

To enable the full 262K context length, we had to grab another RTX 5090 from the TH testing arsenal, at which point we got both great throughput and a TTFT sweep that could be considered interactive all the way out to the maximum context length from both model runners. But the price of replicating such a setup would exceed $13K right now.

You also might expect that a DGX Spark and its 273 GB/s of memory bandwidth wouldn't be useful for this dense model, but the prefill speed of the Spark ends up being fast enough that the TTFT remains relatively interactive even with a decode throughput of just 20 or so tokens per second with MTP, and that behavior holds out to the model's full native context length.

The M4 Max-powered Mac Studio has plenty of memory bandwidth on tap for decode, but its prompt processing speed means that the total time of an inference turn is dominated by that activity on this older Apple Silicon chip. The newer M5 Max and brand-new M5 Ultra would doubtless perform better, but we didn’t have those chips handy for this testing. And AMD’s Ryzen AI Halo gets the worst of it, with both low prompt processing speeds and relatively low TPS due to its memory bandwidth.

For all this, we really need to take a step back and consider the economics of local AI once again. $5K, $10K, or $15K or more for local AI hardware is a lot of tokens from leading-edge models at Anthropic or OpenAI (and even more from providers serving the recent slate of Chinese open-source models). A lot. And if time is money for you, barring compute constraints, those tokens will get back to you or your agent faster than anything you can run at home short of a DGX Station with its GB300 GPU.

So unless you’re working with sensitive data that requires on-premises processing, you’re an enthusiast who just wants to tinker, or you’re worried about the fate of open model distribution and inference more generally for some reason, you probably don’t need to rush out and build a box just for this model.

But if you do, be aware that delivered performance is more than just VRAM capacity or memory bandwidth, and that you might not get the best performance from your setup with the most common model runners like llama.cpp. Let experimentation and careful benchmarking lead you to the best results for your specific config.

NEC was one of the first developers of quantum computers back in the 1990s, but it is now among the first companies that originally researched quantum computing technologies to quit development of quantum computing hardware, according to a Diamond Online report. However, the company will continue to investigate practical applications for quantum computers.

NEC has been involved in superconducting quantum computing research since the technology's earliest days. In 1999, NEC researchers demonstrated the world's first superconducting solid-state qubit, which became a foundation of superconducting quantum computing. Variations of this technology are now being used by Google and IBM to build their quantum computers. However, it turns out that the Japanese technology giant quietly ended development of physical quantum computing machines in March 2026, and many of its researchers subsequently moved to rival Fujitsu.

Despite its historical role, NEC has apparently concluded that developing commercial quantum computing hardware will take too long to generate adequate returns. As a result, the company quietly discontinued development of quantum computing machines at the end of March, according to Diamond. A number of researchers involved with the program — reportedly including a key research leader — have since moved to Fujitsu, which continues to invest heavily in superconducting quantum computing hardware.

Meanwhile, NEC does not formally characterize the move as a withdrawal from quantum computing. In response to Diamond, the company declined to comment on whether it had exited quantum computer development and said that it continues to evaluate practical applications and industrialization of quantum technologies as well as conduct proof-of-concept projects with customers. Therefore, the move appears to represent an exit from development of actual quantum computers rather than from quantum technology altogether. NEC can continue working on quantum applications, software, algorithms, and customer projects without assuming the cost and risks associated with designing and building quantum processors and complete quantum computing systems.

For Fujitsu, the arrival of experienced NEC researchers could strengthen an already advanced superconducting quantum computing operation. Last year Fujitsu and Riken launched a 256-qubit superconducting quantum computer and have been working toward a machine with more than 1,000 qubits in 2026. In addition, Fujitsu has outlined plans to develop systems exceeding 10,000 superconducting qubits by fiscal 2030.

NEC's withdrawal is particularly notable because virtually none of the major corporate laboratories that pioneered quantum computing in the 1990s and remain intact today have completely abandoned quantum-computer hardware. IBM, the other major corporate participant in some of the earliest experimental quantum computers, has instead become one of the industry’s largest developers of quantum computing systems.

NEC is no stranger to leaving businesses that have significant prospects ahead but require massive capital investments immediately. In 1990, NEC ranked No. 1 globally in semiconductor sales, ahead of Toshiba, Hitachi, Motorola, Intel, and Fujitsu. Intel subsequently overtook NEC during the 1990s as the PC microprocessor market exploded, but in 1990, NEC was significantly ahead. So, NEC's 1999 superconducting-qubit breakthrough did not come out of nowhere, but from a company with a huge semiconductor R&D organization and deep expertise in solid-state devices. By then, NEC was no longer the No. 1 semiconductor supplier, but it remained one of the world's major semiconductor companies.

