Month: January 2026

The situation with RAM prices is so desperate that there are already those who build their own memory at home

by

The crisis we are experiencing with RAM memories and its exorbitant price is shaking the technology industry in multiple ways, precisely because they are components that found in the vast majority of devices that we use in our daily lives. Faced with the crazy prices, there are users who have not given up and have resorted to extreme solutions: building their own RAM memories. Untenable. Memory prices have skyrocketed to unsustainable levels. DDR5 modules that previously cost between 100 and 150 euros now easily exceed 350 euros in many markets. You can blame the AI. And the demand for DRAM for artificial intelligence applications has absorbed a large part of global production. OpenAI alone has accounted for 40% of all productionleaving home users paying the consequences. The worst thing is that it doesn’t look like this is going to be solved soon. And it is that according to analysis firm IDC, the shortage could last until 2027. cheap adapters. In recent weeks we have seen how some users have chosen to build their own RAM memories to face the price crisis. One of the approaches is to use SODIMM to UDIMM adapters. In a video The Hardware Canucks YouTube channel shows how they have tested this solution on Ryzen 7000, 9000, Intel LGA 1700 and LGA 1851 systems without too many problems. The approach is simple: buy DDR5 SODIMM modules (the ones for laptops) that are still relatively cheap and connect them to the system using adapters that cost between 10 and 15 euros. It must be said that this method also has its limitations, and that is that the data transfer speed that these adapters achieve is quite a lottery. In the tests carried out by Hardware Canucks they say that some do not exceed 4,800 MT/s stably, while others reach 5,600 MT/s or even 5,800 MT/s, although it depends a lot on the adapter model and the platform. In terms of performance, the good thing is that the difference is practically imperceptible. According to the content creator’s tests, with an RTX 5090 and a Ryzen 9800X3D, the difference is between 5 and 7% in the worst case compared to conventional DDR5-6000 memory. A more radical solution. Another approach is the one that the Russian modder VIK-on has opted for. And just as they count From Videocardz, the enthusiast has built a functional 32GB stick by combining chips from two 16GB SODIMM modules from SK Hynix, a Chinese PCB, and a heatsink from AliExpress. The total cost: 17,015 rubles, about $218. As explained in the media, in Russia an equivalent module costs at least three times more. Images: VIK-on After physically assembling the parts, a process that requires BGA reballing stations and considerable soldering experience, the modder then integrated ADATA firmware to enable an XMP profile that allowed the memory to run at a speed of 6,400 MT/s. In this way, VIK-on has achieved a functional 32 GB stick that any motherboard could recognize and that according to the modder works stably in games. Between the lines. That making a RAM from home is economically viable says a lot about the state of the market. Furthermore, not everyone experiences the situation in the same way, since in some markets such as Russia prices are especially prohibitive. Of course, soldering memory chips is not trivial, as it requires specialized equipment, technical experience and taking risks of damaging expensive components. The adapter method is much more bearable, but it is most likely that these homemade solutions will continue to be a niche. Most users will prefer to pay the extra price rather than risk soldering components or dealing with third-party adapters. Although if the forecasts end up being true and the crisis extends for several years, the emergence of a secondary market for professionally refurbished modules taking advantage of surpluses is no small feat. It would probably be the last solution we would resort to, but in other markets there might be enough demand. Cover image | Andrey Matveev In Xataka | The computers of the future have found an unexpected ally to store information: fungi

Mining Bitcoin has always been an energetic black hole. Someone wants to turn it into your home heating

