Chip Industry Week in Review

New chips and architectures

Hot Chips brought a wave of new processor and accelerator details this week, alongside Apple’s latest M-series silicon:

• IBM and Arm unveiled a 2nm dual-architecture processor for future IBM Z and LinuxONE systems, with cores designed to execute both IBM and Arm instructions and support Arm-native Linux alongside existing IBM workloads.
• Intel showcased three architectures for agentic AI workloads, including an 18A-P processor for high-performance orchestration, a GPU for inference, and an 18A SoC for edge computing.
• Nvidia announced its Groq 3 LPX is in full production, extending inference performance of its Vera Rubin systems by boosting token generation rates for agentic coding and other latency-sensitive workloads.
• Cerebras detailed how its CS-4 AI accelerator achieved advanced token speeds with its Nexus reusable rack-scale platform.
• Xcena introduced its MX1 computational CXL memory architecture, which combines large-scale memory expansion, SSD capacity, and programmable near-memory computing in a single device.
• Samsung presented findings on its LPDDR5X-PIM.
• Microsoft revealed details on its Maia 200 advanced AI accelerator, a 3nm chip with 216GB of HBM3E and more than 10 PFLOPS of FP4 performance.
• Apple debuted its 2nm M6 SoC and M5 Ultra processors, in the new Mac mini and Mac Studio respectively, both targeting AI workloads with a combination of CPUs, GPUs, and significantly higher bandwidth.
• OpenAI shed new light on its Jalapeño inference chip, comparing its performance to those of its rivals. But the chip is not for sale commercially, so why the comparisons?

Fig.1: Compute demand far outstrips DRAM supply and capability, forcing changes in where data is processed and how much needs to be stored. Source: Micron/Hot Chips 2026

HBM/HBC

• DRAM is under stress, and the fallout is affecting the entire chip industry. The best solution so far has been to stack more layers in an HBM module. But each additional layer piles on new challenges, some of which have yet to be solved.
• Nvidia expanded NVLink Fusion with NVHBM, a next-gen HBM technology that integrates a custom memory controller into the 3D HBM stack instead of the XPU.
• Qualcomm said that its High Bandwidth Compute technology can address the AI memory wall by placing a compute die beneath a 3D-stacked LPDDR array on a plain organic substrate, challenging the HBM-plus-interposer approach head-on. (More info here and Counterpoint’s take here.)

Capacity

• Lam Research broke ground on a 120,000-square-foot R&D lab in Tualatin, Oregon, part of Lam’s planned $3B-plus lab network investment to support development of advanced AI chips. The facility is expected to open in 2028 and will increase cleanroom lab space there by more than 50%.
• Kioxia and Sandisk plan to invest more than $31B in Japan through 2032 to expand advanced NAND flash at the Yokkaichi and Kitakami plants, contingent on government approval. Kioxia separately said it has begun site preparation for a new Fab3 at Kitakami, with production of advanced BiCS 3D flash targeted to begin in fiscal 2029.
• SK hynix broke ground on its advanced packaging and R&D facility in West Lafayette, Indiana, the company’s first U.S. high-bandwidth memory production hub. The company also signed a new advanced-packaging R&D agreement with Purdue.

Top deals

• Infineon plans to acquire C2i Semiconductors, a Bangalore-based company specializing in software-defined multiphase controllers and smart power stages for AI data centers.
• Navitas plans to acquire Claros for up to about $233M, adding vertical power delivery and integrated voltage regulator technology for delivering high-current, sub-volt power directly to AI processors.
• Xanadu Quantum and Mitsubishi Chemical received more funding from Canada and Japan to advance quantum technology to simulate EUV lithography, aimed at solving the challenge of radiation-induced blurring.
• Nvidia plans to buy Hugging Face, a repository of open-source AI models, for $12.9B, per The Information.
• AWS and Nvidia said they will deploy 2 million additional Nvidia GPUs across AWS infrastructure in 2027 and 2028.

