compete

Japan was the king of semiconductors in the 80s. Rapidus is its only hope to compete in this market again

by

In the 1980s, Japan did not compete in semiconductors and technology. It was devastating. In 1988, Japanese companies controlled more than half of the world semiconductor market, and NEC, Toshiba, Hitachi and Fujitsu were above giants of the time in the US such as Motorola, Texas Instruments or Intel. That golden era ended with the hyperspecialization that emerged both in South Korea and China and (especially) in Taiwan, but now Japan wants to make a splash again. what has happened. A year ago the technology industry was surprised by the birth of Rapidus Corporationa company born from the alliance of several Japanese giants (Sony, Toyota, SoftBank) with the aim of returning to Japan part of its relevance in the field of semiconductors. The initial plan was very ambitious: they wanted to jump directly to 2 nm by 2027. As we will see later, they have had to delay that forecast, but what has also changed (a lot) is the structure of the company. Japan like main investor. The Japanese government has decided to make Rapidus a centerpiece of national security, and is taking unprecedented control of the company. He will become the largest shareholder, although initially he will only exercise 10% of the voting rights to leave management in private hands. Of course: the State reserves the right to raise that participation above 50% if the company is experiencing difficulties. Total capital has skyrocketed to 420 billion yen ($2.7 billion), when in 2022 the investment did not exceed 50 million. The golden action. The Japanese executive has made use of a legal mechanism by acquiring the so-called “golden shares” with which he can exercise his veto in critical decisions such as changes in management or mergers. The objective is to shield Rapidus against foreign capital acquisitions and guarantee the sovereignty of the project. Which is exactly the same thing we are seeing around the world, of course: each country wants to have its own apples in its basket. Investors who are also clients. Financial support comes from the Japanese government, but also from some large Japanese business groups such as the aforementioned Sony and Toyota or Denso. In total, 32 companies have invested 167.6 billion yen (1.075 billion dollars) and will contribute to this commitment by also being customers of the silicon that Rapidus can produce. They remain just as ambitious… or more. Rapidus CEO Atsuyoshi Koike has adjusted the development plans for his chips, and has delayed the arrival of mass production to March 2028. That’s bad news, but not so much when we discover that the company has plans to go beyond 2nm and is preparing to be able to make 1.4nm chips and even 1 nm. Fast as gunpowder. One of the factors that want to differentiate Rapidus is its promise of rapid delivery of semiconductors. The project aims to automate both the manufacturing, packaging and testing of the chips. These last two are processes with great manual intervention, but at Rapidus they believe they have the key to making them much more autonomous. If they succeed, they could reduce the cycle time of semiconductors by 66% and thus beat even giants like TSMC by the way. Japan turns to chips. Japan’s aspiration is striking, and its Prime Minister, Sanae Takaichi, seems to be clear that the commitment to this segment must be notable. In fact, Japan is investing a proportion of its GDP (0.71%) in semiconductors much higher than that of the US (0.21%) or Germany (0.41%). Challenges. The strategy, of course, has its critics. Takero Doi, professor at Keio University, point “There are many cases in which public-private investment has led to systems that lacked accountability. It is important to clarify who will lead the project, the private sector or the government.” Plan B. Although the plan with Rapidus is ambitious, the country is actually playing both sides. While boosting its own business, the government has made commitments with TSMC to upgrade its manufacturing plants in Japan. This makes it have a hybrid ecosystem: it attracts the experience and knowledge of the semiconductor giant while on the other hand trying to create a national alternative. Image | Xataka with Freepik In Xataka | Panasonic was the bastion of 100% Japanese TVs after Sony’s step back. Now it has surrendered to China