by

The CES 2026 that has just closed its doors has confirmed an inescapable reality: Artificial Intelligence is everywhere, even where it seems to make no sense. From electronic LEGO bricks and wearables with roll-up screens, to more questionable devices like AI hair clippers that adjust the cut dynamically or digital frames that generate art by voice using GPT Image 1.5. Among this tide of “AI even in the soup”, a proposal has emerged that breaks with that trend and has surprised by its pragmatism: is it possible to get hot by mining Bitcoins? The answer is a resounding yes, and this year technology has shown that what was once a nuisance thermal waste is now a valuable household resource. ANDl income generating water heater. The American startup Superheat captured everyone’s attention with the presentation of its Superheat H1a water heater that uses ASIC (application specific integrated circuits) chips to heat a 190-liter tank while processing Bitcoin transactions. Unlike traditional electric water heaters, the H1 has an approximate price of $2,000, placing it 30-40% above the conventional market. However, as detailed in CNETwill be able to generate about $1,000 annually in passive income, always depending on the value of Bitcoin and the difficulty of the network. The science of “thermal reuse”. To understand this phenomenon, you have to turn to basic physics. The mining process requires intensive computational calculations (proof-of-work) that generate a massive amount of heat. Traditionally, this heat was expelled into the air using fans, but companies like Superheat have turned it around: mining is now the primary function and hot water is the secondary benefit. From the user’s point of view, the experience is seamless. The manual for devices like the Heatbit Trio reveals a control system sophisticated where the user can navigate the panel like a professional: Eco Mode: Heats exclusively by mining, limiting consumption to 400W. Target Mode: Combines the mining plate with a conventional heating element to maintain the desired temperature. Air purification: These devices not only heat, but act as purifiers with HEPA filters and air quality sensors (PM 2.5). Europe at the forefront. In the old continent, the proposal focuses on design and structural integration. From Austria, the company 21energy presents the Ofen 2a minimalist design radiator made of steel and aluminum. Unlike industrial miners that emit 75 decibels, this model is around 32-35 dB, being almost inaudible to the human ear. Furthermore, with a consumption of 1000 watts, it generates up to 40 TH/s of mining power, allowing users to recover part of their electricity bill while heating rooms of up to 50 m². On the other hand, in Switzerland, the company RY3T has marked a historic milestone. The RY3T ONE system has already been installed as the main heat source in a single-family house in the canton of Sankt Gallen. According to the companythis system can be more environmentally friendly than a conventional heat pump, as it reuses a computing power necessary for the global financial network instead of requiring additional electricity exclusively to generate thermal friction. A good idea or a technological chimera? Despite the enthusiasm, a report from Interesting Engineering raises critical questions that the consumer should consider: Obsolescence: What happens when mining hardware becomes obsolete? Will the entire heater or radiator have to be replaced? Network Cost: Even though heat is “free,” electricity for Bitcoin mining is often more expensive than natural gas in many countries. Regulation: If a country decided to ban Bitcoin mining, the user’s heating system could be legally compromised. From mining coins to processing Artificial Intelligence. As this report began, AI is the main protagonist of the moment, and its evolution will continue to be talked about far beyond cryptocurrencies. Julie Xu, COO of Superheat, explained at CES that the ultimate goal is to use this network of appliances for cloud solutions and AI inference. Instead of building gigantic data centers that stress the power grid and require massive cooling, homes could house small distributed computing units. However, this future poses a new dilemma: privacy. Experts from iFixit and Consumer Reports They already warn at this CES that “you don’t want a camera in front of your refrigerator watching you all the time” or a constant internet connection on simple devices, since it makes them more expensive to repair and prone to failure. The challenge will, therefore, be to balance the profitability of heating the home with the security of our private data. Image | freepik and heatbit Xataka | The bitcoin business cools down, but some miners have found a new vein: AI fever

It is raining so much in the province of Jaén that the olive oil harvest has had a problem: there is too much water

by

The “liquid gold” market expected a great recovery after years of drought, but the data you have given the Food Information and Control Agency At the end of December 2025, they have had a significant impact. Especially in the epicenter of oil production in our countrysuch as Jaén, where it has been registered a 45% drop in its accumulated production. Although it is something that hides an important economic paradox: it is selling more than ever. The figures. As detailed by the Ministry of Agriculture itselfthe reality of the current campaign is radically different from the previous one. While in 2024 Jaén accumulated almost 300,000 tons at the end of the year, this 2025 it has remained at half speed with 164,841 tons, which represents a variation of 45.3%. Something that has also been noticed at the national level. What has happened? Although everyone might think that we are talking about the drought that has caused there to be fewer olives, the reality is that excess rain has been the problem. The intense rainfall of November and December 2025, although beneficial for the tree in the long termhave been an obstacle to the harvest. Logically, with the mud it is difficult to enter with the machines to be able to pick the olives or work by hand. This has caused the harvest to be delayed and has affected the yield of the fruit. Other factors. Beyond the excess of rain at the end of this year, we must also highlight the high temperatures that were recorded in the month of June 2025, which damaged the weight of the fruit after spring fruit set that promised a lot, but fell short. Besides, according to COAG Jaénthe delay in taking the olive to the olive mill due to the weather has caused part of the fruit to suffer damage, reducing the final yield. Less oil, but more sales. Even though the silos fill more slowly, the market is extremely active. UPA Andalusia has highlighted that, despite the decrease in production, sales have increased by 10% in the last quarter, with a month of November where oil output reached 129,727 tons. This means that the consumer continues to demand olive oil despite the instability of recent years. Exports are also doing well, with a substantial increase of 44% in Andalusia, which puts pressure on current stocks, which are 13% lower than last year. The price. Without a doubt it is the most important point for the consumer, especially when in the past we have already seen really high prices for olive oil due to a bad harvest. Logic dictates that if supply falls and demand increases, prices should increase, but experts call for considerable caution. Right now, the price of Extra Virgin oil at origin moves between 4.20 and 4.29 euros per liter, and what is expected is that it will remain at a stable price during the year 2026, without major drops to maintain the stability of the sector that needs to cover costs. Images | Kostas Morfiris Nazar Hrabovyi In Xataka | Half of Spain has gone crazy with the question of whether olives make you fat or not. But your biggest problem is not calories.

