Category: Computers

  • Noctua Unveils First AIO with Asetek Pump at Computex

    Noctua Unveils First AIO with Asetek Pump at Computex

    Key Takeaways

    1. Noctua introduced a new unnamed all-in-one (AIO) cooler at Computex 2025, showcasing silent cooling and unique design.
    2. The AIO cooler will be available in three sizes: 420mm, 360mm, and 240mm, using classic beige and brown fans.
    3. It features the Asetek Gen 8 V2 pump, known for its strong performance and includes noise-reducing elements like mass dampers and acoustic foam.
    4. An 80mm custom-designed fan accessory will be available to help cool PSU, RAM, or M.2 SSD.
    5. Pricing and availability for the AIO cooler are not yet announced, with a launch expected in 2026.


    Noctua is well-known for its air-cooling products that deliver excellent performance while ensuring quiet operation. Their signature beige and brown color scheme makes them easily recognizable. At Computex 2025, the company introduced an unnamed all-in-one (AIO) cooler, showcasing its skills in silent cooling and unique design. This cooler features an Asetek pump and comes in three radiator sizes.

    A Variety of Sizes

    Like other big brands unveiling new technologies at Computex, Noctua’s initial AIO cooler will be available in three dimensions: 420mm, 360mm, and 240mm. It incorporates the NF-A14x25 G2 (140mm) fans or the newer NF-A12x25 G2 (120mm) fans, both of which, unsurprisingly, sport the classic beige and brown colors. The cooler is also rumored to utilize Noctua’s SecuFirm2+ mounting system, allowing users to switch from a Noctua air cooler to the AIO without needing to change the mounting bracket.

    Advanced Pump Technology

    For this AIO cooler, Noctua has opted for the Asetek Gen 8 V2 pump. Although it’s the latest model from Asetek, the original version was released back in 2022, with the V2 hitting the market in November 2024. Nonetheless, it has demonstrated great performance, making it a popular choice among various brands, such as Asus (ROG Ryuzo III on Amazon) and Lian Li (GA II LCD on Amazon). This setup is expected to include mass dampers and acoustic foam to help reduce any noise produced by the pump.

    Additional Accessories Available

    Alongside the cooler, Noctua plans to offer an 80mm fan featuring a custom design as an accessory, which can be used to cool the PSU, RAM, or M.2 SSD.

    While the company has not disclosed any information regarding the pricing or availability for its first AIO cooler, it is anticipated to launch in 2026. It’s important to note that this AIO cooler is distinct from Noctua’s “Thermosiphon” cooling solution, which was presented last year. However, a working prototype was showcased at Computex, though details on when this sample will transition into a market-ready product remain unclear.

  • Nvidia Eyes Investment in Photonic Quantum Startup PsiQuantum

    Nvidia Eyes Investment in Photonic Quantum Startup PsiQuantum

    Key Takeaways

    1. Nvidia is in talks to invest in PsiQuantum, a start-up focusing on photonic quantum computers, but details of the deal are unclear.
    2. PsiQuantum seeks at least $750 million in funding, with a pre-money valuation of around $6 billion, and BlackRock is already involved.
    3. Nvidia’s strategy is evolving, as indicated by the establishment of a quantum-research lab in collaboration with Harvard and MIT.
    4. PsiQuantum uses a unique method of creating photonic qubits with standard semiconductor processes, facilitating easier mass production.
    5. Even if the investment happens, PsiQuantum’s technology is still experimental, and commercial quantum computers are not expected to be available immediately.


    Nvidia is said to be deep in talks about investing in PsiQuantum, a start-up in Silicon Valley that’s working on big photonic quantum computers. The Information reports that discussions are going well, but neither Nvidia nor PsiQuantum has made any comments, leaving the specifics like size, structure, and timing of the potential deal unclear.

    Investment Insights

    This potential investment comes after a report from March, which stated that PsiQuantum is looking for at least $750 million, with a pre-money valuation around $6 billion, and that BlackRock is already involved. If Nvidia joins in, it would not only enhance the funds for the start-up but also connect it to Nvidia’s growing ecosystem in quantum software.

