Technology

Rapidus to start construction on 1.4nm fab in 2027 — research and development on node to begin next year

Source

Tom's Hardware

Published

Nov 25, 2025

TL;DR

AI Generated

Rapidus, a Japanese chipmaker, plans to construct a 1.4-nanometer fab in 2027, with production set for 2029 in Hokkaido to compete with TSMC. The company will start R&D on the node next year and has received substantial investments from Japanese companies and the government. Despite facing competition from established fabs like TSMC and Intel, Rapidus aims to streamline production with advanced packaging techniques. However, former Intel CEO Pat Gelsinger suggests the need for more advanced offerings to compete successfully.

Read Full Article

Lexar regional manager says that RAM prices are expected to double by the end of the year — 'discounts' and stabilized prices result from distributors getting rid of old stock or sourcing products from other regions

Lexar's Regional Manager for Australia & New Zealand predicts that RAM prices will double by the end of the year due to high demand for memory chips in AI build-outs. While some retailers are offering discounts on RAM modules, this is mainly to clear old stock before new, more expensive supplies arrive. Industry experts now expect RAM prices to keep rising throughout the year, with shortages affecting not just desktops and laptops but also smartphones and other devices. Consumers are advised to purchase RAM now as prices are unlikely to decrease anytime soon.

Tom's Hardware•

1 week ago

Taiwanese startup FormulaV Line wants to break into US market with two new unique cases — company expects products to become available on Newegg later this year

Taiwanese startup FormulaV Line unveiled two new PC cases at Computex 2026: the Air Power G10 with tilting fans for customizable airflow and the Crystal Z3 featuring a unique bottom-intake chamber design. The Air Power G10 can accommodate ATX motherboards and offers various fan configurations, while the Crystal Z3 provides a panoramic view and detachable bottom fans for easy maintenance. These cases are expected to hit the US market later this year on Newegg, with the Air Power G10 priced around $150 and the Z3 around $80. Additionally, FormulaV Line showcased other products like a 1000-watt PSU and cooling solutions at the event.

Tom's Hardware•

1 week ago

G.Skill explains how AMD EXPO ULL unlocks additional performance — expanded profiles allow memory makers to include subtiming tweaks for the first time

G.Skill showcased new memory kits at Computex with AMD's EXPO Ultra Low Latency (ULL) support, aiming to reduce memory latencies for improved CPU performance. The EXPO ULL program allows memory makers to adjust sub-timings within primary timings for lower latencies, enhancing memory performance. While faster memory speeds above 6000 MT/s can add latency on AMD platforms, EXPO ULL kits aim to optimize latency without the need for manual adjustments. These kits require stricter binning during production, potentially making them more expensive but appealing to gamers seeking optimal CPU-bound gaming performance.

Tom's Hardware•

1 week ago

Huawei-led team claims it post-trained DeepSeek's 1.6-trillion-parameter model — 1,000 Ascend 910C chips used in training

A research group led by Huawei Technologies has successfully post-trained DeepSeek's V4-Pro, a 1.6-trillion-parameter model, using a cluster of 1,000 Ascend 910C chips. This achievement showcases Chinese accelerators' capability to handle training-class workloads on domestic silicon, a significant development amid U.S. export controls. While Chinese chips have excelled in inference tasks, they have historically struggled with training, making this accomplishment noteworthy. However, the team's claim lacks specific benchmarks or comparisons with Nvidia hardware, leaving some aspects of the achievement unverified.

Tom's Hardware•

1 week ago

Lexar regional manager says that RAM prices are expected to double by the end of the year — 'discounts' and stabilized prices result from distributors getting rid of old stock or sourcing products from other regions

Tom's Hardware•

1 week ago

Taiwanese startup FormulaV Line wants to break into US market with two new unique cases — company expects products to become available on Newegg later this year

Tom's Hardware•

1 week ago

G.Skill explains how AMD EXPO ULL unlocks additional performance — expanded profiles allow memory makers to include subtiming tweaks for the first time

Tom's Hardware•

1 week ago

Huawei-led team claims it post-trained DeepSeek's 1.6-trillion-parameter model — 1,000 Ascend 910C chips used in training

Tom's Hardware•

1 week ago

Rapidus to start construction on 1.4nm fab in 2027 — research and development on node to begin next year

TL;DR

Similar Articles

Lexar regional manager says that RAM prices are expected to double by the end of the year — 'discounts' and stabilized prices result from distributors getting rid of old stock or sourcing products from other regions

Taiwanese startup FormulaV Line wants to break into US market with two new unique cases — company expects products to become available on Newegg later this year

G.Skill explains how AMD EXPO ULL unlocks additional performance — expanded profiles allow memory makers to include subtiming tweaks for the first time

Huawei-led team claims it post-trained DeepSeek's 1.6-trillion-parameter model — 1,000 Ascend 910C chips used in training

We use cookies

Rapidus to start construction on 1.4nm fab in 2027 — research and development on node to begin next year

TL;DR

Similar Articles

Lexar regional manager says that RAM prices are expected to double by the end of the year — 'discounts' and stabilized prices result from distributors getting rid of old stock or sourcing products from other regions

Taiwanese startup FormulaV Line wants to break into US market with two new unique cases — company expects products to become available on Newegg later this year

G.Skill explains how AMD EXPO ULL unlocks additional performance — expanded profiles allow memory makers to include subtiming tweaks for the first time

Huawei-led team claims it post-trained DeepSeek's 1.6-trillion-parameter model — 1,000 Ascend 910C chips used in training