Hardware News

Motorola is also preparing a “passport-style” foldable smartphone — the Razr Flex will feature a 165 Hz display and an AI shortcut button Motorola is preparing to launch the wide-format foldable Razr Flex smartphone, codenamed Parker, which will compete with the iPhone Duo, Galaxy Z Fold8, and other models comparable in size to a passport. Alleged specifications and images of the device have appeared online, created based on leaked sketches and the current design. The Razr Flex will feature a 6.92-inch internal foldable OLED display with a resolution of 1728 × 2364 pixels and an ultra-high 165 Hz refresh rate. The main display will support HDR10, HDR10+, and Dolby Vision, while the external OLED display will measure 4.97 inches diagonally and have a resolution of 1736 × 1140 pixels. At the same time, according to Android Headlines, the bezels around both screens will be the same width. Inside, the smartphone will have a Qualcomm Snapdragon 8 Gen 5 chip clocked at up to 3.8 GHz, 12 GB of RAM, and 256 GB of internal storage. The internal storage cannot be expanded with a memory card, although Motorola may also prepare other configurations. Battery life will be provided by a 5000 mAh battery consisting of two cells. Qi2 wireless charging will support up to 25 W, while wired charging will operate via USB-PD, although its maximum power is not yet known. The rear panel will house two 50-megapixel cameras — a main camera capable of 30x digital zoom and an ultrawide-angle camera. The device’s connectivity options will include Wi-Fi 7 and Bluetooth 6.0 with LE Audio, while mobile connectivity will be provided by nano-SIM and eSIM support. The smartphone will run Android 17 and offer water and dust protection according to the IP48 and IP49 standards, a fingerprint scanner integrated into the power button, and a separate physical key for artificial intelligence features. When unfolded, the device will measure 110 × 149 × 5.8 mm; when folded, it will measure 110 × 76 × 12.6 mm. The device will weigh 215 grams. Unofficial mock-ups also show a horizontal camera module, buttons in the upper part of the right edge, a centered hole in the external display, and a hole in the upper-right corner of the internal display. The Motorola Razr Flex is expected to be available in PANTONE Coriander (gray-beige) and PANTONE Dark Botanical Garden (dark green). The model is scheduled to launch in mid-December, although the timing may still change. The price has not yet been disclosed, and the company itself has not confirmed the published specifications or the final design. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Spider robots have learned to weld ships — they can walk on ceilings, reach hard-to-access areas, and may take people’s jobs In South Korea, robots have come closer to valuable practical use — they are taking on the role of welders at shipyards. Diden Robotics has introduced the four-legged DIDEN Spider robot for welding, inspection, and other hazardous operations inside vessels under construction. Unlike wheeled platforms, Spider can move across steel floors, vertical walls, and ceilings, using electromagnets in its limbs, which significantly speeds up the work. According to the developer’s estimates, spider robots can accelerate ship construction by 60–70%. This involves not only welding but also the processing of weld seams — inspection, grinding, and painting. Together with Aidin Robotics, a specialist in surface-processing tools, DIDEN Spider is creating a sensor system to determine the forces exerted by robotic tools during automated surface processing. According to Diden Robotics, the company has already signed contracts to supply robots to two South Korean shipbuilding companies and one European customer, while the first commercial units were delivered to a Korean shipbuilder in December 2025. In South Korea, these companies are Samsung and Hyundai, which operate some of the world’s most advanced shipyards. Using robots for welding and surface processing will make them more competitive, which is necessary in light of the growing role of Chinese shipbuilders. Whether the Chinese will stay away from similar innovations is not even a question — they will not, allowing them to continue putting pressure on Samsung and Hyundai. Ultimately, welding robots will become widespread, challenging one of the highest-paid skilled trades. Material on YouTube Incidentally, working with a welding arc creates certain problems for the robots’ electronics. Short high-voltage pulses can spread through the robot’s metal structure and alter the electrical potential used by electronic circuits as a reference. This leads to false signals and interference with operations. Diden Robotics reports that during the initial tests, this caused communication errors and spontaneous restarts of the control boards. Therefore, the company is developing its own electronics specifically for operation in such conditions. The first commercial batches of welding robots are expected to appear in 2027. The company is also developing bipedal platforms for working in narrow openings, as well as a humanoid welding robot. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Money Goes Round, Risks Rise: Nvidia Tries to Draw Insurers into AI Bubble Financing Deals The artificial intelligence boom requires colossal investments with uncertain prospects for financial returns. It is hardly surprising that it has already spawned numerous circular-financing deals in which everyone passes money around in a circle, while no one fully bears responsibility for the end result. Nvidia is trying to balance the industry’s risks by involving insurance companies in indirectly securing deals to build data centers. This is reported by the Financial Times, which says that Nvidia representatives have been negotiating with insurance companies. The idea is to insure against the risk of bankruptcy of a company that has undertaken to build a data center based on Nvidia’s computing accelerators. It is no secret that under normal conditions, these accelerators lose value fairly quickly, and Nvidia itself contributes significantly to this process by releasing increasingly powerful next-generation accelerators every year. Loans for data center construction, however, are issued based on the fixed value assigned to the computing accelerators purchased for the facility. If the borrower goes bankrupt, these accelerators may have to be sold at a steep discount, leaving the lender at risk of losing a great deal of money. Nvidia wants insurers to cover the amount by which the accelerators are discounted in such cases. At the same time, the company boasts about the strong demand for its accelerators, claiming that they lose little value over time amid the AI boom. Some independent appraisers share this optimism. For example, Barkr AI claims that an eight-chip Nvidia system based on the 2022 H100 still costs $320,000—almost as much as a new system cost four years ago. However, this imbalance is possible only because demand for such accelerators is lagging behind supply. Once the market reaches equilibrium, Barkr AI experts believe that a system of this class will lose a third of its value during its first year of operation, and after six years its residual value will not exceed $30,000. According to Nvidia, which is promoting the idea of involving insurers in the indirect financing of the industry through its financial solutions director, Ingemar Lanevi, the availability of specialized insurance programs will attract more capital to the AI infrastructure construction segment. In addition, for counterparties, working with neocloud providers will become just as acceptable as working with market giants that possess substantial financial reserves. At the very least, risk insurance will level the playing field for both major and emerging market participants. Nvidia is even prepared to join investment consortia financing the development of AI infrastructure. It has already undertaken to attract $500 billion to this sector, and it has also assumed financial guarantees amounting to $105 billion for the construction of a large data center for OpenAI. Investments in the AI industry are already being measured in sums so large that even the biggest companies cannot raise them alone, so schemes involving shared liability will become increasingly widespread. Nvidia is already hiring consultants and brokers to work out these matters. In negotiations with potential partners, the company even uses the following arguments: “AI infrastructure is an investment-grade asset because it is uniquely productive, reliable, and fungible.” Initially, Nvidia wanted to cover lenders’ risks using its own funds, but it is now considering insuring such risks. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

TSMC Expands Its Chip Empire in the US: After Six Fabs in Arizona, Another Six Could Appear in Texas Until now, Taiwanese company TSMC’s efforts to localize chip production in the US have been focused on Arizona, where the total number of chip-manufacturing fabs is expected to grow to 12, while four chip-packaging facilities and a research center are also set to appear nearby. New information suggests that TSMC is also interested in Texas as a second US location. The Economic Daily News reports that the company could establish another six advanced facilities in Texas for processing silicon wafers, which are subsequently used to produce semiconductor chips. While TSMC plans to invest up to $265 billion in total in the development of its Arizona cluster, investments in Texas could be even higher, according to Taiwanese sources. TSMC’s customers, which account for more than 75% of its total revenue, are currently experiencing a boom in the US, but most of the company’s facilities are located on the island of Taiwan. TSMC has been actively trying to move closer to its US customers for several years, but so far Arizona is home to the only contract chip-manufacturing facility that produces 4-nanometer components for local customers. Under Biden, US authorities attempted to encourage TSMC to localize production through subsidies, while Donald Trump used tariffs during his second presidential term to exert pressure on Taiwanese businesses. TSMC is reportedly considering establishing a presence in Dallas, in northern Texas, where some infrastructure for developing semiconductor manufacturing operations is already available. Samsung Electronics has two technology parks in the state, although its older facility in Austin can no longer be considered cutting-edge in terms of the technologies it uses. In Taylor, however, the South Korean giant plans to begin producing 2-nanometer chips for Tesla. TSMC is facing water and labor shortages in Arizona, while Texas can offer not only an ample workforce but also well-developed energy infrastructure and a low tax burden. It is noted that TSMC’s board of directors has not yet decided whether to begin construction of facilities in Texas, meaning that the initiative has not been finally approved. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Tesla postpones presentation of flying Roadster sports car by two weeks due to unsuitable weather in Texas On Thursday this week, Tesla planned to hold the next phase of its presentation of the second-generation Roadster sports car—this time in some physically tangible form—but the weather forecast for that part of Texas forced the company to abandon its plans. Due to heavy thunderstorms and strong wind gusts, the event has been rescheduled for October 15. As Electrek reports, viewers invited to the presentation began receiving notifications about the postponement the day before, and Tesla soon officially acknowledged on the social network X that the need to hold the presentation exclusively outdoors, combined with the unfavorable weather forecast, had forced it to set October 15 as the new date for the event. It is still expected to take place at SpaceX’s testing facility in the Texas town of McGregor. The choice of location is presumably driven by Tesla’s desire to demonstrate the second-generation Roadster’s ability to rise above the ground using a system of rocket nozzles. This is what necessitates both holding the presentation in an open space and meeting specific weather requirements. This coming Thursday, figuratively speaking, conditions will be “unfit for flight.” First and foremost, wind gusts in this part of Texas may reach an impressive 50 km/h, while the chance of rain and thunderstorms stands at 82%. Over the following couple of days, wind speeds will decrease only slightly, while the high likelihood of precipitation will remain. Against the backdrop of the roughly six-year delay in the start of production of the second-generation Roadster, an additional two weeks of waiting are unlikely to be a problem for most of the public. Nevertheless, the seriousness of Tesla’s intentions regarding the completion of this “long-running construction project” is partly demonstrated by the fact that the company has resumed accepting pre-orders for delivery, although its current timeline remains unspecified. It is generally believed that mass production of the car will not begin before 2027 or 2028. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Full-size Nothing Headphone (1) Pro headphones unveiled with improved sound and noise cancellation The first full-size headphones from Chinese-British company Nothing were released last summer — they were priced at $299. Just over a year later, the company announced their “professional” version: priced $100 higher, they offer higher-quality sound, more effective noise cancellation, studio-grade tuning options, and an aluminum body. Nothing Headphone (1) Pro headphones are equipped with ten microphones, compared with six in the previous model. The manufacturer changed the microphones’ positioning and improved the earcup design, which made it possible to enhance the quality of ambient-noise pickup and the effectiveness of the active noise-cancellation system. Compared with the standard model, the level of external-noise reduction in the Nothing Headphone (1) Pro has increased by an average of 5 dB, while the frequency range available to the system has expanded by 40%. The manufacturer paid particular attention to sound quality: the single 40 mm dynamic driver has been replaced by a three-driver architecture. There is one dynamic driver for low frequencies, one for mid frequencies, and a high-frequency xMEMS driver. The bass driver delivers clean sound without distortion at high volume, the midrange driver helps eliminate unwanted resonances, and the xMEMS driver is designed to make instruments sound more detailed and natural, bringing them closer to how they are perceived in reality. The 1,060 mAh battery in the Nothing Headphone (1) Pro provides up to 30 hours of operation with active noise cancellation enabled and 68 hours with it disabled. In this respect, the new headphones fall short of their predecessor, but just five minutes of charging can provide up to two hours of playback with noise cancellation enabled or four hours with it disabled. In addition to the improved hardware, the headphones include new software features. Developed in cooperation with London’s Metropolis Studios, these include a range of audio functions, such as a Flat EQ mode switch (neutral equalizer), a new dynamic spatial-audio mode with five profiles supporting head-position tracking, and the ability to customize the equalizer. The headphones support Bluetooth 6.1, Auracast technology, simultaneous connection to two devices, Google Fast Pair, the LDAC codec, and Hi-Res Audio in both wired and wireless modes. As is traditionally the case with Nothing products, design is one of their strong points. The Headphone (1) Pro are made of aluminum, titanium, and glass. The device’s frame is made of aluminum, which is intended to improve the tactile experience when using the headphones; the earcup holders are made of titanium, providing structural strength while keeping the weight low; and the outer sides of the earcups feature scratch-resistant Panda Glass with a hardness rating of 9H, allowing part of the “internals” to be seen. The headband has become wider and thicker, while the ear cushions are softer, which should provide a more comfortable fit. The recommended retail price of the Nothing Headphone (1) Pro is $399. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