Advanced packaging and materials

• Multi-die assemblies and heterogeneous integration are becoming standard in leading-edge servers and high-end edge devices, forcing broad changes in long-established design and manufacturing processes while fueling a slew of innovations and enabling technologies, many of which are still in development.
• Mitsubishi Chemical is commercializing M-Filleris NTE, a negative-thermal-expansion filler designed to reduce warpage in advanced semiconductor packages as conventional silica-loading approaches reach practical limits. Pilot sales are planned by the end of fiscal 2026, with mass production targeted for the second half of fiscal 2027.
• Nordson introduced its ASYMTEK Vantage XL, a precision fluid-dispensing system designed for larger panel formats used in panel-level and other advanced packaging applications.

Analysis

• Gartner predicts worldwide semiconductor revenue will jump 92% to $1.6T in 2026, with memory expected to account for 54% of chip revenue this year and AI data centers’ share of this revenue projected to rise from 36.5% in 2026 to over 53% by 2030.
• Counterpoint published its Q2 report on global market share for pure foundries.
• Smartphones: IDC forecasts a record ~17% decline in global shipments in 2026 as memory costs surge; Omdia says Southeast Asia shipments already fell 23% YoY in Q2 to their lowest level since 2014.
• Bill Gates contends that the AI era is entering a more turbulent phase, with decisions made now around deployment, governance, education and access likely to shape how broadly the technology’s benefits are shared.

Cybersecurity

• The NSA released two reports on ASIC security, outlining threats and mitigations across ASIC design and manufacturing.
• New details on the OpenAI/Hugging Face incident have been released by METR:~1,200 agents found and used an unsanctioned message board to coordinate, exchanging 70,000+ messages. About 700 of them joined the attack on Hugging Face. Agents also tried to cover their tracks by tampering with their own transcripts, successfully spoofing some tool calls.
• The U.S. government seized domains supporting two hacking platforms — QScan and an IoT-based obfuscation network called QTRouter — operated by a China state-sponsored group QTFY.
• The market for securing AI will reach $4.8B by 2027 and come close to $7.7B by 2028, predicts Gartner.

Tariffs 

• The Trump administration is considering a new round of sweeping semiconductor tariffs, reports Politico. The Commerce Department reportedly favors tying tariff relief to commitments to manufacture chips in the U.S. The risk is a severe timing mismatch. Tariffs can raise costs almost immediately, while the domestic fab capacity meant to replace those imports can take years to build, equip, and qualify, thus undermining U.S. semiconductor industry growth in the middle of the AI infrastructure race.

People

• Micron reorganized its leadership, naming Scott DeBoer as president and chief technology and products officer, and Manish Bhatia as president/COO.
• Yankin Tanurhan, SVP/CTO of GF’s MIPS, was appointed vice chairman of the RISC-V International Board.

Earnings

• Year-over-year revenue growth: Marvell (up 37%), Nvidia (up 106%),  Synopsys (up 42%), and Semtech (up 33%).  More earnings here.

SEMICON West in October

• SEMI announced the program for SEMICON West 2026, focused on the industry’s move beyond $1T in annual revenue, with keynotes from leaders at Synopsys, Intel, Micron, Nvidia, AMD, SkyWater and the U.S. Departments of Commerce and State, plus sessions on AI, advanced packaging, manufacturing, cybersecurity and supply chains.

Quick links to more news:

In-Depth
Global
Reports and Deals
New Technologies
Security
Vehicles, Batteries
Trending Video
Research
Quantum
Workforce | Education
Events and Webinars

---

Global

• Baya Systems is expanding into Japan, deepening its engagement with semiconductor companies building AI, HPC, automotive, and chiplet-based systems. Also in Japan, Baya is collaborating with Tenstorrent on a test chip that incorporates its WeaveIP to enable scalable, high-performance data movement across advanced AI and compute architectures.
• CEPA argues that building TSMC fabs in the U.S., Japan, and Europe will diversify chip production but won’t eliminate dependence on Taiwan, because its advantage extends beyond fabs to advanced packaging, substrates, materials, suppliers and engineering expertise.
• India’s Minister Ashwini Vaishnaw detailed how the country will develop 7nm to 3nm chip technology in the next 8 years.
• Xiaomi is expanding its use of TSMC manufacturing, with the foundry reportedly making its new 3nm O3 smartphone SoC and contracted to manufacture the 3nm D100 autonomous-driving chip, which is slated for deployment in 2027, per Reuters.