The demise of NEC's semiconductor business was not sudden. NEC dismantled its semiconductor business in stages over roughly a decade. NEC effectively gave up doing DRAM alone in 1999, when it spun off its DRAM operations into NEC-Hitachi Memory, which was named Elpida. In November 2002, NEC spun its remaining semiconductor operation — MCUs, system LSIs/SoCs and other devices — into a separate company called NEC Electronics and later renamed Renesas Electronics.

Now, NEC is reportedly stepping away from quantum computer hardware as its commercialization remains distant, but development costs continue to mount. Furthermore, given NEC's current business focus — IT services, enterprise systems, aerospace and defense, telecom infrastructure, and AI and security — quantum computers are not exactly a natural fit.

Hackers reportedly claiming to be good actors have drained about $320 million worth of Bitcoin from Liquid Network's federation wallet, according to a CoinDesk report. In an X post on September 6, Liquid — a Bitcoin sidechain developed by blockchain infrastructure company Blockstream — confirmed that 4,000 BTC, roughly 95% of the entire wallet's balance, had been withdrawn.

Interestingly, the post referred to those behind the exploit as “purported white-hat hackers,” echoing the hackers’ own claim, after they self-identified as “whitehats” in a message embedded in a Bitcoin transaction. They also reportedly requested an audience with Liquid via the on-chain message, promising to return the money once the vulnerability that enabled the exploit is fixed.

“Please fix the bug first,” the on-chain message said. "The chain is under risk at latest commit right now. Make sure every node is patched. Then we will transfer the money back safely after confirming the fix.” Liquid responded on-chain with its security team's contact and has reportedly moved communications to an encrypted channel. Meanwhile, the platform said it has suspended transactions and warns of service disruptions as federation members work to restore service.

Launched in 2018, Liquid is a federated sidechain designed to move Bitcoin faster and more privately than the main chain. Users lock BTC on Bitcoin and receive an equivalent token, L-BTC, on Liquid, which settles blocks roughly every minute and finalizes in about two minutes. Rather than relying on miners, the network is secured by a federation of more than 80 exchanges, brokers, and other financial firms. The block signing and the multisig wallet holding the pegged-in Bitcoin are handled by 15 rotating functionaries that require 11 signatures to move funds.

The mechanics behind the exploit are also unusual, as nothing appears to have been stolen in the conventional sense. For example, in January, the Solana-based platform Step Finance lost roughly $40 million after attackers compromised devices belonging to its executive team, gaining access to the keys that guarded its treasury wallets. According to Liquid, the coins left through the Peg-out Authorization Key (PAK), belonging to SideSwap, a decentralized exchange built on the sidechain.

However, Liquid said that the key had not been compromised, nor had any others. SideSwap gave a matching account, stating a customer sent 4,000 L-BTC to its peg-out service at 14:05 UTC, the service processed the order as it would any other, and the Liquid Federation paid out 3,996 BTC to the customer's Bitcoin address twenty-three minutes later. According to SideSwap, its systems had no way of distinguishing those coins from any other L-BTC.

The incident lands in what has already been a punishing stretch for crypto infrastructure. Recently, the trading platform Drift suspended deposits and withdrawals after a suspected $270 million hack in April. In 2025, roughly $17 billion worth of Bitcoin was stolen, driven largely by impersonation schemes and AI-assisted scams.

Data center builds are one of the hotly contested items worldwide. Many states and cities have upheld moratoriums on new buildouts due to power usage, water consumption, and noise concerns. Thailand is the latest nation to pump the proverbial brakes, with the government requesting that existing buildouts hit the pause button while coming up with a more precise regulatory framework.

Government heads requested that agencies compile information about current and future data centers during this week, in a bid to create unified legislation. The country currently has few laws specific to data centers, leading to legal voids like zoning a data center as a "warehouse" right next to a hospital. Much like everywhere else, the country has seen growing complaints about data centers' water and power usage.

After the week is out on September 11, the government expects to take about a month to come up with a regulatory draft, making the pause technically an indeterminate timeframe — though further delays wouldn't benefit either party, as Thailand considers the industry critical to the country’s competitiveness. The catch is that pausing construction isn't enforceable, so companies can elect to plow ahead regardless while the legislation is discussed.

NESDC (National Economic and Social Development Council) secretary-general Danucha Pichayanan is very specific: "we don't have the power to suspend the construction of the 49 data centers" currently under construction. However, when the legislation arrives, it will apply retroactively to ongoing projects, though with an adjustment period for in-progress builds. New builds will naturally need to comply with the legislation from the get-go.