    Nvidia’s Evolving Strategy

    Nvidia’s attitude seems to be changing. Back in late 2024, CEO Jensen Huang launched a quantum-research lab in Boston with Harvard and MIT, even after saying that practical quantum computers were likely two decades away. These simultaneous initiatives imply that Nvidia might be preparing for a different timeline.

    PsiQuantum takes a unique path compared to its competitors, who usually depend on superconducting or trapped-ion technology. The company creates photonic qubits using standard semiconductor processes, which could make mass production easier. They are already working with both federal and state agencies to set up two demonstration systems—one in Brisbane, Australia, and another in Chicago, Illinois.

    Future Prospects

    Even if the deal goes through, it won’t lead to a commercially viable quantum computer right away; PsiQuantum doesn’t have any products available, and the technology is still in the experimental phase. However, Nvidia’s interest highlights the strategic importance that major chip manufacturers now attribute to quantum computing as a potential next step or addition to traditional AI accelerators. All information is still tentative until an official announcement is made by the companies.

    Source:
    Link

  • Asus Launches 3000W PSU for Four RTX 5090 Graphics Cards

    Asus Launches 3000W PSU for Four RTX 5090 Graphics Cards

    Key Takeaways

    1. Asus introduced the Pro WS 3000W Platinum PSU, capable of powering four RTX 5090 GPUs simultaneously.
    2. The PSU is 80+ Platinum rated and adheres to ATX 3.1 standards with advanced cooling features.
    3. It includes protective measures against moisture and dirt, and has enhanced heatsinks and a long-lasting fan.
    4. The PSU supports high loads, allowing for a total of 2300 to 2400W for four GPUs, with additional capacity for other components.
    5. Pricing and availability for the Pro WS 3000W, along with 2200W and 1600W models, have not been announced.


    Computex showcases many exciting hardware innovations that often push technology boundaries, and Asus had something remarkable with a 3000W power supply. This ATX PSU can support not just two or three but four RTX 5090 GPUs simultaneously. The company made sure to focus on cooling features to maintain optimal performance over extended periods. This PSU is part of a lineup that also offers 2200W and 1600W models. Currently, there is no information on pricing or when it will be available.

    Details on the Pro WS 3000W

    The Asus Pro WS 3000W Platinum PSU, as indicated by its name, is an 80+ Platinum rated power supply with a capacity of 3000W. It adheres to the latest standards for power efficiency and connectivity. It is compatible with ATX 3.1 and features four native 12V-2×6 connectors with gold-plated pins, which can simultaneously power four GPUs, including the high-end RTX 5090s. However, having four GPUs in a gaming setup is not very common, so this PSU is primarily designed for intensive AI tasks.

    Advanced Features and Design

    The PCB is safeguarded with a protective layer to avoid short circuits caused by moisture or dirt. It also has heatsinks that feature “twice the volume of standard designs” and a dual-ball bearing fan that boasts a lifespan of 80,000 hours. Additionally, the Pro WS 3000W Platinum PSU is equipped with etched cables that are claimed to work at temperatures 50 degrees Celsius lower than the safety threshold. Asus backs this PSU with a solid 10-year warranty.

    In regard to its ability to support four RTX 5090s, each GPU has a maximum load rating of 575W, meaning four of them could draw between 2300 and 2400W when fully loaded. This would leave around 600 to 700W for other components, which is likely sufficient.

    Availability and Other Models

    As previously mentioned, Asus has not disclosed the pricing or availability details for the Pro WS 3000W Platinum PSU. This series also includes 2200W and 1600W power supply units, both of which are also rated 80+ Platinum and comply with ATX 3.1 standards.