Telefónica is already selling its minicenters to compete in the era of real time

by

For years they have told us that the future of artificial intelligence lies inincreasingly larger data centersmore powerful and more demanding in energy consumption. And it’s true that computing muscle matters. But there is an equally determining factor that is talked about much less: distance. In the era of real time, it’s not just how much you process that matters, but where you do it. Every millisecond that data takes to travel can disrupt the ability to react instantly. This nuance, apparently technical, is beginning to become a strategic issue for Spanish companies. Telefónica’s bet. The company has activated the commercialization of its edge computing services for B2B clients in five Spanish cities, Madrid, Valencia, Seville, Bilbao and A Coruña, as part of a broader deployment that includes 17 nodes in this initial phase. This means that companies and administrations can now hire these processing and storage capacities close to the point where the data is generated. Closer data. Edge computing involves processing information where it is generated, rather than constantly sending it to distant data centers. As Microsoft explainsis about moving computing and storage capacity to peripheral network locations, such as factories, stores, offices or distributed infrastructures. In practice, local devices and servers analyze and filter data on site and only send what is relevant to central systems. The goal is to reduce latency, alleviate network traffic and enable real-time responses, complementing rather than replacing traditional cloud. The deployment. Telefónica’s Edge Plan plans to reach 17 nodes in this first phase throughout this year. According to the company, 12 infrastructures are already deployed: to the five with active B2B services, other nodes are added in Madrid, Barcelona, ​​Málaga, Palma de Mallorca, Valladolid, Terrassa and Mérida. This same year, the incorporation of Zaragoza, Las Palmas de Gran Canaria, Gijón, Santa Cruz de Tenerife and Santiago de Compostela is planned. Many of these facilities are located in old copper plants converted into Edge centers, adapted to availability and security requirements. Basic and Smart. Telefónica does not sell “edge” in the abstract, but rather two concrete ways of using it. The first is Basic Edge, a stable layer that brings computing capacity closer to the territory and focuses on data control and compliance with national, regional or local regulatory frameworks. Each node acts as an availability zone, allowing applications to be deployed with additional guarantees of continuity and resilience. The second is Smart Edge, which introduces dynamism: selection of the most appropriate node at all times, creation of instances on demand and operation with FTTH or 5G SA connectivity depending on the scenario. Beyond physical infrastructure. Telefónica integrates computing capacity with GPUs into its portfolio for artificial intelligence loads, available as a service and deployed in Edge nodes. This allows companies and institutions to run high-performance models without purchasing their own hardware and maintaining processing within the defined regional environment. The company also mentions the incorporation of RAG agents and capabilities to adapt models to specific contexts. Overall, the strategy seeks to bring AI closer to data under criteria of sovereignty and regulatory compliance. When the millisecond rules. An example helps to dimension the scope of this architecture. Telefónica developed with CAF a pilot that combines Edge and 5G Stand Alone for the railway sector, providing artificial vision solutions that process data close to the asset instead of depending on centralized infrastructure. According to the company, this approach avoids installing processing nodes in each car and keeps responsiveness at levels compatible with real-time operations. Images | Xataka with Gemini 3 Pro In Xataka | We had suspicions, but Sam Altman has confirmed it: AI is just an excuse to fire

Samsung’s new way to compete with OLEDs without using organic pixels

by

Samsung has been building its own strategy for premium televisions for years based on avoiding traditional OLED and exploring alternative paths. First it was QLED, then Mini LED and, more recently and almost reluctantly, QD-OLED. At CES 2026, which has already kicked off, the South Korean firm has added a new milestone to that roadmap with the presentation of your first 130-inch Micro RGB TVa non-OLED technology, neither MicroLEDbut that seeks to differentiate itself by bringing the color directly to the light source. Samsung’s proposal is not just about a new type of panelbut of a strategic move to expand differentiate its premium television catalog beyond OLED, at a time when a good part of the industry is converging on very similar solutions. What is Micro RGB and why it is not just another LED In current LCD televisions, even the most advanced Mini LEDs, the backlight is based on a series of white or blue LEDs arranged in a matrix located behind an LCD panel in which different color filters are combined. The tMicro RGB technology breaks with that approach by changing those white or blue light LEDs for an array of micro-LEDs that directly emit reds, greens and blues independently, each of them with a size of less than 100 micrometers. That is, color is no longer generated by filtering and decomposing white light, but is emitted directly from tiny RGB bulbs located on the panel. This reduces light intensity losses, improves color purity and allows much more precise zone control of light, even though the image still passes through a conventional LCD panel. The result is a color volume much higher than that of traditional LCDs (Samsung claims that it offers complete coverage of the BT.2020 space used in the film industry), and a brightness capacity greater than that offered by OLEDs, which continues to be one of the strong points of LCD compared to organic technologies. Micro RGB vs. QD-OLED: two paths to the premium market The comparison between Micro RGB and OLED (in any of its variants) is inevitable, especially since Samsung already sells televisions QD-OLED like the S95F. At this point, the difference is not so much visual quality as technological approach. OLED and QD-OLED are self-emissive panel technologies, capable of turning off each pixel individually. This guarantees absolute blacks, extreme contrast and a uniformity that is difficult to match because the level of control over lighting is much more precise. Micro RGB, continuing to rely on a backlight system, cannot replicate that behavior: there is always some residual light, especially in very dark scenes, but the tiny size makes that lighting control has improved a lot compared to the MiniLED and even the Micro LED. In exchange, Micro RGB eliminates the risk of burn-in inherent to OLED screens, allowing a higher level of brightness to be achieved and offering greater long-term stability. These are relevant advantages in very large televisions, in intensive use or in bright environments, where OLED continues to have physical limitations, although its performance has greatly improved with the latest generation panels. More than a replacement, Samsung presents Micro RGB as a second premium path, parallel to QD-OLED, with different strengths and aimed at different audiences and formats. One of the challenges of Micro RGB is to manage thousands of RGB emitters efficiently. To this end, Samsung has announced at CES the development of new processing engines such as Micro RGB AI Engineresponsible for analyzing the image in real time and adjusting brightness, color and contrast by area in these new panels. This advanced processing seeks to minimize classic LCD effects such as blooming and improve detail in shadows, without promising absolute blacks of a self-emissive panel. It is a clear example of how the leap in quality no longer depends only on the panel, but on the electronics that govern it. From 130 inches to the living room: the challenge of scaling Samsung already presented in 2025 a television with this technology with a diagonal of 115 inches, but at CES 2026 it has taken a step further in the development of Micro RGB panels, growing to 130 inches. This giant screen format It works more like a technological showcase of the potential of Micro RGB technology with which Samsung gains muscle demonstrating that already in its first panel versions it is capable of reaching sizes where OLED cannot reach, but it is not the final destination of the technology. Samsung has made it clear that its intention is to bring Micro RGB to smaller sizes starting at 55 inches throughout 2026, something that fits with its differentiation strategy compared to other manufacturers focused almost exclusively on OLED. Here appears the main obstacle: the cost of production. As it is a new technology, with high precision RGB micro-LEDs and complex assembly processes, manufacturing a Micro RGB panel is more expensive today than producing a conventional OLED, even than the QD-OLEDs found in the brand’s S90 range. So that Micro RGB has a real commercial journey in household sizes (between 55 and 77 inches), Samsung needs to make manufacturing cheaper and simplify processes. Only then will it be able to compete on price with high-end OLED and Mini LED, something that will not happen immediately. A strategy to not depend on a single technology As its implications are analyzed, Micro RGB fits as part of Samsung’s broader strategy: not to rely on a single panel type in the premium segment. Samsung already competes in QD-OLEDmaintains a strong position in Mini LED and now adds a third option that combines brightness, color volume and scalability in size. Micro RGB does not seek to dethrone OLED, but rather to occupy its own space between premium LCDs and OLED technologies. If Samsung manages to reduce costs and move this technology to smaller sizes, it will not only change its catalog, but also the current balance of a market that seemed to have opted almost entirely for the new advanced OLED technologies. In … Read more

