Week 2026-34

Commutator maker Kaizhong announced its intention to expand into data centers and other industrial sectors, while also targeting new markets in Southeast Asia.

Nvidia formed a political action committee to increase its lobbying presence amid ongoing government debates regarding artificial intelligence regulations.

Samsung anticipates its first-generation zHBM products by 2029, noting that heat management remains a primary engineering challenge for stacking HBM over AI accelerators.

Samsung presented its advanced semiconductor packaging roadmap, focusing on FOPLP using glass substrates and emphasizing AI's role in future research.

2.5D/3D advanced packaging, Glass-core substrates

SK Hynix is considering a stronger partnership with Kioxia Holdings concerning NAND flash, signaling an extension of its US expansion strategy into this competitive memory area.

YMTC is strengthening its domestic semiconductor supply chain in China by expanding partnerships with 17 local firms in Wuhan that provide materials and equipment.

Leading-edge foundry, Fab materials, masks, gases

Vanguard International reported a fire at one of its fabs in Taiwan but stated that the incident did not interrupt production operations.

Leading-edge foundry, Fab materials, masks, gases

Anthropic introduced a hardware standard intended to simplify the integration of AI agents with physical equipment globally for research and manufacturing.

ASIC design services, IP & EDA, General contractors & design-build

Renesas opened a physical AI and robotics lab in Beijing to accelerate the development and testing of robotic systems.

Wah Lee Industrial aims to advance its Physical AI initiatives by targeting an Original Design Manufacturer (ODM) phase in the second quarter of 2027.

OpenAI announced the performance of its custom inference chip, Jalapeño, demonstrating improvements in both throughput and latency.

Hyperscaler custom ASICs, Rack-scale systems

MIPS joined RISC-V International as a Premier Member and appointed its SVP & CTO to the Board to advance open standards in semiconductor technology.

Server CPUs, Leading-edge foundry

The EuroHPC JU extended its funding for the EuroCC and CASTIEL projects to bolster Europe's High Performance Computing infrastructure.

Merchant GPUs / accelerators, Rack-scale systems

The Schwarz Group is committing a €5.6 billion investment to build a data center facility in Mecklenburg-Vorpommern, Germany.

Utilities, IPPs & power providers signing data-center deals, Hyperscalers, neoclouds & model labs

Federal funding is allocating $2.2B to support the deployment of microreactors at military installations, enhancing site resilience and next-generation nuclear power viability.

SMRs, advanced reactors, components, fuel, Large power transformers (345–765 kV) & substations

At Hot Chips 2026, Oxmiq Labs analyzed the potential of HBF compared to HBM in the context of AI compute.

Merchant GPUs / accelerators, HBM

Glass-core substrates, which Intel had promised as replacements for organic chip packaging, are nearing final qualification but have yet to reach commercial availability.

Developers engaging with local governments often use NDAs, which limits public access to information regarding data center projects.

Data center developers, operators & REITs, Hyperscalers, neoclouds & model labs

The Proton data center in Frankfurt experienced a cooling failure and subsequently went offline.

Rack manifolds, Quick disconnects

OpenAI unveiled its Jalapeño AI ASIC accelerator, noting it offers good performance per watt and low latency for inference tasks compared to Nvidia's Blackwell.

Merchant GPUs / accelerators, Hyperscaler custom ASICs

A company is breaking ground on a new data center facility in Graz, Austria.

Hyperscalers, neoclouds & model labs, Prefab / modular data centers & power/cooling skids

Google initiated an AI for Energy Accelerator program listing numerous startups across the US and Europe.

Utility Georgia Power will build the necessary power infrastructure for OpenAI’s planned 3.2GW data center in Effingham County, Georgia.

Chillers & compressors, Floor PDUs, RPPs, static transfer switches, ATS, Circuit protection, fuses, surge, grounding, Hyperscalers, neoclouds & model labs

The data center startup Fossefall will use Armada to deploy five modular Leviathan data centers in the Nordics.

A significant discount made Corsair's 32GB DDR5-6000 memory kit the cheapest option at those specific speeds.