    Source:
    Link

  • Lenovo Yoga Slim 7 14 Review: Great OLED, Poor Chassis Quality

    Lenovo Yoga Slim 7 14 Review: Great OLED, Poor Chassis Quality

    Key Takeaways

    1. Lenovo’s latest models feature a 2.8K OLED display with peak brightness over 1,000 nits, enhancing image quality significantly.
    2. The device includes accurate color profiles for sRGB and P3, contributing to excellent color reproduction.
    3. The Yoga Slim 7 is powered by an AMD Ryzen AI 7 350 processor, with 32 GB RAM and 1 TB SSD for robust performance and storage.
    4. Despite a good keyboard design, the chassis quality is lacking, with noticeable creaking sounds that detract from the overall experience.
    5. Additional modern features include USB 4.0, Wi-Fi 7, and Copilot+, meeting current technological standards.


    Lenovo has made significant strides in the display quality of its latest models. The new 2.8K OLED panel features 120 nits and boasts peak brightness exceeding 1,000 nits, making it one of the finest displays available today. While the SDR brightness is slightly below 500 nits, which may not seem impressive, it’s still superior to many current OLEDs from Samsung. The device comes preloaded with accurate color profiles for sRGB and P3, delivering excellent image quality. The high PWM frequency, coupled with low amplitude, contributes positively, although the strong reflections can be a bit distracting. Naturally, Lenovo has integrated these new screens across various series, and we’ve had the opportunity to test several models, including the latest Yoga Slim 7 14 G10.

    Technical Specifications

    The Yoga Slim 7 also presents some remarkable technical specifications. Its AMD Ryzen AI 7 350 (Krackan Point) processor performs well, while the 32 GB RAM and 1 TB SSD offer ample space for storage needs. Additionally, the other features such as USB 4.0, Wi-Fi 7, and Copilot+ are quite satisfactory and meet modern requirements.

    Chassis Concerns

    Nevertheless, there are certain concerns regarding the chassis. The slimmer design in comparison to the Yoga Pro 7 14 does not compromise the quality of the keyboard, which has a nice travel of 1.5 mm. However, the overall build quality is not particularly impressive. The screen lid exhibits noticeable creaking with each movement, and the base unit also produces faint creaking sounds. In daily usage, you’ll notice this whenever you open the display or adjust its angle. It gives off a very cheap impression, which is unexpected from a well-known brand like Lenovo.

    For more details on the new Yoga Slim 7 14 G10, including comprehensive benchmarks and measured values, check out notebookcheck in-depth review.

    Source:
    Link


  • SK Hynix Engineer Indicted for Stealing HBM Tech for Huawei Firm

    SK Hynix Engineer Indicted for Stealing HBM Tech for Huawei Firm

    Key Takeaways

    1. A former SK Hynix engineer, Kim, was arrested for allegedly stealing sensitive semiconductor information to share with HiSilicon, linked to Huawei.
    2. Kim collected around 11,000 pages of proprietary information, including plans for CMOS sensors and HBM production methods, after receiving a job offer from HiSilicon.
    3. The stolen information involved hybrid bonding technology, crucial for enhancing bandwidth and energy efficiency in future high-bandwidth memory.
    4. Kim is facing serious legal consequences under South Korea’s industrial technology protection laws, with potential prison sentences of up to ten years and fines.
    5. This case highlights a worrying trend of intellectual property theft aimed at Chinese companies, prompting South Korean authorities to strengthen protections for domestic innovations.


    A former engineer from SK Hynix has been arrested and officially charged for allegedly taking advanced semiconductor information, particularly regarding next-generation high-bandwidth memory (HBM) packaging methods, and trying to share it with HiSilicon, a company linked to Huawei. The prosecutors label this technology as a “future growth engine” for South Korea’s chip industry.

    Employee’s Actions

    The Seoul Central District Prosecutors’ Office reported that the individual, referred to as Kim, was employed at SK Hynix’s branch in China. He reportedly started collecting sensitive documents after he received a job offer from HiSilicon. Investigators claim Kim printed internal documents, erased their security labels, and took photos of around 11,000 pages of proprietary information, which included confidential plans for complementary metal-oxide-semiconductor (CMOS) image sensors and steps for HBM production.