They were carried out with “passenger” planes

by

An idea is beginning to gain traction among many military experts. The episode takes place in the first attack of the US campaign against Venezuelan vessels that Washington linked to drug trafficking, a coup that occurred on September 2, 2025 in the Caribbean that left 11 dead. Everything indicates that it was with something very similar to a “civilian” plane. A non-military aircraft. Yes, the singular thing, according to the information of New York Times and the Washington Postis that for that first action an aircraft with a paint scheme and silhouette not identifiable as “typically military” would have been used, to the point of seeming a civil or passenger aircraft. Not only that. It is suggested that he also carried the ammunition inside the fuselage instead of displaying them on external supports. Also because the device emitted a military identifier via transponder, although it is not clear what real capacity a vessel of that type would have to perceive or interpret it during contact. The double impact. The initial attack was marked by the decision of hit twice: after the first impact there were survivors who managed to remain on the remains of the hull and who later died in a second attack which ended up sinking the remaining structure. The boat reportedly changed course after detecting the plane and, at some point later, the survivors appeared to gesture toward the aircraft, without it being clear whether they understood what had caused the initial explosion. This moment is what fueled much of the public debate because it enters into especially sensitive issues of the law of armed conflicts, in particular the treatment of people in vulnerable situations after a sinking. Perfidy. The legal point currently being discussed is whether the use of a civilian-looking aircraft to carry out an offensive attack could fit into the figure of perfidyunderstood as feign protected status (like the civilian) to obtain a tactical advantage, causing the adversary to lower his guard or stop taking precautions. Various specialists cited in the media they explain that the analysis would depend on factors such as the degree to which the aircraft was truly “unidentifiable” as a combatant, whether it came close enough to be seen by those on the boat, and whether that appearance could have influenced its behavior (for example, not evading, not surrendering, or not taking survival measures). It is also highlighted that the obligation to distinguish military forces from the civilian population It is a reiterated principle in manuals and doctrinal guides of the United States. The official position. The administration defended the legality of the attacks, presenting them as part of a confrontation against drug trafficking and associated violent activities, and maintained that the actions were consistent with the applicable law. In parallel, the Government argument that there is an armed conflict against certain criminal groups and cartels, a statement that has been discussed by different experts, but which serves as a framework to present the targets as “combatants” within that category. From the Pentagon, public responses focused on pointing out that the systems and platforms used undergo legal reviews and internal validation processes, and the exact model of the aircraft used in the initial strike was avoided. The mystery of the plane. It has not been identified with certainty which aircraft it was, although hypotheses and indirect clues: from models derived from commercial aircraft as variants of the Boeing 737 in military service, to devices with clear liveries and minimal markings, occasionally seen in operational environments. It has also been confirmed that the Navy operates P-8A Poseidon (based on 737 and with very clear schemes) and that there are other 737 military transport aircraft, as well as more discreet aircraft with civil registration plates linked to corporate structures little transparent. At the same time, is remembered that open flight trackers do not necessarily show all military activity, so the absence of conclusive public identification does not allow specific platforms to be confirmed or ruled out. The technical factor. They counted the analysts at TWZ that the technical plausibility of an airplane “that looks like a passenger” but at the same time can attack is based on already existing solutions to integrate ammunition in a discreet way, especially through the so-called Common Launch Tube (CLT), a system that allows loads to be housed and launched from inside an aircraft without the need for visible supports. The known capabilities The CLT can accommodate munitions and small drones (including light glide bombs, compact missiles and unmanned vehicles) and mounts that can be integrated into ramps, doors or internal compartments, with reloading options and tactical flexibility. The approach has been associated with platforms such as AC-130Jweapons kits on KC-130J and drones such as the MQ-9, suggesting that, by design, it would a priori be adaptable to a wide variety of aircraft, including those that due to their external appearance can be confused with civil or transport devices. Image | x In Xataka | The US did not need to shoot to enter Caracas. All it took was an invisible weapon and unexpected “help” from Russia In Xataka | The mission in Caracas revealed that the best kept secret in the US is not a drone: it is called DAP and you will not see it in the movies