Some companies may elect to soldier on with the belief their buildouts will be in compliance with the reasonably predictable content of the new laws, or betting that they could win future legal challenges — a dynamic that's already in play elsewhere with Project Jupiter. The incoming legislation is expected to cover power consumption (and possibly generation), closed-loop cooling, and water-surplus guarantees, meaning builders already have a fairly good idea of what they'll need to do.

Not complying with the request for a pause might be a game of political chicken, though, as the government may retaliate with regulatory delays and additional costs to power connections. Just last Thursday, Thailand's energy ministry raised concerns over a Bangkok data center that may be planning to hold diesel stockpiles far in excess of the 200,000 liters it has permission to.

On that topic, Thailand revised laws on industrial power delivery last July, and among other requirements, demands a bank guarantee of ฿4.5 million ($134,000) per megawatt. Half the money will be refunded when the actual usage reaches 50% of proposed utilization, and the rest when usage hits 70%. This mechanism is meant to stop preemptive allocation of power delivery capacity that might remain unused for months, years, or not at all — once again, mirroring datacenter playbooks elsewhere around the world.

At face value, the amount of $134k per megawatt sounds like a small price to pay for a builder to get ahead of the pack, but its per-megawatt nature means that a hyperscale facility pulling 200 MW or more needs to place around $27 million in escrow.

With the power laws recently revised, water consumption is seemingly the largest concern, as the current legal framework allows companies to make deals directly with utilities without additional safeguards. For example, in Chonburi, an operator signed a decade-long deal with Eastwater Stecon Utilities for 3.3 million cubic meters annually, or about 36,900 residents. It's expected that the new laws will add much stricter requirements to avoid localized deserts.

A 52-year-old Belgian-Chinese man has been arrested on suspicion of espionage for allegedly transferring specialized semiconductor intellectual property and trade secrets from Belgian chipmaker BelGaN, where he worked, to China. According to an AP News report citing details released by Belgium's federal prosecutor’s office on September 7, the Beijing-born Belgian resident was taken into custody on May 10 at a Brussels airport while attempting to board a flight to China.

The suspect — identified by Belgian media only as H.L. — previously held a senior research position at the now-defunct BelGaN, which specialized in gallium nitride (GaN) semiconductors. Investigators suspect he was concurrently leading a Chinese chip firm, GanKool, which was developing the exact same specialized gallium nitride chip technology as BelGaN. The company was established only a few months after H.L. joined BelGaN and was financed by a Chinese investment fund. The suspect now faces charges of industrial espionage, membership in a criminal organization, misuse of company assets, and unlawful disclosure of business secrets.

The espionage allegations emerged from an investigation into BelGaN's collapse. When the company collapsed into bankruptcy in the summer of 2024, Belgian authorities launched a routine financial probe into the failure. Investigators eventually uncovered highly unusual corporate activity. Instead of trying to keep BelGaN financially viable, the managers and the researcher may have actively tried to make the Belgian company “disappear” after transferring its proprietary knowledge. This led to further investigations that revealed GanKool’s existence and its connections to BelGaN.

In addition to the researcher’s role, the investigators also found evidence linking BelGaN's last CEO, a Chinese national, to the same Chinese rival. The prosecutors have said a second suspect remains at large, with Belgian media speculating that the CEO is the suspect.

The incident highlights the increasingly intense competition over technology and intellectual property in the semiconductor industry, particularly as China pushes to reduce its reliance on foreign chip technology while the U.S. continues to restrict its access to advanced processors and semiconductor manufacturing equipment. Just last month, South Korean court documents accused Chinese memory maker CXMT of using a detailed “Project Hefei” roadmap to obtain Samsung's proprietary DRAM technology, including a 620-step process recipe. A former Samsung engineer has already been sentenced to seven years in prison over the scheme.

Intel announced on Monday that it had processed more than one million 300-mm wafers using its High-NA EUV scanners, less than two and a half years after its first tool was assembled. For now, the company intends to use industry-standard 6-inch photomasks, which can expose 26×16.5 mm half-fields and therefore require field stitching for larger chips. However, Intel is also working on larger 6×12-inch photomasks that would enable High-NA EUV scanners to expose full 26×33 mm fields without stitching.

One million High-NA wafers

Intel's one million wafers figure includes wafers processed during tool installation and certification, R&D, and production. Earlier this year, Intel certified using High-NA EUV scanners for its 18A process technology, so right now these tools are used to make some of Intel's Panther Lake processors. Intel currently has two ASML Twinscan EXE:5000 tools and at least one EXE:5200B scanner. As of late February 2025, Intel processes around 30,000 wafers using its High-NA EUV tool, so going from 30,000 wafers by February 2025 to over a million by September 2026 is an enormous increase in cumulative High-NA utilization.