when Group C appeared on the streets because they wanted to compete on the circuits

by

Speaking of cars, my father has always told me “why do you want 200 HP if you can only go 120 km/h?” Someone had to say that to the manufacturers who, in the nineties, registered beasts with more than 600 HP designed for driving on the street. Le Mansbut with which someone could go on a picnic on a Sunday morning. They are the heirs of Group C. And they could only have been possible in one era: the 90s. Supercars with license plate The world of motorsports has a lot of rules when we talk about competition. Logic tells us that technological advances should result in increasingly faster and, above all, powerful cars. However, the organization that is in charge of regulating all this four-wheel motor competition, the FIA, has been imposing a series of rules so that the power does not get out of control. The Lancia Delta S4, the Ford RS200, the Peugeot 205 T16 and the Audi Quattro, legendary group B rallies An example we saw it in the rally world. The category is extreme, with cars that accelerate like a racing motorcycle and display enormous speed. However, in the 80s, manufacturers began to modify both the engine and the chassis, taking it to the extreme and creating spectacular machines. Accelerations from 0 to 100 in two seconds on land. It was truly crazy. In five years, cars advanced a lot and what had to happen happened: uncontrolled power, maximum competition and pressure, insufficient safety measures and some negligence caused fatal accidents. One of the most remembered is that of Portugal in March 1986, when Joaquim Santos’ RS200 lost control and ran into a crowd, killing three spectators instantly, putting a fourth in the hospital and injuring around thirty people. In May of that year, those who died were those who were driving the car. Toivonen and Cresto lost control and fell off a cliff. The FIA ​​decided that would cut off the development of Group B because, directly, it had gone too far. And if I tell you all this nonsense it is because, in parallel to this extreme development of rally cars, Group C was also emerging. It was in 1982 when this group was introduced, designed for the competition of purely prototype sports cars. While in other categories the FIA ​​limited the engine displacement, braking power, in Group C the limitation came due to fuel. They were endurance racing cars. and control was achieved through 100 liters of capacity with a minimum of five refueling stops every 1,000 kilometers. That allowed 600 liters per 1,000 km. A stupid thing. The FIA’s intention was for manufacturers to limit themselves to improving power through turbocharging. For 20 years, Group C cars put on a show at endurance races and Le Mans, with legendary machines and racing technologies. Formula 1 who were adapting to that competition. The result? Perfect machines that reached average speeds above 200 km/h in Le Mans and peaks of 330 km/h in the Mulsanne straight. But after two glorious decades, the FIA ​​did what it does best: change everything and distort the competition. Within six years, the organization announced that it wanted non-turbo engines and races of 430 km at most (when before they were 1,000). That completely distorted the competition and the meaning of Group C. Furthermore, although the new engines would supposedly be more economical, developing them from scratch would be a great effort for the teams, so they abandoned them, and before the start of the 1993 season, the competition and the category were cancelled. This is how the GT1 was born and manufacturers like Toyota, Nissan, Porsche, Jaguar and Mercedes found themselves with hundreds of millions that were going to waste. And all this context for the girito: unless they took advantage of those supercars that, with a couple of changes, they could register and sell as a street car, taking advantage to finance the development of the cars of the newborn GT1. The Mercedes CLK that had nothing CLK, the most exclusive Nissan and the flying Porsche Taking advantage of this technology and development, the companies used the prototypes created for Le Mans to give life to a series of street supercars that shared many characteristics. They used to be carbon fiber monocoques, they had very high-power engines with sophisticated electronic management, transmission made for racing, active aerodynamics in some cases, very low weight and, in some cases, space for a cabin suitcase. The Porsche 962 When brands like Nissan, Toyota or Mercedes raced in Group C, they didn’t need to manufacture vehicles with street versions: they only focused on the most untamable beasts. However, heThe GT1 category required the production of some registrable units before validating the racing prototypes. The companies took advantage of some regulatory loopholes to get racing, but that need to have a street version caused wild racing cars to circulate directly on the streets. Our colleagues from MotorPassion They have reviewed some of the most representative specimens of this crazy period, and some stories are unbelievable. Heirs of the Porsche 962 Dauer 962 Le Mans It was one of the most representative cars at Le Mans and its chassis was taken as a reference by three manufacturers. One was the Dauer 962 Le Mansa car modified with the help of Porsche itself that had Kevlar panels, a flat floor for stability, a second leather seat, hydraulic suspension and a trunk in the front. The engine had 730 HP and, as it was one of the firstachieved approval by producing only 13 copiesnot the 25 street prices that would be requested later. How did they manage to homologate a racing car so that it could circulate on public roads? Through a hydraulic suspension that allowed the car to be raised up to 10 centimeters and, after passing some emissions and crash tests, the German ITV gave the go-ahead. There were some more heirs from 962, such as Schuppan 962R of which only … Read more