AI Hasn’t Replaced Engineers but Has Made Them Three Times Faster — OpenAI Reveals the Secrets Behind Jalapeno’s Development In June, OpenAI introduced its Jalapeno AI chip, developed with the active participation of Broadcom specialists. In an interview with Tom’s Hardware, OpenAI Vice President of Hardware Richard Ho explained how Jalapeno was developed and how the experience gained could benefit the semiconductor industry as a whole. First, as the OpenAI representative noted, the extensive use of AI during the chip-design stage reduced the time required for the entire process from the traditional 18–24 months to nine months. However, it would be premature to say that artificial intelligence has replaced the work of engineers, since a relatively small team of specialists ultimately became more productive. Certain compromises also had to be made in terms of architecture, as the tight development schedule was the priority. In the future, if OpenAI needs to create a more complex chip, it will be able to do so faster than with traditional methods by relying on AI and an existing design. According to Ho, OpenAI was driven to create its own chip by the desire to improve the efficiency of its computing infrastructure, since computing resources are in short supply in the AI sector and must therefore be managed more rationally. In this context, inference chips are particularly important for AI applications today, as they largely determine the user experience. Jalapeno is therefore specifically designed for these purposes, reducing latency when processing requests. At the same time, Jalapeno should perform well with AI models from various providers, not just OpenAI’s own models. In practice, even models that were not taken into account when the chip was created can be adapted for efficient operation on it within a couple of months. Ho believes, however, that OpenAI’s internal demand for computing resources is so great that it will be a very long time before Jalapeno is used by external customers. OpenAI is willing to share the experience it gained from using AI to develop Jalapeno with other market participants, as it is interested in bringing new chip models from other suppliers to market more quickly—chips that OpenAI also uses in its infrastructure. OpenAI will not abandon components from other developers, such as AMD and Nvidia, and will continue to use them in its computing infrastructure. The main requirement is that the data center’s hardware configuration ultimately deliver efficient performance at a given cost level. OpenAI plans to bring new generations of its own chips to market as they become ready, and there are no predetermined requirements in this area regarding the frequency of such announcements. Each new generation should provide a noticeable performance improvement over the previous one, so the transition between generations will not occur excessively often. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

The founders of Anthropic will retain full control as long as at least two of the seven remain at the company At the end of last week, it became known that following the IPO, the seven founders of the Anthropic startup will control 50.1% of the votes on strategic decisions thanks to special shares they will receive. It is now becoming clear that Anthropic’s founders will retain their influence over the company’s development as long as at least two of them remain at the company. Reuters has learned that, in order to reduce the influence of investors’ market interests on Anthropic’s humanitarian mission, a structure called Founder LLC will be established as part of preparations for the IPO, bringing the seven founders together. Moreover, even after the company goes public, Anthropic’s formal status will not change, as the startup will remain a public benefit corporation. This organizational structure is expected to allow the founders to balance the interests of all humanity with those of the investors who directly provide capital to Anthropic. The seven co-founders will control 50.1% of the votes through a single Class F share, allowing them to influence all decisions concerning Anthropic’s management, with the exception of the appointment of certain board members. In total, the company’s shares will be divided into five classes, with ordinary investors receiving the least “influential” Class A shares. The seven Anthropic co-founders have demonstrated strategic unity throughout the five years since the startup was established, and some of them first met ten years ago. Nevertheless, Anthropic’s governance structure includes mechanisms