Harbert acquired Mid-Georgia Cogen to operate a cogen plant that provides peak power and waste heat for both Georgia Power and an adjacent Frito-Lay facility.

Heat reuse & district heating interfaces, Backup generation: gensets, fuel systems, HVO, Gas turbines (onsite / behind-the-meter) & HRSGs

According to the NRDC, Trump's energy policies could significantly reduce the potential build-out of renewable energy capacity in the United States.

This article advises on designing a dependable 24V power path by selecting appropriate MOSFETs, diodes, TVS devices, and capacitors to manage surges, reverse polarity, and inrush current.

Board-level power: VRMs, power stages, PoL, power semis, UPS (static, modular, rotary/DRUPS) & UPS power semis

Following persistent pressure from a customer, Micron offered a refund and eventual replacement for Crucial X9 4TB SSDs that were initially reported as unavailable.

Physical AI mandates complex silicon designs that require multi-vendor supply chains and full lifecycle security.

Merchant GPUs / accelerators, Hyperscaler custom ASICs, ASIC design services, IP & EDA

SpaceX plans a massive $100B spaceport in Louisiana with facilities for propellant production and self-sustaining power.

Structure, steel, precast, envelope, site equipment, Prefab / modular data centers & power/cooling skids

Although Huawei's Ascend 910C is sold out in China, the inability of domestic chips to fully replace Nvidia's H200 suggests ongoing global AI compute constraints.

Merchant GPUs / accelerators, Hyperscaler custom ASICs

US export restrictions on advanced AI chips are forcing China's AI companies to shift focus from model performance to the ability to operate models at scale using domestically manufactured chips.

Merchant GPUs / accelerators, Hyperscaler custom ASICs, ASIC design services, IP & EDA

Increased data center demand for high-frequency optics is causing a severe indium phosphide (InP) substrate shortage, leading companies to sign long-term contracts for supply stability.

Silicon photonics & co-packaged optics, Optical transceivers

Due to surging AI chip demand, outsourced assembly and test providers warn that traditional packaging capacity, particularly wire bonding, could face a deficit exceeding 20% by 2027.

2.5D/3D advanced packaging, Power shelves & PSUs

Revenue for IntelliEPI in Q2 2026 was impacted by a temporary shutdown of a Japanese substrate supplier, which limited the availability of InP substrates used in high-speed optical communications.

Critical minerals: gallium, germanium, indium, tungsten, beryllium, Substrate materials

China is intensifying efforts to secure helium supplies due to rising demand from semiconductor manufacturing and commercial space sectors.

Strong AI demand is causing shortages and long lead times for various ICs and passive components, with Data Image securing a one-year supply of industrial control products.

Leading-edge foundry, Semiconductor test & burn-in

High memory component pricing may force cloud operators to allocate a significantly larger portion of their capital expenditure to DRAM and NAND.

Nvidia's forecast indicates that demand for AI infrastructure remains robust and will drive continued price pressure across components including memory, power systems, cooling units, and networking hardware.

Merchant GPUs / accelerators, Memory modules, Coolant distribution units, CRAH / CRAC / fan walls / in-row units, Energy storage: UPS batteries & BESS

Nvidia reported massive revenue growth, driven by the global expansion of AI infrastructure and increased purchases from hyperscalers.

Merchant GPUs / accelerators, Hyperscaler custom ASICs, Rack-scale systems

Marvell Technologies raised its revenue outlook for fiscal 2027 and 2028, driven by custom silicon and scale-up optics.

Optical transceivers, Silicon photonics & co-packaged optics

Marvell's total revenue reached a record $2.74 billion in Q2 FY2027, driven primarily by its data center segment which contributed 79% of the total.

NICs / SuperNICs, DPUs

Nvidia anticipates that sales of Vera Rubin systems will represent a rapid and significant increase, contributing 20% of its data center revenue mix in the third fiscal quarter.

The PJM Interconnection's wholesale power costs increased 46% year-to-date due to rising energy and capacity expenses.