November has been a black month for consoles. They no longer compete against each other, but against TikTok

by

November 2025 is a month that many video game lovers have “celebrated” as the twentieth anniversary of Xbox 360. Pure nostalgiabecause beyond memory, November 2025 will be remembered by Sony, Nintendo and Microsoft as a black month. The reason? It is the worst November for console sales since, precisely, November 2005. In the recent report from Circana we can see a figure very striking: 27%. That is how much spending by American consumers – the largest market for video games – has fallen during November of this year compared to November 2024. Another fact: with 1.6 million consoles sold that month, it is the worst November since 1995, the year of launch of PlayStation. It is relevant because it was released a year ago PS5 Pro and this month of may nintendo switch 2but the high prices of both machines and video games, which have experienced a rise in recent months, have not been able to convince players. Not even on Black Friday. And the key here may not be that thousands of video games are released every month or the price of the console itself. The key may be that the console war has ended and a very different one has begun: the attention war. Console war? War of attention At Xataka we have discussed the topic on more than one occasion. Our ability to focus is broken. in a task due to the enormous amount of stimuli to which we are subjected. Everything competes for our attention. Matt Booty, one of the Xbox heavyweights, said A few months ago Xbox’s competition was not PlayStation. Neither does Nintendo. The competition was TikTok. It was not a mistake, since Satya Nadella, absolute boss of the American company, also stated that “the competition of video games is not other video games, but short format video.” The interesting thing is that it is not an unreasonable statement. To Netflix, especially as a result of the final season of ‘Stranger Things’you are being accusing of producing empty series so that people have them in the background because they assume that they will be consuming short videos on their mobile phones while watching the series. That’s why there are short dialogues and a long opening exposition at the beginning of the season so you can “forget” about having to follow anything else. We are in a moment in which we spend the day unfocused, without being able look at your phone every 15 minutes as a reflex act and where we have to look for strategies so that multitasking and division of attention does not affect us. Matthew Bell, one of the most influential analysts of the video game market, already told it a few months ago in his book ‘The State of Video Gaming 2025‘. In your radiographypointed out how the video game industry no longer competes against itself, but against a tremendously fragmented digital entertainment ecosystem. Our time is finite. If we take away the hours of work, rest, transportation and food, we have little time for the rest. In the United States, there are studies who do not agree on how much time an average user spends on TikTok. The data varies between 58 minutes and 95, but whatever it is, then there is YouTube, Instagram or Facebook. This, in addition, is having cognitive consequences: have less attention span than a fish has. If in 2020 the average human attention span was twelve seconds, now it is eight. A fish pays attention for nine seconds, and you have to pay attention to a video game. And the threat of TikTok? The AI There are those who are catching this situation on the fly and that is why microdramas have appeared. At first, the fever of series with one-minute episodes occurred in Chinabut is climbing. And it is logical that you think: if the competition for consoles is TikTok, who is TikTok’s competition? The answer is also easy: artificial intelligence. Those minutes in which users They use ChatGPT as if it were their psychologista virtual friend or even a coupleare minutes that are not spent on TikTok. It is still something as accessible as opening an app, exactly the same as on TikTok, but perhaps waiting for that response from an AI that is characterized by being tremendously flattering is more comforting for our brain than the umpteenth quick video created with… AI –the slope-. Because they are social networks, video platforms, consoles, artificial intelligence and even the metaverse – someday, if that -, the objective is the same: to keep us glued to the screen. And we cannot attend to everything. In Xataka | An unknown console has overtaken Xbox in sales: it is just the beginning of more ambitious plans