The housing crisis in the Pyrenees opens the debate on the limits of camping

by

“I am a temporary worker and I come to work in the Tena Valley. The rents are 800 euros and I am not going to share a flat. I prefer to live in the van with my cat (…) I don’t understand why they focus on me when I don’t break the regulations.” The words are those of an indignant woman. One to which the Civil Guard demands movement when it understands that it is camping illegally. One that, if what is said in the video is true, is completely right and the regulations are on its side. What happened? The video is brought to X by a user of the social network and in it you see a woman who, between irony and indignation, explains that she works as a seasonal worker in the Tena Valley (Huesca), that she lives in her van and that the Civil Guard is asking her to leave because, supposedly, she is camping. The video shows the conversation between the woman and one of the agents. He defends that “it’s been going on for a month and a half now” and that “that can be considered camping.” She, for her part, alleges that she does not take any items outside and that she also moves, which the agent also acknowledges. “Rents are 800 euros”. In her video, the protagonist points out that the cost of housing is very high in the Tena Valley and that she does not want to share a flat. “I prefer to live in the van with my cat,” he emphasizes. The truth is that this enclave right next to the Pyrenees is very tense. In fact, it was only a few months ago that he was born. Decent Housing Viello Aragón in defense of access to housing in this Aragonese area. They point out that the area is full of tourists. The problem is not just a matter of underused second homes. The group points out that the European funds dedicated to sustainable tourism are causing the attraction of more and more visitors with the conversion into apartments and houses as rural complexes, which worsens the conflict. They defend that it should be stopped the delivery of new licenses for tourist use, they propose the creation of a public housing pool, declaring the region as a stressed area in order to control prices and punish the owners of more than four residential properties who have empty homes. The final objective is to facilitate access to housing and so that the economy can diversify so as not to live solely on tourism. Yes, the law is on your side (sort of). The truth is that if the author of the video complies with what she says, the law is on her side. As explained by the colleagues of Motorpassionthere are no regulations that specify a maximum time in which a person can spend the night in their car. Because camping in a place not authorized for it is illegal but spending the night is allowed. The PROT Instruction 2023/14 It is the one that collects these differences. It states that “parking is not camping” as long as: That the vehicle, with the engine stopped, is only in contact with the ground through the wheels (stabilizing legs or any other device is not used, except for chocks, provided for by the General Traffic Regulations). That the vehicle does not occupy more surface area than what it occupies when closed, that is, without the deployment of projectable elements, chairs, tables, etc., elements that can invade a surface larger than that delimited by the perimeter of the vehicle, understood as the plan projection of the same. That the vehicle does not emit any type of fluids or noises to the outside. One but. And then, the instruction also states the following: All of the above will be understood without prejudice to the powers of the town councils, through their municipal regulations, to limit or regulate, without discrimination based on the type of vehicle, the stopping and parking points under criteria of physical organization of traffic, commerce or environmental criteria, or with the purpose of favoring the arrival of this type of motorhome tourism, establishing for this purpose, parking zones or areas or, where appropriate, camping. Likewise, the regional regulations that have been approved or could be approved for the purpose of tourism promotion in the national territory will be taken into account. to promote a new type of accommodation that in any case will coexist with the full application of national legislation on traffic and road safety and especially the regulations relating to stopping and parking maneuvers. The matter here is a little more delicate because the video does not explain the situation regarding any of these concepts. What the instruction makes clear is that the City Council can delimit where a motorhome can park. In the images it appears that the van is located on the outskirts of a town but it is not clear whether or not the vehicle may be parked there, specifically. Without knowing the municipality, we also do not know if the town’s mobility ordinance imposes a maximum parking time. In the case of not wanting to park in the town, the situation is more delicate. In this case, regional regulations require you to spend the night in an area expressly authorized for this purpose since the Tena Valley is a protected environmental space. What do we get clear? In a municipality, a person can spend the night without any problem in their car or caravan as long as the vehicle is parked correctly and they do not take belongings and objects outside (from chairs to awnings). In that case, the driver is considered to be camping and this can only be done in an area authorized for this. Therefore, living inside a van and making life in it is not illegal as long as no noise or fluids are generated that are emitted to the outside. … Read more