    Details of the Stolen Information

    The stolen documents allegedly included information on hybrid bonding, a stacking method being examined by both SK Hynix and Samsung for future HBM versions. Hybrid bonding can reduce die size while significantly enhancing bandwidth and energy efficiency—features considered essential for artificial-intelligence accelerators and other demanding tasks.

    Legal Consequences

    Kim’s scheme began falling apart earlier this month. Investigators in industrial technology monitored his actions for several months and apprehended him at Incheon International Airport just before he was about to leave. He is now facing charges under South Korea’s Act on Prevention of Divulgence and Protection of Industrial Technology, which can impose prison sentences up to ten years and fines approximating $71,000; even stricter penalties apply when “strategic sectors” like semiconductors are involved.

    Authorities indicate that this case is part of a broader trend of attempted intellectual-property transfers to Chinese companies, referencing previous incidents with Samsung. Prosecutors have committed to a “stern response” to protect domestic innovations and the nation’s economic interests as the demand for advanced memory technologies grows.

    Source:
    Link

  • AMD Ryzen AI 400 Gorgon Point Laptop CPUs Leaked Online

    AMD Ryzen AI 400 Gorgon Point Laptop CPUs Leaked Online

    Key Takeaways

    1. Zen 6, named Medusa Point, will debut on laptops with an increased core count, featuring Zen 6, Zen 6c, and LP cores.
    2. Four new CPUs (Ryzen 9 475, Ryzen 9 465, Ryzen 7 450, Ryzen 5 445) have been identified as 28 Watt components for laptop use.
    3. These CPUs are likely part of the Gorgon Point series, using the FP8 socket, and are essentially refreshed Strix Point dies.
    4. The Gorgon Point lineup includes three tiers: Gorgon Point 1, Gorgon Point 2 (trimmed version), and Gorgon Point 3 (refresh of Krackan Point).
    5. Medusa Point will be released under the Ryzen AI 500 name, with a GPU limited to RDNA 3.5, as RDNA 4 is not expected for laptops soon.


    Multiple leaks have revealed that Zen 6 is set to debut on laptops under the name Medusa Point. The most recent leak mentioned an increase in core count and highlighted the addition of three distinct CPU cores (Zen 6, Zen 6c, and LP) on the chip. Additionally, another leak has provided insights into AMD’s upcoming laptop SKUs that won’t feature Zen 6 cores.

    New CPUs Spotted

    Renowned leaker Olrak29_ has identified four interesting CPUs: the Ryzen 9 475, Ryzen 9 465, Ryzen 7 450, and Ryzen 5 445 (these names are provisional) on a shipping list. While the specific core counts remain unclear, all of them are designed as 28 Watt components, confirming that they are intended for laptop usage.

    Gorgon Point Details

    Another source, @9550pro, added that these CPUs are probably part of the Gorgon Point series since they utilize the FP8 socket, which is the same as that used by Strix Point. This leaker also mentioned that these processors are basically refreshed Strix Point dies, suggesting they will launch with a combination of Zen 5 and Zen 5c cores.

    Lineup Breakdown

    Earlier, @Olrak29_ shared information about three tiers within the Gorgon Point lineup: Gorgon Point 1, Gorgon Point 2, and Gorgon Point 3. It is suggested that Gorgon Point 2 will be a trimmed-down version of Gorgon Point 1, whereas Gorgon Point 3 will serve as a refresh of the Krackan Point.

    If this information is accurate, it indicates that Medusa Point will be released under the Ryzen AI 500 name, aligning with AMD’s typical naming conventions. The GPU will be limited to RDNA 3.5, as RDNA 4 is not expected to make its way to laptops in the near future.