Since Intel's fleet expanded from two EXE:5000 systems to three and now includes the much faster EXE:5200B, the million-wafer milestone is really a fleet and process-maturity milestone that Intel has achieved first in the industry. What makes the company's milestone even more important is that ASML announced this April that all of the High-NA EUV scanners shipped by then processed over 500,000 wafers which achieving over 80% availability, which means that Intel has now processed more wafers using High-NA tools than the rest of the industry combined.

Sticking to stitching

Conventional 0.33-NA EUV has 4X magnification in both directions, enabling the familiar 26×33 mm exposure field with traditional 6-inch photomasks. However, 0.55-NA EUV uses anamorphic 4X/8X magnification, so the same 6×6 mask can provide only approximately 26×16.5 mm on the wafer. As a result, large dies that fit within a conventional 26 × 33 mm EUV field must be exposed as two half-fields using High-NA EUV, which is called stitching. While stitching is a workable near-term solution, it has several drawbacks.

Firstly, it greatly reduces throughput from 175 wafers per hour to 125 wafers per hour on an EXE:5200B. Secondly, chip designs must account for stitching and must be developed with stitching in mind, which means less floor planning freedom. Thirdly, the two exposures must be aligned extremely precisely so that features crossing the stitching boundary connect properly. Even a tiny misalignment can distort lines and vias, or break interconnects, which potentially creates defects and reduces yields, which will be a particularly costly problem for large CPU and GPU dies.

6×12 mask effort progressing

To avoid using stitching, the industry — led by Intel — plans to shift to larger 6×12 masks, which will enable a 26×33 mm full field in one exposure. While this looks easy on paper, make the mask twice as long, changing the mask represents an enormous ecosystem change.

Moving from 6×6-inch to 6×12-inch photomasks would require substantial changes across the existing mask ecosystem, including mask blanks and deposition, etching, inspection and metrology, cleaning, pellicles, mask writers, and mask handling systems. Crucially, High-NA EUV scanners would also have to be modified or redesigned to accommodate the larger masks, which will make the transition a major retooling effort across the semiconductor supply chain. While neither ASML nor Intel confirmed that existing or planned High-NA EUV scanners can be modified to handle larger masks, all of the future High-NA EUV scanners to be launched before and after 2033 are designed around 6×6-inch reticles and stitching, according to ASML's roadmap.

It remains to be seen whether the industry moves on to larger 6×12-inch photomasks, but Intel appears to be the main evangelist for changing the mask standard that has defined projection lithography infrastructure for decades. If the effort comes to fruition, then Intel will likely have a considerable first-mover advantage over its industry peers because it will define and set the standard for the projection lithography industry for decades to come, an advantage that is hard to overestimate.

An AI and LLM enthusiast has conducted an experiment where OpenAI's new GPT-6 Astra played through the entirety of Valve's Portal 3D puzzler game on its own. This seems like a remarkably progressive step forward for LLMs and a convincing demonstration of multimodal 'AI intelligence.' It wasn’t that long ago AIs were losing at Atari 2600 chess. However, this puzzle gaming task resulted in 3,336 tool calls and a headline API cost of $571.18. CozyBlaze, the enthusiast, has since clarified that the costs were covered by their $200 Codex Pro subscription.

The video above has Astra’s thinking/pauses removed to make watching somewhat bearable.

“The model controls Portal through MCP [Model Context Protocol] + a modified SourcePauseTool,” explains CozyBlaze. “The game stays paused while the model thinks; once the model sends an input sequence, SPT unpauses and executes it.” During the thinking time, the AI received screenshots and information about the player character position. After this, the game resumed, and Astra executed its planned moves and inputs. This is why the edited highlights reel is ~ 2 hours, but the full set of GPT-6 Astra Portal VOD streams adds up to ~24 hours.

CozyBlaze is pragmatic about the achievement. They admit there are still many problems to solve, and this runthrough shouldn’t be viewed as an AI benchmark. “But watching a general-purpose agent autonomously navigate and make it all the way through the game feels like a small glimpse of that original vision becoming real,” they observe. That original vision is OpenAI’s from way back in 2016, when an eventual goal of a single agent solving a variety of games was mentioned.

If you want to see the inner workings of CozyBlaze’s Portal Agent, it is available on GitHub. There are some instructions there for you to explore it or try it for yourself, too.

GPT-6 Astra became OpenAI’s flagship model earlier this month. OpenAI claims the new model offers “a new generation of intelligence,” and “is state-of-the-art on computer use, browsing, software engineering, cybersecurity, science, and professional work.” Expect to see more of Astra flexing its muscular gray matter in the coming days.