India. It will no longer be possible for another reason: Europe

by

In recent years, cars are not the only thing that has risen in price disproportionately. The general escalation of prices has affected also to the world of motorcycles and, to try to contain even greater increases, in recent years the most common practice by large manufacturers has been to move part of their production to India to save costs. The EU has not been amused in the slightest. New tariffs. The month of January starts with new tariffs on motorcycles and scooters made in India. And not a single model manufactured outside European territory is spared. 8% tariff for models up to 250cc Tariff of 6% for larger cylinder capacities In a market with tight margins, a tariff of this amount will force large European manufacturers to completely rethink their strategy. Who is affected?. KTM, Triumph, BMW, Suzuki, Aprillia and mainly Royal Enfield will be the main affected after this measure by the European Union. KTM manufactures its small displacement motorcycles in India, such as the KTM 390 Duke and Adventure. BMW and its best-selling contender, the GS 450is manufactured together with TVS Motor, an Indian conglomerate. Aprillia manufactures together with Bajaj its 457. Triumph has been allied hand in hand with Bajaj for years to manufacture its stars of displacement 400. Why is it important. Given the increase in prices in recent years, manufacturers made a very clear bet: to commit to manufacturing medium-low displacement models in India. Seeing premium brands like Triumph or BMW launch 400 models that aspire to be bestsellers is a true reflection of the state of the sector. The key? Manage to sell competitive models between 5,000 and 6,000 euros, in a segment ideal for everyday use and occasional outings. The 6% tariff will shake this pricing strategy. Because now. Although the measure seems like an attack from Europe on external manufacturing, the context goes further. The EU has the Generalized System of Preferences (GSP) to regulate the flow of exports from developing countries to Europe. But there is fine print: The EU requires certain minimum standards when manufacturing the product. It also requires that, if certain export volume thresholds are exceeded, preferential tariffs end. Although there is no official data on how many motorcycles manufacturers are exporting from India, the threshold has been exceeded, and with it tariffs are once again applied without any type of bonus. The great beneficiary. China is sweeping Spain with economical models and much more equipped than the rest of its competitors, and China will continue to sweep if tariffs negatively affect the main European manufacturers. With Zontes making its way as the third manufacturer in our country, the countdown to see a podium headed by motorcycles made in China. Image | bmw In Xataka | The 11 cheapest electric motorcycles with the most autonomy: the best quality-price options

The Spanish business that Vodafone sold as ballast is now worth three times as much. Zegona has shown that the problem was the owner

by

according to further Populi Voicea medium with a good track record in telecom exclusives, Telefónica has started talks with Zegona to acquire Vodafone Spain. The negotiations are recent (just a few weeks) and it was Movistar who picked up the phone first. Telefónica wants to close the operation in the first half of 2026. The rumors come from months ago. The problem is that arrive late, and that has a price. A little more than two years ago, Zegona bought Vodafone Spain for about 5,000 million euros. Vodafone (the British parent) was selling a problematic asset: It was the third operator in a market of four. He was caught between the scale of Telefónica and the agility of the low-cost He inherited a network that required constant investment. And he also inherited a tarnished reputation after years of complaints. For the British group, Spain was a drain of money and effort. For Zegona, a poorly managed gold mine. And in just two years, the fund has proven that he was right: Has returned to its shareholders 1.4 billion euros in dividends (28% of what was paid by Vodafone Spain). Has reduced the number of shares in circulation by 69%. And yet its current capitalization is around 3.6 billion. For fund shareholders, the return has been spectacular: The stock went from 345p when they bought Vodafone (less than 100 when they announced their intentions) to over 1,565p now. It has multiplied by 4.5 in two years. Vodafone Spain generates around 4.5 billion annual revenues and, with more focused management than before and without the bureaucracy of a global giant, it has become a profitable operation that Zegona can continue to exploit… or sell to the highest bidder. Telefónica is now negotiating from a weak position. It needs the operation (Marc Murtra has repeated that Movistar must lead the consolidation of the Spanish market) and the market knows it. An ERE of 4,500 people has just closed. And while Telefónica prepared the house to add more furniture, its price has fallen 27% since the end of October. Zegona, however, its value has skyrocketed. The price of this indecision is between 2,000 and 7,000 million extra euros. regarding what the purchase of Vodafone Spain would have cost in 2023. Zegona is in no hurry. It can wait, it can squeeze, it can even stay as it is. Telefónica now cannot afford that luxury because buying Vodafone Spain is not an expansionist move, it is an almost defensive necessity: needs critical mass before Europe forces further consolidation where Movistar is the main course, not the diner. But when negotiating is a necessity and the other side knows it, the price stops being a variable and becomes a toll. If the operation crystallizes, it will create a giant with more than 45% of the Spanish market, great cost savings by eliminating duplications (headquarters, networks, contracts…) and intense regulatory scrutiny from Brussels. Although not as brutal as it would have been with Vestager because Ribera has another look. Telefónica knows it and so does Zegona. The difference is that one is late and the other can afford to wait. That changes everything in a negotiation. In Xataka | The great dilemma of Spanish telecos: either they become giants or China swallows them Featured image | Vodafone, Telephone