    Source:
    Link


  • Huawei MateBook Fold: 18″ Foldable Laptop with OLED Display

    Huawei MateBook Fold: 18″ Foldable Laptop with OLED Display

    Key Takeaways

    1. The Huawei MateBook Fold Ultimate Design is a foldable PC featuring HarmonyOS 5.
    2. It has a lightweight design at 1.16 kg, with a thickness of 7.3mm when unfolded and 14.9mm when folded, and includes a unique water drop hinge.
    3. The device boasts an 18-inch OLED LTPO display that folds to 13 inches, with a 3.3K resolution, 1600 nits brightness, and a 92% screen-to-body ratio.
    4. It includes 32GB of RAM, storage options of 1TB or 2TB SSD, dual-band Wi-Fi, Bluetooth 5.2, and features like an 8MP camera and six speakers.
    5. The MateBook Fold is priced at CNY23,999 (~$3,328) for the 1TB model and CNY26,999 (~$3,745) for the 2TB model, surpassing the price of Lenovo’s ThinkPad X1 Fold 16.


    Years after Lenovo introduced the first ever foldable PC, Huawei has now revealed its own version of a foldable computer. Named the MateBook Fold Ultimate Design, this new device is also one of the first from Huawei to run on HarmonyOS 5.

    Design and Build

    Weighing in at 1.16 kg, the MateBook Fold has a thickness of 7.3mm when unfolded and 14.9mm when it’s folded. According to Huawei, this computer features a special water drop hinge designed specifically for foldable devices. This hinge allows the laptop to maintain positions ranging from 30° to 150°. Additionally, it includes an integrated kickstand for supporting the device in landscape orientation.

    Display and Performance

    The MateBook Fold boasts an 18-inch screen that conveniently folds down to a more portable 13 inches. The display is a tandem OLED LTPO panel, showcasing a 92% screen-to-body ratio, alongside a 3.3K (3296 x 2472) resolution and a maximum brightness of up to 1600 nits. It also supports 1440Hz PWC dimming and features an impressive contrast ratio of 2,000,000:1.

    Inside the MateBook Fold, Huawei has packed 32GB of RAM with storage options of either a 1TB or 2TB SSD. While the specific processor hasn’t been disclosed, it’s very likely to be a Kirin chip, similar to other devices in the Huawei MateBook range. Connectivity options include Bluetooth 5.2, dual-band Wi-Fi, two USB-C ports, and a fingerprint scanner located beneath the power button. It also features an 8MP camera at the front, six speakers, and four microphones, along with a 74.69Wh battery.

    Accessories and Pricing

    While the MateBook Fold comes with a virtual keyboard, Huawei is also providing a physical keyboard in the box. This keyboard includes a touchpad, and has a key travel of 1.5mm, weighing only 290 grams and measuring a slim 5mm in thickness. The keyboard claims a battery life of up to 24 days, and users can opt to pair it with a Bluetooth keyboard and mouse.

    Apart from the keyboard, the MateBook Fold package includes a 140W power adapter, a braided cable, a screen cleaning cloth, a large carrying bag for the laptop, and a smaller bag for accessories.

    Offered in Black, Blue, and White, the MateBook Fold is priced at CNY23,999 (~$3,328/€2,953) for the 1TB model and CNY26,999 (~$3,745/€3,322) for the 2TB variant. This price point surpasses the launch cost of the Lenovo ThinkPad X1 Fold 16, which is available on Amazon for $2899. The MateBook Fold 6 is set to launch on June 6 but can already be pre-ordered.

    Source:
    Link


     

  • GMK Unveils EVO-T1 Mini PC with Intel Core Ultra 9 285H

    GMK Unveils EVO-T1 Mini PC with Intel Core Ultra 9 285H

    Key Takeaways

    1. Powerful Processor: The EVO-T1 will feature the Intel Core Ultra 9 285H, a high-performance CPU with 16 cores and 16 threads.

    2. Gaming and AI Capabilities: The mini PC is designed for gaming and AI tasks, featuring an Intel Arc 140T integrated GPU that performs well against competitors.

    3. Unique Design: The EVO-T1 will have a more compact and rounded design compared to the EVO-X series, setting it apart visually.