Since Iryo and Ouigo compete with Renfe, we have had ultra-cheap high-speed tickets. Everything has an end

by

There is a problem in the supply of high-speed trains in Spain. We believed that with the arrival of competition to Renfe we ​​would see ticket prices reduced. This has been the case during the last four years and in different regions, but now the three operators have begun to raise their rates in most corridors during the third quarter of 2025, according to the latest report of the National Markets and Competition Commission (CNMC). The change of trend. In the Madrid-Barcelona corridor, the busiest in the country, prices rose by an average of 25% compared to the same period in 2024. Iryo led the increases with an increase of 52.9%, placing the average ticket at 63.82 euros. Renfe AVE raised its fares by 13.3% to 70.58 euros, while Ouigo, traditionally the cheapest option, increased its prices by 20.2% to reach 51.86 euros on average. The context that explains the rise. The withdrawal of Avlo, Renfe’s ‘low cost’ brand, from the Madrid-Barcelona corridor in September after cracks were detected in the bogies of its Avril trains, has reduced the supply of tickets economical on the most popular route. This has caused the remaining operators to adjust their rates upwards. Despite the increase in prices, tickets are still 26% cheaper than before the liberalization of the sector in 2020, as indicated by the CNMC. The exception: Andalusia. In this Autonomous Community, the evolution is different. The entry of Ouigo in January 2025 on the Madrid-Seville and Madrid-Málaga/Granada routes caused a price war which has kept rates down. On the Madrid-Málaga route, only Iryo raised prices (+2.6%), while Renfe AVE lowered them by 8.9% and Avlo by 15.3% to compete with the 32.54 euros on average offered by the French operator. In Madrid-Seville, the average price fell by 2.8% despite the fact that individual operators such as Iryo (+12.5%) and Renfe AVE (+0.9%) did make their tickets more expensive. The Levantine corridor. Regarding routes to the Valencian Community, these show moderate increases. In Madrid-Valencia, prices rose by 1.3% to 30.56 euros on average, the cheapest ticket on the entire network. In Madrid-Alicante, the increase was 1.5% to 37.96 euros. Iryo was the one that increased its fares the most on both routes (with increases of 24.6% and 23.9% respectively), while Ouigo maintained its low price strategy with slight reductions. The thing is about profitability. The Minister of Transport, Óscar Puente, has been publicly demanding this price increase in recent months, going so far as to accuse Ouigo of operating at losses and dragging Renfe into an unsustainable dynamic. And while the rest of the operators have been gaining ground, we are now at a point where they are looking for the economic viability of their operations, and that is where the price increases come in. The balance of passengers. Despite price increases, demand remains robust. High speed will reach almost 40 million travelers in 2024, 77% more than in 2019, before liberalization. In the third quarter of 2025, routes such as Madrid-Málaga/Granada (+17.7%), Madrid-Seville (+13.2%) and Madrid-Alicante (+8.9%) reached record passenger numbers. Only Madrid-Barcelona registered a slight decrease of 0.3%, possibly weighed down by the withdrawal of Avlo and the increase in fares. The future of the sector. After four years of aggressive offers by the rest of the high-speed operators, the sector seems to be entering a phase of maturity in which it seeks to attract travelers without losing sight of the sustainability of the business. Renfe maintains a market share of 62% in most corridors, although in Madrid-Valencia it is already at 50%. It will be very interesting to know the figures in the coming quarters to know how the panorama evolves. Cover image | Jose Garcia Nieto In Xataka | High speed in Madrid is at risk of collapsing. And that’s why Adif wants to send her to Parla