There is a material on which the future of the iPhone and AI depends. And almost everything is manufactured by the same Japanese company.

by

More than 100 years ago two Japanese textile companies called Fukushima Boseki Co., Ltd., and Katakura Seishi Iwashiro Bosekisho they joined forces to become Nitto Boseki Co. Ltd, also known as Nittobo. A century later we have encountered a giant on which a critical material for the future of our chips depends: glass fabric. Technological glass artisans. The Japanese company was the first in industrially producing carbon fiber. They did it in 1938, almost right at the same time as Owens Corning Fiber Glass in the US. Later, in 1969, they developed the “crystal fabric” or “glass cloth” (glass cloth), a material that began to be used in printed circuits Hello, T-glass. That material evolved and in 1984 they launched their T-glass, an even more specialized glass fabric that began to be used as a substrate in chips of all types. This material is different from the common fiberglass like that used in surfboards or in insulation solutions. Thus, it has a very low coefficient of thermal expansion, which ensures its good performance even when the chips are operating at maximum performance. Japan, we have a problem. As indicated on Nikkeiexperts warn that the lack of this material has become a major obstacle to chip manufacturing and the advancement of AI in 2026. Nittobo is practically the only company in the world capable of manufacturing this glass with the necessary quality. Its glass fabric is extremely thin, bubble-free and heat-resistant, which has made it a fundamental part of chips such as those used in iPhones. Apple, in fact, was one of the first major technology companies to reach an agreement with Nittobo to use this material. Everyone loves Nittobo. The good performance of this material has now made companies like NVIDIA, Google or Amazon also demand T-glass for their chips, and that has generated a worrying competition due to inventory that is quickly depleted and it is not clear that it can cope with demand. Apple asks for help. The situation is so tense that Apple has sent some managers to Japan and has even asked the Japanese government to intervene to ensure supplies from Nittobo. Once again the objective is to guarantee the launch of its key products, and at Nikkei they point directly to the expected foldable iPhone. The fiberglass fabric is a critical layer on the chip substrate and ensures that everything works perfectly even under heavy workloads. Source: Nikkei. Capacity will grow, but not immediately. At Nittobo they know very well what the situation is like, but they can’t do anything to remedy it, at least in the short term. A company executive quoted in Nikkei indicates that “if we do not have additional capacity, it means that we do not have additional capacity no matter how much pressure is put on Nittobo. The way I see it, the situation will only improve significantly when Nittobo’s production increase becomes a reality in the second half of 2027.” Looking for alternatives. Apple and Qualcomm are looking for plans B, and their initiatives to find new suppliers in China or Taiwan are already underway. However, the demand for the quality of this type of material is very high: an error in the quality of the glass of the chip substrate cannot be repaired, and would ruin entire batches of components. AI causes chaos again. We already saw it with memories: the AI ​​industry needs immense quantities of DRAM and NAND memory chips, and that has now meant that the rest of the world is suffering from a huge rise in prices. The same thing is happening with this glass fabric: AI chip manufacturers have an exaggerated demand for this material, which harms the rest of the “traditional” chip manufacturers and, therefore, the users. bad business. And as happens with memories, in the end the material is sold to the highest bidder, which are usually companies like NVIDIA that have exceptional profit margins. That leaves consumer electronics manufacturers in a vulnerable position and with declining sales forecasts. Nittobo does not want to saturate the market. And as happened with the memory market, Nittobo does not want to oversize its business in the face of this demand and prefers to be cautious. Japanese suppliers already suffered losses from overstocks in 2022, so they are now reluctant to expand their factories aggressively. It is precisely the same speech that Micron made, which already suffered from excess inventory after the pandemic: although they could now manufacture more memory chips, for them that means risking history repeating itself. In Xataka | A thousand-year-old mystery allowed us to put nanotechnology into modern screens. Today the discovery has a Nobel Prize