    4. Connectivity Features: The front of the device is expected to include a Thunderbolt port, although the exact version (Thunderbolt 5 or otherwise) is not confirmed.

    5. Launch Timeline: The EVO-T1 will be presented at the GMKtec x Intel event on May 21, with an anticipated release in July 2025.


    GMKtec has given a sneak peek at its upcoming mini PC, the EVO-T1, which will join the EVO series. This new addition is distinct from the EVO-X series, as it looks to be powered by Intel technology, specifically the Intel Core Ultra 9 285H processor.

    Powerful Processing Capabilities

    The Intel Core Ultra 9 285H is a high-performance Arrow Lake CPU with 16 cores and 16 threads. GMK indicates that this processor will work alongside 64 GB of memory, which is expected to enable the mini PC to run the DeepSeek R1 32B model locally.

    Gaming and AI Performance

    In addition to managing intricate AI operations, the Core Ultra 9 285H is likely to transform the GMK EVO-T1 into a solid gaming mini PC. To elaborate, this Arrow Lake chip includes the Intel Arc 140T integrated GPU, which competes effectively against the Radeon 890M found in the Ryzen AI 9 HX 370 regarding performance-per-watt. The integrated GPU is also capable of smoothly running most AAA titles at 1080p resolution.

    Design and Features

    While GMK has not released extensive specifications for the EVO-T1, the teaser suggests a design that differs from the EVO-X2 (which is available for pre-order on Amazon priced at $1,499.99 using a coupon). The new mini PC seems to be slightly more compact compared to the recently launched AMD mini PC, featuring a more rounded shape instead of a traditional boxy design.

    The teaser image has revealed that the front of the device will include a Thunderbolt port, although it remains uncertain whether it will be Thunderbolt 5. The anticipated launch is set for July 2025, but the device will be presented at the GMKtec x Intel event on May 21. More information regarding the EVO-T1 is expected to be unveiled in the near future.

    Source:
    Link


  • Intel Arc Pro B50 B60 Launch: Affordable Workstation GPUs 24GB VRAM

    Intel Arc Pro B50 B60 Launch: Affordable Workstation GPUs 24GB VRAM

    Key Takeaways

    1. Intel is launching two new professional GPUs, the Arc Pro B50 and Arc Pro B60, during Computex 2025, both built on the Battlemage architecture.
    2. The Arc Pro B50 is priced at $299, featuring 16 Xe2 cores and 16 GB of graphics memory, focused on efficiency and value for AI workstations.
    3. The Arc Pro B60 utilizes all 20 Xe2 cores, offering up to 24 GB of VRAM and variable power usage between 120 to 200 watts, claiming superior performance in AI tasks.
    4. Intel’s Project Battlematrix enables users to combine up to eight Arc Pro B60 GPUs for a total of 192 GB of VRAM, enhancing processing capabilities for larger AI models.
    5. Both GPUs are expected to be available in Q3 2025, with the Arc Pro B50 at $299 and the Arc Pro B60 at around $500, initially sold as part of complete workstations.


    Following the fairly successful debut of the Intel Arc B580, Intel is showcasing two new professional GPUs built on the same Battlemage architecture during Computex 2025. The first model, the Intel Arc Pro B50, is priced at $299, matching the Arc B580, but it comes with 16 Xe2 cores instead of 20 and has a power draw of 70 watts.

    Focus on Efficiency

    Intel is increasing the graphics memory to 16 GB while cutting down memory bandwidth to just 224 GB/s. The main goal of the Arc Pro B50 is to deliver efficiency and good value for its price, as using several Arc Pro B50s together could create AI workstations with a huge amount of graphics memory at a much lower cost compared to a GeForce RTX 5090. Intel’s tests show that the Arc Pro B50 is more than two times faster than the Arc Pro A50, boasting double the performance per dollar when compared to the Nvidia GeForce RTX A1000.