Sam Altman is trying to buy his own rocket company to compete with SpaceX. The key: data centers

by

The rivalry between Sam Altman and Elon Musk has just reached its highest point: space. And all so that OpenAI can deploy its own data centers in space. The news. As revealed by the Wall Street Journalthe CEO of OpenAI has been exploring the purchase of Stoke Space, a Seattle startup that develops reusable rockets, with the goal of building data centers in space. Although talks with Stoke Space cooled in the fall, the move confirms a trend we’ve been observing for months: Silicon Valley is outgrowing the Earth to fuel AI. Sam’s plan. According to the Journal’s sources, Sam Altman was not looking for a launch provider, but rather an investment that would ensure OpenAI majority control of Stoke Space. Stoke Space, founded in 2020 by former Blue Origin engineers, is developing a fully reusable rocket called ‘Nova’ to compete with SpaceX’s Falcon 9. So that. Altman maintains a tense rivalry with Elon Musk, so the logic of this move would be to reduce OpenAI’s dependence on Musk’s rockets in the event that it decided to deploy servers in space. But above that there is a purely energetic motivation. The computing demand for AI is so insatiable that the environmental consequences of keeping it on Earth will be unsustainable. In certain orbits, however, solar energy is available 24/7 and the vacuum of space offers an infinite heat sink to cool equipment without wasting water. The fever of space data centers. Altman is not alone in this race. What until recently seemed like an eccentricity has become a serious project for big technology companies: And what does Musk say? The irony of Altman pursuing his own rocket company is that the industry’s undisputed leader, Elon Musk’s SpaceX, already has the infrastructure in place. While his competitors design prototypes and seek financing, Musk has cut off the debate with his usual forcefulness: in the face of the discussion about the need to build new orbital data centers, He assured that there is no need to reinvent the wheel: “It will be enough to scale the Starlink V3 satellites… SpaceX is going to do it.” Images | Brazilian Ministry of Communications | Village Global In Xataka | Building data centers in space was the new hot business. Elon Musk just broke it with a tweet

To the question of what sense it makes to compete with Google, OpenAI or Anthropic in AI, Mistral has an answer: small and local models

by

French startup Mistral AI Mistral 3 has been launcheda family of 10 open source artificial intelligence models that represent its most ambitious commitment to date. The Parisian company, which is often considered the main European hope in the development of AI, seeks to differentiate itself from the large American technology companies by betting on flexibility and deployment in all types of devices instead of raw power. Under these lines we tell you all the news. What Mistral has presented. The Mistral 3 family includes a flagship model called Mistral Large 3, with 675 billion parameters, and nine compact models grouped under the name Ministral 3 (in three sizes: 14,000, 8,000 and 3 billion parameters). All models are released under Apache 2.0 license, allowing unrestricted commercial use. The large model also has multimodal capacity, being able to process text and images. It is also multilingual, with a special emphasis on European languages. On the other hand, small models can run on devices with just 4 GB of memory, making them perfect for modest laptops, mobile phones and embedded systems without the need for an internet connection. Why strategy matters. While OpenAI, Google and Anthropic focus on increasingly powerful and closed systems with agentic capabilitiesMistral has focused on the breadth and scope of its models, efficiency and what its co-founder Guillaume Lample calls “distributed intelligence.” According to declared told VentureBeat, the company believes the future of AI is defined not by scale, but by ubiquity: models small enough to run in drones, vehicles, robots and consumer devices. The economic and practical argument. Lample explained It means that in more than 90% of cases, a small, specifically tuned model can get the job done, especially if it is trained with synthetic data for specific tasks. According to Lample, this is not only cheaper and faster, but it eliminates concerns about privacy, latency and reliability. The company also has teams that work directly with customers to analyze specific problems and fine-tune small models that perform specific tasks. This, above all, can attract companies that become frustrated when choosing the best possible model for a specific task and, if it does not perform adequately, they end up giving up. Europe is lagging behind. If we talk about innovation and technology around AI, we do not hesitate to say that Europe is leagues away of what companies in the United States and China are offering. This is why Mistral AI advocates a different approach in which it prioritizes massive deployment in devices and the flexibility of its smaller models. The capacity offered by open models can be a great asset to continue betting on these technologies. In China, for example, the open models of DeepSeek, Alibaba or Kimi are emerging widelyabove in certain tasks even competitors as large as ChatGPT. Lample explained that most leading Chinese models are exclusively text-based, with separate image processing systems. For this reason, they also want to opt for a multimodal approach. A complete ecosystem. Mistral no longer only offers language models. The company has built an entire ecosystem that includes Mistral Agents APIwith connectors for code execution, web search and image generation; Masterlyyour reasoning model; Mistral Code for programming assistance; and AI Studioan application deployment platform that also has analytical and logging capabilities. Furthermore, his assistant Le Chat It has incorporated a deep research mode, voice capabilities and a list of more than 20 enterprise integrations. Thus, in addition to its model offering, the company can provide other companies with a whole layer of personalized products and services, with the aim of being their main source of financing. Digital sovereignty. Although Mistral is often characterized as Europe’s answer to OpenAI, the company prefers to consider itself as ‘a transatlantic collaboration’. Its CEO, in fact, is in the United States, has teams on both continents and trains these models in collaboration with American teams and infrastructure. However, its positioning as a defender of European digital sovereignty has earned it strategic partnerships with the French army, the country’s employment agency, the Luxembourg government and various European public organizations. The European Commission presented in October a strategy to promote European AI tools that provide security and resilience while boosting the continent’s industrial competitiveness. Offline capabilities for democratization. The use cases that Mistral has designed for its small models include, above all, local applications, such as factory robots that use sensor data in real time and without relying on the cloud, drones in natural disasters or rescues that operate offline, and smart cars with functional AI assistants in remote areas. Lample stood out that there are billions of people without internet access but with laptops or cell phones capable of running these small models, which he considers potentially revolutionary. Additionally, by running on the device, these apps preserve the privacy of user data. Real “open source” debate. Not everyone celebrates Mistral’s approach. Some critics question his decision to opt for models’open weight‘, that is, free to access but providing less information about their code than truly “open source” models, which provide the code and training data necessary to train a model from scratch. Andreas Liesenfeld, assistant professor at Radboud University and co-founder of the European Open Source AI Index, declared to the Financial Times that data at scale is the missing key in the European AI innovation ecosystem and that Mistral does not contribute to that at all. The long-term strategic bet. Lample recognize that their models are “a little behind” the most advanced closed systems, but argued that the important thing is that “they are catching up quickly.” Time will tell if Mistral’s approach to low-cost, versatile models with local applications ends up working for them to end up positioning themselves as one of the great European bets on AI. Cover image | Mistral AI In Xataka | China already has an army of 5.8 million engineers. His new plan involves accelerating doctorates