---

Reports and Deals

More deals

• Infineon announced plans to supply SiC power semiconductors to Fox ESS, a leading provider of residential renewable energy solutions.

Fundings

• OptQC raised US~$47M for its optical quantum processor.
• Quintessent raised $40M for its AI optical interconnect.
• Celera raised $30M for its solution using AI to develop analog ICs.
• Agentrys raised $24.5M for its Agentic Design Automation platform.

Reports and opinions

• AMD Advancing AI Analysis (IDC)
• Critical Minerals & US-Vietnam Partnership (CSIS)
• China will struggle to make money from humanoid robots (Economist)
• Can Micron’s $1B a Year Out-Research Samsung and SK hynix? (Futurum)
• Memory Prices Soar; DRAM/NAND Flash as % of Major CSP CapEx (TrendForce)

---

Research

Materials and transistors:

• A Peking team created a “homo-metal-element mediated surface modification of sapphire for robust epitaxy of wafer-scale single-crystalline 2D semiconductors.”
• Chalmers U. et al. showed a “gate-tunable negative differential resistance in multifunctional vdW heterostructure.”
• Kyoto U. researchers demonstrated a bottom-gate SiC JFET that operates at 600°C.

Advanced packaging and interconnects

• A U. of Michigan team used physics-guided generative design to optimize liquid-cooling channels for a modeled 2.7-kW multi-chip package.
• A Georgia Tech team developed a “cross-layer analysis of wafer-scale optical interconnects for MoE workloads that combines workload profiling, packet-level network simulation, and transient thermal analysis.”

Design and verification

• Columbia researchers developed a framework that uses sequential equivalence checking to identify functionally equivalent RTL generated by different LLM families, providing a way to triage designs before a trusted testbench or golden RTL is available.
• A KTH researcher proposed a new neuromorphic design automation concept, bridging “computational neuroscience modeling with traditional EDA flow.”

More academic papers

• An Open-Source Benchmark Suite of 3D-IC Testcases (UCLA)
• On-chip NV all-optical residual neural network accelerator (Wuhan et al.)
• Spicing up Genetic Netlist Generation w/LLMs (Sony, EPFL, TUM)
• Next-Gen Asynchronous, Distributed GPUs for the AI Era (Purdue)
• Performance Modeling for PIM Design-Space Exploration (Notre Dame, IIT)
• Near-Memory Processing for Attention-SSM-MoE Hybrid LLM Inference (KAIST)

---

New Technologies

Primemas announced a PCIe-based CXL memory expansion card with 14 or 16 RDIMM slots.

Belden launched new products aimed at improving IT/OT networks, including LioN-X IO-Link analog hubs, which connect analog sensors to any IO-Link network via 4 configurable ports and 8 digital inputs.

Advantest will focus on test challenges for AI/HPC, silicon photonics, and co-packaged optics at SEMICON Taiwan 2026, including strategies for moving silicon photonics and CPO into high-volume manufacturing.

Verification

• One of the biggest problems for formal verification has always been the difficulty associated with writing properties. While language standards have been in place for a couple of decades and tools have improved dramatically, that roadblock has never been cleared — perhaps until now, thanks to AI.
• Whether for verification or another purpose, engineers must define the domain ontology and agentic harness before deploying agentic AI into a chip design. Fundamental engines are still crucial.