Nvidia reported revenue exceeding $96 billion and committed to purchasing up to $160 billion in memory, bringing its total commitments to $279 billion amid strong AI demand.

Merchant GPUs / accelerators, Memory modules, CXL controllers & memory expanders

Nvidia's revenue is expected to increase significantly, driven by high profitability per unit of data center power.

Merchant GPUs / accelerators, Rack-scale systems

SK hynix broke ground on its new fabrication facility in Indiana, aiming to establish a local base for AI memory production.

Leading-edge foundry, Memory modules

SK hynix held a groundbreaking ceremony for its HBM production facility in Indiana.

SK hynix held an event to begin building a new production facility for HBM in Indiana.

LG CNS implemented direct-to-chip cooling for Naver Cloud's data center located in Goyang City.

ASIP Technologies is partnering with APACT to build an OSAT facility in India's Visakhapatnam region.

A Section 232 duty on advanced logic chips was implemented, but an exemption was issued for imported computer processing units used in US data centers.

Server CPUs, DPUs

The Trump administration is considering expanding semiconductor tariffs to include servers and other finished products containing chips, potentially revoking previous data center exemptions.

Merchant GPUs / accelerators, EMS, system integration & rack-level test

The state's public counsel criticized Duke Energy Florida for its proposed rate structure, stating it fails to comply with provisions of the new data center law.

Grid equipment manufacturers require clearer rules regarding software and digital products when determining the country of origin.

HV switchgear, GIS, breakers, protection relays, MV switchgear, E-houses, power skids

An environmental group is suing the EPA over two newly fast-tracked chemicals approved for semiconductor manufacturing that lack adequate safety safeguards.

Delaware passed a clean energy mandate for data centers, affecting hyperscalers and owners.

Utilities, IPPs & power providers signing data-center deals, Hyperscalers, neoclouds & model labs

Proposed EPA regulation changes could eliminate the need for public comment on air pollution permits issued near data centers.

Industrial & specialty gases, Air handling, economizers, filtration, MEP & electrical contractors

SK hynix began construction of an HBM production facility in Indiana as part of a new US-Korea AI effort.

Lam Research started an expansion in Oregon as part of a $3 billion global investment focused on accelerating AI chip development.

Leading-edge foundry, Fab materials, masks, gases

SK hynix initiated construction on a US$4 billion advanced packaging facility in Indiana dedicated to handling High Bandwidth Memory (HBM).

2.5D/3D advanced packaging, HBM

YMTC aims to become the world's largest NAND flash producer by 2027, surpassing market leaders Samsung and SK hynix.

SK hynix is constructing a new advanced packaging production base in Indiana to support HBM manufacturing.

HBM, 2.5D/3D advanced packaging

SK hynix is investing over $4 billion in an Indiana fab to establish U.S.-based mass production of next-generation HBM starting in the second half of 2029.

Kioxia is building a new facility (Fab3) to increase its production capacity for advanced three-dimensional flash memory.

Gateway Capital plans to construct an 81MW data center facility in Sydney on a former industrial site.

Data center developers, operators & REITs, General contractors & design-build, Structure, steel, precast, envelope, site equipment

Amazon Web Services is planning to deploy two million extra Nvidia GPUs into their infrastructure.

Merchant GPUs / accelerators, Rack-scale systems

Kioxia and Sandisk plan a significant investment of over US$31 billion in Japan to expand flash memory capacity, supporting global AI infrastructure demand.

The Salim Group purchased Keppel's stake in the Indonesian data center venture, which is now relaunched as IndoData.

NewPower Worldwide expanded its credit facility to $750 million to support global growth and customer demand across the electronics supply chain.

NICs / SuperNICs, Board-level connectors & sockets

SK Horizon was launched with $2.2 billion in investment from KKR and IMM, focusing on AI data center infrastructure.

Hyperscalers, neoclouds & model labs, Prefab / modular data centers & power/cooling skids

OpenAI unveiled Jalapeño, a custom, purpose-built ASIC designed from the ground up specifically for AI workloads.

Amazon's Annapurna Labs is set to collaborate with NVIDIA on the development of NVHBM technology alongside NVLink Fusion for next-generation AI workloads.

