Micron Reports Record Fourth-Quarter Revenue and Projects Further Growth
United StatesMemoryEarnings & Guidance
On September 30, Micron reported fiscal 2026 fourth-quarter results for the period ended September 3. Revenue reached US$54.23 billion, compared with US$41.46 billion in the preceding quarter and US$11.32 billion a year earlier. Full-year revenue totaled US$133.19 billion.
For fiscal 2027 Q1, Micron guided to revenue of US$61.5 billion, plus or minus US$1.5 billion. The company plans further investment in technology, products and manufacturing. This guidance represents management expectations, rather than completed sales.
Micron also began sampling 512GB high-capacity DDR5 RDIMMs and highlighted customer adoption progress for high-speed server memory and data-center SSDs.
R&A View: AI platforms require greater memory capacity, bandwidth and storage performance. Buyers should assess server memory, enterprise SSDs and mainstream memory separately. Aggregate revenue growth does not establish inventory availability or lead times for individual part numbers.
Samsung Expects HBM to Use Nearly 30% of Industry DRAM Wafer Capacity in 2027
South KoreaHBMDRAM Capacity
Reuters reported on September 29 that a Samsung Electronics executive expects HBM to account for nearly 30% of DRAM manufacturers’ total wafer capacity in 2027, up from approximately 20% today.
Samsung noted that HBM and conventional DRAM compete for wafer capacity, so expansion in HBM could constrain conventional DRAM supply. The estimate concerns capacity allocation, not HBM revenue share or an equivalent percentage reduction in conventional DRAM output.
The forecast highlights the continuing influence of AI memory demand on manufacturing resources.
R&A View: Buyers outside AI markets should also monitor memory capacity allocation. Plan by DDR generation, density, temperature grade and qualification status, and assess alternatives early. An industry forecast does not mean every part number will face the same price increases or shortages.
VSMC Opens Singapore 300mm Fab, Targeting Volume Production in Q1 2027
SingaporeWafer ManufacturingSpecialty Processes
On September 28, VSMC, the joint venture between NXP and Vanguard International Semiconductor, opened its first 300mm fab in Tampines, Singapore. The facility has entered risk production and plans to begin volume production in the first quarter of 2027.
The fab supports processes from 130nm to 40nm for mixed-signal, power-management, analog and interposer applications, serving high-performance computing, mobile, automotive, industrial and consumer markets.
At full capacity in 2029, the facility is expected to produce approximately 44,000 12-inch wafers per month. Opening, risk production and full production are separate milestones; target capacity should not be treated as fully available supply today.
R&A View: This investment adds a medium-term supply option for mature and specialty processes. Monitor product transfers and qualification progress, while continuing to base near-term purchasing on part-specific production status, allocation and confirmed delivery commitments.
Infineon Opens Thailand Backend Facility, Expanding Assembly and Test Capacity
Germany / ThailandAssembly & TestManufacturing
On October 1, Infineon announced the opening of its new backend manufacturing facility in Samut Prakan, in the Bangkok area of Thailand. The site covers semiconductor backend manufacturing and wafer testing for automotive, industrial and energy-infrastructure applications.
The site begins with Module A and has space for expansion to five modules. Infineon said it currently employs around 350 people, with headcount expected to reach approximately 1,000 as the first building ramps up.
The project complements Infineon’s frontend manufacturing expansion and diversifies its global production footprint. It is not a new frontend wafer fab.
R&A View: Assembly and testing are essential to chip delivery. Buyers should track affected product families, qualification requirements and product-change notices. The opening supports long-term supply flexibility, but does not establish immediate lead-time improvements across all Infineon devices.
Renesas Introduces Dual-Side-Cooled 650V GaN Device for AI Data-Center Power
JapanGaNAI Power
On September 30, Renesas introduced the TP65H020G4PLSGBD GaN device in an 8 × 8mm dual-side-cooled PQFN package. It has a nominal on-resistance of 20mΩ and targets high-density power conversion in AI data centers.
Renesas says the package reduces board footprint by 57% compared with a 10 × 15mm TOLT package. The device addresses conversion stages within 800V power architectures; implementation must respect its device ratings and the chosen circuit topology.
Samples are available to major AI data-center OEMs and ODMs, with volume production planned for mid-2027.
R&A View: Server power upgrades require thermal design, packaging and switching performance to be evaluated together. Validate gate drivers, controllers, magnetic components and cooling as a system. Distinguish sample availability from volume supply when scheduling new-product introductions.
Synopsys and Amazon Sign Multiyear IP Agreement Worth More Than US$1 Billion
United StatesSilicon IPCustom Chips
On September 30, Synopsys announced a multiyear agreement with Amazon worth more than US$1 billion, expanding application-optimized silicon IP for custom chips through a licensing and royalty model.
The collaboration also covers EDA, simulation and analysis, and AI-assisted engineering. The companies plan to optimize Synopsys multiphysics software for Trainium and Graviton platforms. Synopsys will also use AWS computing, storage and Amazon Bedrock to support development.
This is a multiyear technology and licensing agreement. Its value should not be interpreted as chip purchases in the current period.
R&A View: Cloud providers continue investing in design IP and tools for custom silicon. Supply-chain teams should track subsequent platform specifications, packaging requirements and production plans. The licensing agreement cannot be directly translated into near-term component orders.
AMD Plans US$8.2 Billion Acquisition of World Labs to Strengthen Physical AI
United StatesPhysical AIProposed Acquisition
On September 28, AMD announced a definitive agreement to acquire World Labs in an all-stock transaction valued at approximately US$8.2 billion. World Labs focuses on spatial intelligence and AI models that understand and simulate the physical world.
According to the announcements, Fei-Fei Li will join AMD as Executive Vice President and Chief Scientist following completion. The companies have already collaborated on optimizing model training and inference on AMD GPUs.
The transaction is expected to close by the end of 2026, subject to regulatory approvals and other customary closing conditions.
R&A View: Chipmakers are adding model and software expertise to shape future computing platforms. Robotics, simulation and edge-AI teams should follow the resulting hardware and software roadmaps. The announcement alone does not imply immediate product changes or additional volume-production demand.
OpenAI and Synopsys Partner to Develop GPT-Synopsys for Chip Design
United StatesEDAAI-Assisted Design
On September 30, Synopsys and OpenAI announced a multiyear strategic collaboration to develop GPT-Synopsys, combining specialized AI models with Synopsys EDA tools and chip-design expertise.
OpenAI will license Synopsys EDA tools for model development. The companies also plan joint research, development and go-to-market activities under a revenue-sharing framework.
The aim is to help models operate EDA tools, interpret results and iteratively improve designs. The announcement describes a joint development program, rather than deployment across all customer workflows.
R&A View: AI-assisted design must prove its value through engineering validation, including repeatability, design-rule checks and power, performance and area targets. Faster design does not automatically shorten manufacturing lead times; design, tape-out and volume-production milestones still require separate planning.