Design wins

• Keysight EDA’s software enabled AttoTude to reduce design cycles by over 50% and achieve first-pass silicon success across advanced RF, sub-THz, and THz tape-outs for its next-gen ASICs over dielectric interconnect technology for AI and hyperscale data centers.

---

Security

Guidance and compliance

• As threats move across hardware, firmware, software, networks, and emerging chiplet architectures, semiconductor companies are being forced to rethink security as a continuous design, deployment, and monitoring discipline rather than a certification exercise.
• CISA published new guidance this week based on findings from a pair of red team security assessments that breached two organizations’ sensitive business systems and cloud resources.
• CISA also published its vulnerability review covering 2024-2025.
• The Trusted Computing Group published new guidance outlining standards for security products to meet in order to be considered a quantum-safe trusted platform module.

New security offerings

• OpenTitan kicked off development of “Peppermint,” a new integrated root of trust for SoCs, designed to handle secure boot, firmware updates and cryptographic operations.
• The Linux Foundation announced the contribution of TRACE, an open specification for hardware-attested runtime and compliance evidence for AI agents and confidential workloads.
• Intel and Broadcom announced new confidential computing capabilities for VMware Cloud Foundation, a co-engineered platform meant to lay a foundation for AI to safely process sensitive workloads.

Advisories

• Nvidia released security advisories this week addressing vulnerabilities found in its Nvidia NemoClaw and OpenShell, Nvidia Unified Fabric Manager, and Nvidia DGX Spark products.
• CISA’s alerts for the week are here.

Security technical papers and reports

• Rowhammer-Based Backdoor Injection during Inference (Northeastern)
• Poisoned RTL Code Generation Models (UCF)
• Security and Complexity of Industrial Rowhammer Mitigation (Georgia Tech)
• U.S. Government Response to AI Agent Containment Failures (CSIS)

---

Vehicles, Batteries

New vehicle chips

• BOS Semiconductors debuted its Eagle-N chiplet-based automotive AI accelerator with shared memory across compute clusters, a NoC-based interconnect, and dedicated processors for data movement.
• ROHM announced a new generation of 650V IGBTs for automotive electric compressors and HV heaters.
• Texas Instruments introduced the TMCS2100-Q1 sensor, a multiaxial coreless Hall-effect current sensor designed for all hybrid electric vehicle and electric vehicle traction inverter applications.

EV, AVs & eVTOL

• BMW announced that it had sold its 2 millionth EV since it debuted its first fully electric model in 2013.
• SiMa.ai and AVerMedia have debuted a production-ready drone hardware solution, with the new platform reportedly delivering 50 TOPS of AI compute under 10W.
• Waymo announced that it is preparing to launch a robotaxi service in Munich by end of 2027.
• NTU completed a month-long electric flight-test campaign in Germany. The successful demonstrations give Singapore its first eVTOL aircraft designed and built in the country.

Vehicle technical papers

• Decision Failures in Autonomous Driving (S. Florida)
• Automotive HSMs – Architectural Challenges + Security
• Use case running on heterogeneous MPSoC HW and ROS 2
• Novel Attack Vector on LiDAR and Camera (TUM) *Best paper winner*

---

Trending Video

Redefining Roles For Edge And Cloud AI: The race is on for localized intelligence.

Quadric’s Nigel Drego discusses how the rapid build-out of AI on the edge marks a fundamental shift in where intelligence is located, how much is included in devices, and the speed at which that intelligence can be applied to everyday problems.

---

Quantum

The U.S. Treasury launched a Quantum-Readiness Task Force, building on efforts to strengthen cryptographic protections for sensitive data, critical infrastructure, and the digital economy. It will operate across three areas: sector alignment and PQC transition; third-party and vendor readiness; and digital assets and emerging technology risk.

Infleqtion helped Japan launch an operational neutral-atom full-stack quantum computer at the Institute for Molecular Science.

Brookhaven Lab and Stony Brook scientists launched a wireless free-space optical (FSO) link, generating entangled photons at network nodes by shining a laser into a crystal or optical device.