Merchant GPUs / accelerators, HBM

Amazon Web Services and NVIDIA announced a major expansion of their collaboration to deploy two million additional GPUs and advanced infrastructure for AI workloads.

Merchant GPUs / accelerators, Rack-scale systems

Cerebras unveiled its roadmap for wafer-scale accelerators and presented the Nexus rack design for accelerating performance using stacked DRAM.

Merchant GPUs / accelerators, Memory modules

VCI Global has launched a prefabricated modular platform designed for AI data centers capable of supporting up to 500MW of compute capacity.

Nvidia introduced NVHBM, a custom high-bandwidth memory product offering superior performance and efficiency compared to standard HBM4e for its NVLink Fusion partners.

The NVMe 2.4 update enhances security features, power management capabilities, and virtualization support for various cloud and AI workloads.

NICs / SuperNICs, DPUs, FPGAs

Meta introduced the MTIA 400 chip, an AI accelerator that offers speed improvements over Blackwell but has not yet replaced competitive offerings from AMD or Nvidia.

Merchant GPUs / accelerators, Hyperscaler custom ASICs

Nvidia's VP of hardware presented the architecture of the Groq 3 LPX rack and released a third-party benchmark for its performance.

Merchant GPUs / accelerators, Rack-scale systems

Nvidia presented the Rubin GPU and highlighted that power remains a major constraint on data center capacity, demonstrating compute efficiency within a fixed 100MW power budget.

Rack-scale systems, UPS (static, modular, rotary/DRUPS) & UPS power semis

Fujitsu unveiled its 144-core Monaka server CPU at Hot Chips 2026, detailing a shift to dual 256-bit SVE2 vector units for the new generation.

OXMIQ discussed High Bandwidth Flash (HBF) at Hot Chips 2026, clarifying that its practical use is restricted and highlighting a new memory format that balances HBM performance with NAND flash capacity.

d-Matrix showcased Raptor, a 3D DRAM accelerator that bonds a TSMC 4nm compute die directly onto a custom DRAM die.

Merchant GPUs / accelerators, HBM, 2.5D/3D advanced packaging, MEP & electrical contractors

Arm revealed details about its AGI processors, which include up to 136 cores and utilize a 2 TB/s UCIe fabric link alongside a 12-channel memory controller.

Server CPUs, DPUs, 2.5D/3D advanced packaging, Memory modules

SK hynix showcased a comprehensive range of memory solutions optimized for AI computing infrastructure at the DTF 2026 conference.

Memory modules, Memory interface / buffer chips

SK hynix displayed a complete range of memory solutions optimized for AI infrastructure at DTF 2026.

HBM, DRAM, Memory modules

SK hynix presented a comprehensive range of AI-optimized memory solutions at DTF 2026.

Memory modules, CXL controllers & memory expanders

SK hynix showcased its range of memory solutions designed for AI infrastructure at DTF 2026.

Memory modules, CXL controllers & memory expanders

OpenAI presented its custom AI accelerator, Jalapeño, at Hot Chips 2026.

Hyperscaler custom ASICs, Silicon photonics & co-packaged optics

Google unveiled its new eighth-generation TPU family (TPU 8t for training and TPU 8i for inference) at Hot Chips 2026.

Hyperscaler custom ASICs, Scale-up fabric

SambaNova showcased its latest-generation RDU, the SN50, at the Hot Chips conference.

Merchant GPUs / accelerators, Hyperscaler custom ASICs

Microsoft unveiled details about Maia 200, a second-generation AI accelerator designed for server inference, at the Hot Chips 2026 conference.

Merchant GPUs / accelerators, Hyperscaler custom ASICs

Samsung unveiled LPDDR5X-PIM, which incorporates logic directly into the memory to dramatically accelerate AI inference performance.

Memory modules, Memory interface / buffer chips

OpenAI debuted an in-house chip, claiming it offers superior performance and efficiency compared to Nvidia's high-end GPUs.

Hyperscaler custom ASICs, Merchant GPUs / accelerators