A Chinese startup claims to have created its own TPU to compete with NVIDIA. The only problem is that it is three years late

by

A Chinese startup called Zhonghao Xinying (known internationally as CL Tech) has come to the fore with a bold promise. The company claims to have developed an AI chip that not only circumvents Western intellectual property restrictions, but also outperforms NVIDIA’s A100 chip. Which is very good, but also a little bad. Chana arrives. The chip in question has been named “Chana”, and according to SCMP we are dealing with a GPTPU (General Purpose Tensor Processing Unit). Unlike NVIDIA GPUs, aimed at accelerating AI workloads, this is an ASIC, that is, an application-specific integrated circuit designed from the ground up for neural network workloads. promise. According to Zhonghao Xinying Chana, it offers up to 1.5 times the performance of the NVIDIA A100 based on the Ampere architecture. Not only that: it achieves that performance with 30% lower consumption. The startup highlights that the computational cost per unit would therefore be less than half of that offered by the A100 chips. A little history of the company. Behind Zhonghao Xinying is Yanggong Yifan, an engineer formed at Stanford and the University of Michigan. He worked on the development of several generations of Google TPUs and also on the development of Oracle chips, and in 2018 founded this startup in Hangzhou together with Hanxun Zhengan engineer who worked at Samsung for several years. They were joined by other engineers from Microsoft, Oracle, NVIDIA, Amazon and Facebook, they indicate. on Baidu. We are therefore faced with several of those cases of “boomerang talent” with Chinese engineers who are forged in the US and then return to China to create solutions for their own industry. Solutions that do not depend on the West. Yanggong affirms that its chip features “fully self-controlled IP cores, a custom instruction set, and a fully in-house computing platform. Our chips do not rely on foreign technology licenses, ensuring long-term security and sustainability from an architectural perspective.” But. Although the achievement is striking, it is necessary to put it in perspective. The NVIDIA A100 is a 2020 AI GPU, and even with the improvements that this Chinese startup promises, its performance is, for example, far from H100 chips with Hopper architecture that appeared in 2022. Not to mention of the latest Blackwell Ultra chipswhich are currently NVIDIA’s greatest exponent in terms of AI chips. There are also no details about who makes the chip, and one of the candidates it would be SMICwhich has 7nm technology. They are very far away, and they have another problem. The technical achievement of these engineers is certainly notable, but everything indicates that they are still far from what NVIDIA and its competitors are achieving. like AMD or Google with its recent TPU Ironwood. There is another element that works against them: Chinese manufacturers continue without having direct access to the most advanced photolithography on the market, and although it also there is progress from Chinese manufacturers in that sense, competing is certainly complicated without access to the most advanced technologies. Pressure. In 2024 the company achievement revenues of 598 million yuan (73 million euros) with a net profit of 85.9 million yuan, but in the first half of the year the income was only 102 million yuan and had losses of 144 million yuan. The firm has reached an agreement with its investors by which it will have to go public at the end of 2026, or else it will be forced to buy back shares. The financial pressure is therefore notable for the company, which must demonstrate in the coming months that its roadmap is truly competitive. In Xataka | China was no longer supposed to be able to get its hands on NVIDIA’s most advanced chips. Until he found a shortcut in Indonesia