Chinese startups have been relying on NVIDIA chips to train their models for years. That is already changing

by

The name of the Chinese startup Zhipu AI (Z.ai) may not sound familiar to you, but perhaps GLM, its AI model, does a little more than its latest version, GLM-4.7already competes with Claude Sonnet 4.5 or GPT-5.1. The real surprise of this “Chinese AI tiger” is the launch of GLM-Image…and not so much for what he does, but for how he has managed to do it. what has happened. GLM-Image is a multimodal generative AI model that focuses on image generation. The idea, of course, is to compete with options like Nano Bananafrom Google. That’s interesting, but even more striking is the fact that the model has not been trained with conventional chips. Trained with Chinese chips. According to those responsible for Z.ai, this model is the first developed in China that has been fully trained with “local” chips. Specifically, it has been trained with Huawei’s Ascend chips thanks to the use of servers Huawei Ascend Atlas 800T A2 and a framework called MindSpore. Thus, traditional NVIDIA AI chips, which are usually the usual choice for AI model developers in Chinese startups, have not been used. Turning point? This milestone demonstrates the real feasibility of training high-performance generative AI models on a platform developed entirely in China. We are not dealing with something minor: it is validation that it is possible to continue innovating in this area despite the restrictions imposed by the US. In fact, Zhipu AI — included last year on the US blacklist — has intensified its collaboration with other local manufacturers, such as the promising firm Cambricon that has risen from the ashes thanks to tariffs. Threat to NVIDIA. The news comes at a unique time, because NVIDIA has not stopped pressuring the US government to once again allow it to sell its advanced AI chips to Chinese companies. He has obtained that permission—which It won’t be free—, but now the one that might not be interested is China, which he hasn’t said anything at all. That chips from companies like Huawei are a valid alternative for training quality AI models can change many things in this area. Zhipu goes like a shot. The Chinese startup has also just gone public, and since it has done so its shares they have shot up more than 80%. Investors see the company no longer as a rival to Google or OpenAI, but as a banner. One that shows that it is possible to compete without depending on the US and its companies. Huawei, great beneficiary. If the trend continues, Huawei can become the Chinese NVIDIA, and the company prepares an increase in production of its AI chips. It is not the only one: Cambricon plans triple your production by 2026, which seems to make it clear that the Chinese industrial machinery is moving quickly to neutralize the impact of US vetoes. Challenges…Despite everything, Zhipu already has warned that the price war in the AI ​​sector will become international. If Chinese companies end up controlling the entire chain (or rather, their chain), they could offer AI services at much lower costs than their Western competitors, who must pay NVIDIA’s margins and Big Tech’s cloud infrastructure. …and unknowns. This technological achievement raises other questions. One of the most important is how powerful and capable Huawei chips are compared to NVIDIA’s in these processes: is training much slower? Is it more expensive in time and resources? The efficiency of the MindSpore framework compared to Pytorch or TensorFlow is another of the key components of these developments. In Xataka | Faced with the US strategy, China has a plan to revive its technology industry: that AI belongs to everyone