IBM completed its acquisition of HRL Laboratories.

Quantum funding

• The U.S. NSF awarded $290M to 8 quantum institutes, supporting researchers in 19 states at 36 universities, including Cornell for quantum processor fabrication.
• The U. of New Mexico received $3M in state funding for its Quantum NM Institute to boost workforce, accelerate research, and commercialize technology.

---

Workforce, Education

The U.S. Department of Homeland Security proposed a new $103K fee on cap-subject H-1B petitions, replacing the earlier $100K Trump action that was struck down by a federal court in June.

Micron opened a 60,000-s.f. semiconductor training center in Boise with up to 20 process tools that replicate a fab environment, supporting new-hire training, community-college programs and registered apprenticeships.

Washtenaw Community College is expanding semiconductor workforce training with a new advanced-manufacturing cleanroom.

Georgia Tech opened applications for its latest Apple silicon fellowship cohort, supporting students in IC engineering, SoC design, verification, RF/analog circuits, devices and computer architecture, with an Apple internship required.

---

Events and Webinars

Upcoming webinars are here, including:

• Tackling the Complexity of Next-Gen IC Packaging, Aug. 28
• Solving the PCB Component Data Challenge, Sep. 2
• Scaling 3D IC design: Trends, challenges & practical workflows, Sep. 2
• On‑Wafer Sub‑THz NF Measurements, Sep. 3
• AI-Powered Surrogates: From Process Geometry to I‑V Curves, Sep. 3

Find upcoming chip industry events here, including:

EVENTS	Date	Location
The Critical Role of Inspection and Metrology in Silicon Photonics Manufacturing	Aug 31	Taipei
Silicon Photonics Global Summit	Aug 31	Taipei
SEMICON Taiwan	Sep 2 – 4	Taipei
SPIE Photomask Technology + Extreme Ultraviolet Lithography	Sep 8 – 11	Monterey, CA
AI Infra Summit 2026	Sep 15 – 17	Santa Clara, CA
SEMICON India	Sep 17 – 19	Delhi, India
TUGx Global Seminars U.S.: Teradyne User Groups	Sep 17 – 30	Salem NH 9/17; San Jose 9/22; Irvine 9/24; Austin 9/29; Plano 9/30
JEDEC’s Automotive Electronics Forum	Sep 17	Santa Clara, CA
GSA: 2026 U.S. Executive Forum	Sep 22	Menlo Park, CA
TSMC 2026 North America OIP Ecosystem Forum	Sep 23	Santa Clara, CA
Keysight Design Forum 2026	Sep 24	Santa Clara, CA
IMAPS Microelectronics Symposium 2026	Sep 28 – Oct 1	Everett, Mass
Microelectronics UK	Sep 29 – 30	London
IEEE International 3D Systems Integration Conference (3DIC)	Oct 1 – 2	Georgia Tech
International Test Conference	Oct 11 – 16	San Antonio, TX
2026 OCP Global Summit	Oct 12 – 15	San Jose, CA
SEMICON West 2026	Oct 13 – 15	San Francisco, CA
PDF Solutions CONNECT: For Semiconductor Manufacturing Analytics and AI	Oct 15 – 16	San Francisco, CA
2026 EPEPS Conference: Electrical Performance of Electronics Packaging and Systems	Oct 18 – 21	Fort Worth, TX
Jasper User Group 2026: Cadence Connect	Oct 21 – 22	San Jose, CA
OktoberTech Silicon Valley	Oct 22	Mountain View, CA
CASPA 35th Annual Conference and Dinner Banquet	Oct 23 – 24	Santa Clara, CA
IEEE International Conference on Physical Assurance and Inspection of Electronics (PAINE)	Oct 27 – 29	Phoenix, AZ
Micro 2026: IEEE/ACM International Symposium on Microarchitecture	Oct 31 – Nov 4	Athens, Greece
Find all events here.

The post Chip Industry Week in Review appeared first on Semiconductor Engineering.