Saudi Arabia is looking for someone to build its new high-speed train. And a battalion of Spanish giants are going to compete

by

Saudi Arabia has put one of the most ambitious railway projects in the Middle East on the table, and the response from the global industry has been especially strong: 145 international companies have officially expressed their interest for participating in the new high-speed line that will connect Riyadh with Qiddiya, a newly created city dedicated to tourism and entertainment. And as it could not be otherwise, among the candidates stand out several Spanish companies with great experience when it comes to cooperating in Saudi projects. What exactly is this project. It is about the Qiddiya High-Speed ​​Railalso known as Q-Express, a high-speed rail line that will link King Salman International Airport and the King Abdullah Financial District (KAFD) in Riyadh with Qiddiya City, according to the Royal Commission for Riyadh City (RCRC). The trains will reach speeds of up to 250 km/h and the intention is for them to complete the journey in about 30 minutes. Qiddiya is one of Saudi Arabia’s five official mass tourism-oriented gigaprojects and is expected to occupy some 376 square kilometers. The city will include 12 amusement parksa Formula 1 circuit and is projected to house 500,000 inhabitants. Several Spanish companies interested. Between companies that have shown interest There are Spanish names with weight in the railway sector. CAF and Talgo appear in the category of manufacturers of rolling stock and railway systems, where they compete with giants such as Alstom, Siemens Mobility, Hitachi Rail or Stadler Rail. Renfe and Alsa, for their part, are among the 12 interested railway operators, along with Deutsche Bahn, Ferrovie dello Stato Italiane or SNCF. In construction, FCC Construction and Copasa stand out, while in technical consulting, Sener, Ayesa, Idom and Typsa are present, competing with international firms such as Aecom, AtkinsRéalis or Systra. Previous experience in the country. Several of these Spanish companies are not new to Saudi Arabia. Some were part of the consortium that developed the well-known AVE to Mecca (Haramain train), which connects the holy cities of Mecca and Medina. Currently, Renfe operates precisely that high-speed line. The president of the company, Álvaro Fernández de Heredia, visited Saudi Arabia just a few weeks ago to participate in an international railway meeting, and where reaffirmed the company’s commitment to collaborate with Saudi Arabia Railways on new projects. For its part, Alsa It already has a guaranteed presence in Qiddiya: a €500 million contract was recently awarded to operate the city’s future buses. Fierce world-class competition. He complete list of interested parties gives clues to the magnitude of the project. The 68 main contractors include companies from China (eight companies, including China Railway Construction Corporation and Aviation Industry Corporation of China), Turkey (with Gülermak, Kalyon or Yapı Merkezi), Italy (Webuild and Saipem), South Korea (Hyundai Engineering and Samsung C&T), France (Bouygues Travaux Publics), India (Larsen & Toubro) and Portugal (Mota-Engil), among other countries. 16 capital investors and 23 design and project management consultancies have also shown interest. How it is going to develop. The project will be executed under a public-private partnership model (PPP), as announced by the RCRC in collaboration with the National Center for Privatization and the QIC. The registration period where companies could show interest in the project opened on September 12 and closed on October 12. Although it was initially planned to be developed under a conventional model, the Saudi authorities finally opted for a public-private collaboration scheme. What comes next. The development includes two phases. The first will connect Qiddiya with KAFD and King Khalid International Airport. The second phase will extend from a development known as North Pole, which includes the Public Investment Fund’s two-kilometre-high tower, to New Murabba, King Salman Park, central Riyadh and the Industrial City south of Riyadh. In addition, the 65-kilometer Riyadh metro line 7 will also connect the capital with Qiddiya City in the future. With so many high-level companies competing for this megaproject, now it’s time to find out which consortiums manage to position themselves as favorites in the bidding. Cover image | HE In Xataka | The electrification of the railway passes through Valencia: the Stadler plant will be in charge of building 200 hybrid locomotives

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.