Michel Foucault was convinced that “visibility is a trap.” And without knowing it I was talking about our lives with AI

by

I never thought I’d write this, but I’ve been thinking about it for days. Michel Foucault more than I would like. And a back pain is to blame. It was a couple of weeks ago, it was one in the morning and the house had been quiet for a while. That’s where the puncture came. I could have woken up my wife who was 30 centimeters away and, well, she is a doctor; I could have searched on Google; I could have even asked on an Internet forum. And yet, I opened ChatGPT, asked what was bothering me, and shortly after turned off my phone to go to sleep. And I fell asleep right away. But a few days ago, this analysis by Javier Lacort about ChatGPT Health It left me thinking. Not because AI was fully entering the world of health and “medical advice” (something that, on the other hand, I knew firsthand); but because of something that was commented on in it: that “we prefer to ask a chatbot have to wait three weeks for an appointment or have to bother a friend at eleven at night. It hurt a little. There was something interesting there. Eleven at night; one in the morning “The ChatGPT Competition”, Lacort continued“it’s not so much with the doctors as with the emotional support network that we used to have. We asked our mother, our partner, the friend who studied nursing.” But for some time now, “upsetting someone has become emotionally costly.” That last phrase is devastating because it contains the key to something that goes far beyond chatbots with medical uses. Something that goes through Millennials’ problems with calls, with the fishmongers, with sex or with any interaction that is not mediated by a screen: the deep cultural aversion that the modern world has generated to ‘social friction’. And it is curious because, although only in recent years do we see the most striking consequencessociology and cultural analysis have been pointing out what was happening for decades. We have Norbert Elias, for example, who I was convinced that (as part of the prolongation of the civilizing process) the thresholds of shame and discomfort are shifting. What fifty years ago was perfectly normal—calling without warning, asking a favor from an acquaintance, interrupting someone with a question—today borders on the intrusive. What’s more, today we have internalized it. Sennet spoke of the decline of the public sphere (we know how to handle ourselves in privacy and in public transactions, but not in the middle ground); the sociology of emotionstalks about the success of therapeutic lexicon and how that has changed the way we relate; Hartmut Rosa cblame social accelerationprecariousness and lack of time, the loss of effectiveness of reciprocity networks. That is to say, we have many theorists thinking about the same thing: that we are a new type of subject. A subject who has internalized the rules, who manages himself, who evaluates his relationships in terms of emotional cost-benefit and who, above all, experiences direct reciprocity as something frictional, uncomfortable and potentially invasive. And, just then, chatbots appear. I’m not talking about the technology behind it, nor its ultimate nature: I’m talking about the same historical process that has created subjects like this, has created something that “listens to them”, that “is empathetic”, that does not judge them and that helps them as and when it can. Honestly, it would be strange not to throw ourselves into his arms. Can Foucault help us understand all this? Google DeepMind That’s where, I’m afraid, Foucault becomes interesting. In his courses at the Collège de France from the late 70sthe French philosopher explored a whole series of different dimensions of power that, although not obvious, were inseparable from the Modern State. In the past, the State was mainly about controlling borders and collecting some money. But not anymore: now the State manages populations (what it called ‘biopolitics‘ and includes things such as vaccination programs or birth policies) and, at the same time, deals with each subject in its particularity (the so-called ‘pastoral power‘ who through family doctors, social workers, school counselors or psychologists listen to us, advise us and “lead us”). He called the combination ‘governmentality‘: a power that (excuse the ‘expletives’) is at the same time totalizing and individualizing. And those, totalizing and individualizing, are features that seem half-made of technological solutions such as ChatGPT Health. A chatbot that, on the one hand, advises users about their problems, listens without judging, guides us in micro-decisions and knows us (or ‘pretends to know us’) in our particularity; and, on the other, it performs triage, implements protocols, normalizes thresholds, generates aggregate data and, in a short time, will integrate with insurers and health systems. Pastoral and biopolitical, at the same time. And with an incredible infiltration capacity. The difference, and this Foucault could not foresee, is that now this power does not depend on the State, but on a corporation. What was previously a community or ecclesiastical function, then partially state, is now outsourced to private, for-profit infrastructures. It is a privatization of power. The tentacles of the State In the previous section I said that “Foucault could not foresee it”, but I think that is not accurate. It is true that when this thinker theorized about “pastoral power” or “biopolitics,” he was thinking about public officials operating in state institutions. But the wickers were there. After all, Foucault himself, in his last courses (especially in ‘Birth of biopolitics‘, dedicated to analyze ‘neoliberalism’ as arts of government), described a decisive mutation of our time: the State no longer thinks of itself as a provider of services but as a guarantor of the conditions for the market to function. The functions that were previously assumed directly (educate, heal, advise, care) can be outsourced to private agents. In this sense, chatbots are neither an accident nor a distortion; are the logical culmination of the historical process of the development of modern power. From a very specific formulation of … Read more

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.