to address the departure or death of several co-founders. Essentially, the founders’ influence will fade only if fewer than two members of the original group remain at the company. The co-founders will be able to appoint their successors within Founder LLC, meaning that this influence will not be limited by the span of their lifetimes. In addition to CEO Dario Amodei, Founder LLC includes his sister Daniela, who also chairs the board of directors. In addition to the two representatives of this family, another director—whose name is not yet known—will be appointed following a vote by the holders of Class A and Class F shares. The remaining four board members will be appointed by the Long-Term Benefit Trust. It has also become known how much compensation Anthropic’s co-founders received in 2025. Dario Amodei received nearly $18 million, primarily in the form of shares, while his sister Daniela received $16.4 million. Going forward, the startup’s co-founders have committed to directing up to 80% of the assets transferred to their personal control toward charitable causes. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

Anthropic Plans to Spend $518 Billion on Infrastructure Over the Next Few Years As the AI startup Anthropic is gearing up for an IPO in full force, it is now time to share financial information with potential investors. The Reuters news agency obtained an opportunity to review Anthropic’s prospectus, and some of the figures in the document are staggering. In particular, the startup’s net losses for last year amounted to $42 billion, while over the next several years it plans to spend $518 billion on infrastructure. At the same time, Anthropic’s revenue is growing at a rapid pace—it increased twelvefold over the past year to nearly $4.6 billion, although the company’s operating losses during the same period exceeded $8 billion. As previously reported, Anthropic’s valuation following its IPO is expected to exceed $2 trillion. Last year alone, the startup spent $7.33 billion on computing infrastructure and computing, three times its corresponding expenses in 2024. In total, Anthropic’s operating expenses reached $12.65 billion last year. However, the net loss figure for 2025 includes approximately $34 billion that will be converted into company shares and written off after the IPO, meaning that the total net loss of $42 billion is not determined solely by direct expenses. Anthropic notes that approximately one-quarter of the startup’s total revenue last year came from its two largest customers, making this dependence a risk factor. At the end of December last year, the company held a total of $20.28 billion in cash and highly liquid assets. If you value what we do, you can support our projects on Boosty. With love, NSWTL.

The Volkswagen ID.Tiguan Electric Crossover Will Replace the ID.4 Over the past five years, Volkswagen has spent its time trying to find its place in the rapidly changing passenger car market. The ID electric vehicle lineup was kept separate, but this year the key crossover in this segment will be designated the ID.Tiguan, demonstrating the continuity of a popular name under new conditions. On the European market, the ID.Tiguan will replace the ID.4 and its sportier ID.5 version as early as the beginning of next year, although production of the eponymous electric vehicles will continue in China until the new global model enters the local market. In the United States, production of the ID.4 was discontinued back in April of this year. Although the Volkswagen press release shows the ID.Tiguan from various exterior angles and with two bumper design options, the vehicle is still covered in camouflage film, and the interior has not been shown. The ID.Tiguan is scheduled to premiere in early October, so the manufacturer is anticipating the event with explanations regarding the future composition of its model lineup. Since 2021, the ID.4 has sold 950,000 units, so it cannot be called a failure in terms of sales. It is expected that the lack of physical control buttons, which has deterred some buyers, will be addressed in the ID.Tiguan’s interior, following the example of the ID.3 NEO and ID.Polo. Alongside the electric ID.Tiguan, a conventional Tiguan with an internal combustion engine or hybrid powertrain will also be sold. At present, the eponymous crossover remains the most popular model not only in Volkswagen’s lineup but also among the products of the entire parent group. Last year alone, the Tiguan and its related Tayron sold more than 1 million units. From a marketing perspective, using a well-known name for a new electric vehicle certainly makes sense. Although the ID.Tiguan adopts some stylistic features from the ID.4 (pictured above), its body shapes are becoming more angular, as is the lighting design. The more compact ID.Polo and ID.Cross were also recently unveiled, with the former proving quite popular in Europe during a period of rising fuel prices at the pre-order stage, as waiting times stretched to ten months. Clearly, Volkswagen is placing its main bet on the ID.Tiguan in terms of sales volume. Last year, Volkswagen sold approximately 382,000 electric vehicles worldwide. If you value what we do, you can support our projects on Boosty. With love, NSWTL.