    Enhanced Performance

    The Arc Pro B60, similar to the Arc B580, makes use of all 20 Xe2 cores of the GPU, and its power usage can be varied from 120 watts to 200 watts. In this version, Intel includes 24 GB of VRAM with a bandwidth of 456 GB/s. Thanks to the increased VRAM, the Arc Pro B60 is claimed to be much faster than the Nvidia RTX 2000 Ada or even the GeForce RTX 5060 Ti 16 GB in AI tasks like DeepSeek R1 or Llama3, as shown by the benchmark results below.

    Project Battlematrix

    Intel’s Project Battlematrix allows for the combination of up to eight Arc Pro B60 graphics cards to form a system with an impressive 192 GB of VRAM, which helps in processing larger AI models compared to competitors like the AMD Ryzen Strix Halo or the Apple M4 Max. Project Battlematrix is built on an Intel Xeon processor with PCIe 5 support, and Intel provides Linux-based software for system management.

    The Intel Arc Pro B50 is anticipated to be available in the third quarter of 2025 for $299, while the Arc Pro B60 is expected to launch around the same time for approximately $500. Initially, both models will be offered as part of complete workstations, meaning consumers may not be able to purchase the GPUs individually until the fourth quarter.

    Source:
    Link


  • TSMC 2nm Wafers Expected to See 10% Price Increase

    TSMC 2nm Wafers Expected to See 10% Price Increase

    Key Takeaways

    1. TSMC’s N2 node is in high demand from major companies like AMD, Nvidia, Apple, and Qualcomm, leading to higher pricing for wafers.
    2. Wafer prices are expected to rise further due to construction costs of fabs, economic uncertainty, and TSMC’s need to recover significant investments.
    3. The cost of TSMC’s 2 nm wafers is projected to increase by 10%, potentially reaching around $33,000 for a 300 mm wafer.
    4. Smaller OEMs might shift to older nodes (N3, N4) or consider Samsung Foundry’s 2 nm options due to rising TSMC prices.
    5. Intel’s innovative 18A and 14A nodes are becoming competitive alternatives, featuring technologies not available from TSMC and Samsung Foundry for several years.


    With the increasing demand from companies like AMD, Nvidia, Apple, Qualcomm, and even Intel, which has its own fabrication plants, TSMC’s N2 node is becoming one of the most wanted in recent times. This high demand allows the Taiwanese semiconductor manufacturer to set higher prices.

    Rising Wafer Prices

    A new report from Ctee suggests that wafer prices might rise even further in the upcoming weeks. Insiders believe this is due to the increasing expenses of constructing fabs in countries like the United States, the overall uncertainty in the global economy, and the need to recover some of the $38-42 billion that TSMC intends to invest this year for capital projects.

    Cost of TSMC’s 2 nm Wafers

    A wafer at TSMC’s 2 nm node is expected to be 10% pricier than previously. Last year, the estimated cost was around $30,000 for a 300 mm wafer, which means the new price could reach about $33,000. The report mentions that Nvidia’s CEO, Jensen Huang, finds TSMC’s advanced nodes to be “very worth it.”

    Nvidia plans to use the N2 node for its upcoming Rubin Next architecture. It will also be utilized in AMD’s Zen 6 processors, Apple’s A21 and M5 chips, MediaTek’s Dimensity SoC, and potentially even in the CPU tile of Intel’s Nova Lake desktop series.

    Smaller OEMs and Alternatives

    This situation might lead many smaller OEMs to consider older options like the N3 and N4 nodes. Alternatively, they may explore Samsung Foundry’s 2 nm offerings (SF2/SF2P), which Qualcomm is reportedly looking into. If TSMC’s wafer prices continue to escalate, Samsung Foundry could have a genuine opportunity for a resurgence, assuming they can stabilize their yields.

    Intel’s 18A and 14A nodes are also becoming strong contenders due to their innovative technology like backside power delivery, which TSMC and Samsung Foundry won’t have access to for a few more years.

    Source:
    Link