Chip Industry Week In Review

The energy and buzz at this year’s Design Automation Conference surprised everyone — despite initial concerns about its location in Long Beach, California, which is well off the beaten path for semiconductors. Hot topics included AI automation, multiphysics simulation, compute-in-memory, chiplets, physical AI, and the development and orchestration of AI agents from design through manufacturing. At the conference: Synopsys unveiled new autonomous agentic AI workflows for chip design, developed in collaboration with Microsoft. Synopsys also worked with Nvidia on new long-running, fully agentic workflows spanning from EDA to CAE. Cadence, in collaboration with Nvidia, introduced an agentic AI platform focused on PCB and advanced packaging design that coordinates specialized agents across planning, implementation, and multiphysics analysis. Siemens EDA introduced self-verifying agentic AI workflows for IC design that use Nvidia AI to help long-running, domain-specific agents reason, act, and continuously validate decisions against deterministic, physics-based EDA engines. Fig.1: Photos from DAC. Source: Semiconductor Engineering’s Liz Allan Advanced packaging ASE is raising its 2026 capital expenditures by $2B to about $10B due to strong demand. Intel Foundry is expanding advanced packaging at its Rio Rancho, New Mexico, operations using Foveros, EMIB and EMIB-T to support larger chiplet-based AI systems, with package scaling reaching roughly 8× the reticle limit today and more than 12× by 2028. Lens Technology and Intel announced a strategic collaboration to develop glass substrate-based advanced packaging aimed at increasing interconnect density, performance and power efficiency for AI and data center systems. UMC approved a phased capacity expansion that will add cleanroom space and equipment at its Phase 4 facility in Singapore while beginning construction of a new fab shell at its Tainan campus in Taiwan. The Singapore expansion will increase silicon photonics capacity, while the Tainan project will house future Phase 7 and Phase 8 facilities and expand UMC’s advanced packaging footprint for AI and edge-computing applications. A 7.1-magnitude earthquake in Japan’s Kumamoto region disrupted manufacturing across Kyushu, prompting a temporary halt to operations at automotive and chip manufacturing plants, including TEL and TSMC‘s JASM. China has begun producing domestically developed immersion DUV lithography systems, with state-backed Shanghai Aishengna leading an effort, reports Reuters. The systems are expected to be delivered this year, although they remain well behind ASML in the throughput, overlay, yield and reliability needed for high-volume manufacturing. Market analysis The SIA released its State of the Industry Report, highlighting that global chip sales reached a record $796B in 2025 and are projected to hit $1.5T in 2026. U.S.-headquartered companies held 53% of the market, while domestic semiconductor investments announced since 2020 exceeded $770B across 160 projects. Global silicon wafer shipments rose 7% year over year and 9% sequentially in Q2 2026, SEMI reported, driven by AI demand and improving industrial and automotive markets. Wafer fab equipment sales are expected to rise 25% to about $162B in 2026 and approach $220B by 2031, according to Yole Group, fueled by AI, advanced logic, memory and packaging investment. TrendForce expects DRAM shortages and rising prices to continue in 2027 as HBM and AI servers absorb capacity, while NAND supply may loosen as new fabs and higher-layer products increase output. CHIPS Act funding The U.S. Department of Commerce signed letters of intent with seven companies for up to $874M in CHIPS R&D incentives, covering integrated photonics, advanced packaging, substrates, materials, novel memory, compute architectures, and supply-chain security. Advance next-gen silicon photonics, optical materials, wafer technologies and advanced packaging, including 3D hybrid bonding for near-packaged and co-packaged optics. Thermodynamic sampling units (TSUs) that use natural thermal fluctuations to probabilistically solve complex problems spanning simulation, optimization and AI at a fraction of the energy consumed by conventional computing approaches. Large-diameter, indium-phosphide-free substrates for photodetectors and lasers used in AI photonic interconnects. Notable deals SK Group and Nvidia are planning a $500B+ partnership, which includes a 2-gigawatt-scale Vera Rubin AI cloud built by SK Telecom, and a long-term agreement for SK hynix to supply memory to Nvidia, including HBM. Intel Foundry completed the RAMP-C program with the U.S. Department of Defense, delivering validated prototypes and design enablement on Intel 18A for commercial and defense customers. The program now feeds into Secure Enclave, which is intended to provide trusted, high-volume U.S. manufacturing of leading-edge semiconductors using Intel 18A and advanced packaging. Silvaco and Nvidia are collaborating on physics-based digital twins for semiconductor design and manufacturing. Raking it in ChipAgents raised $60M to expand its agentic AI platform, which is capable of executing multi-step chip design and verification tasks. Taiwanese prosecutors detained an Nvidia employee as part of an expanding Supermicro investigation into a suspected scheme to route servers containing export-restricted Nvidia chips toward China, reports Reuters. The investigation began in May with the seizure of 50 servers. Nvidia itself was not accused of any wrongdoing. AMD will partner with Korea’s Ministry of Science and ICT to boost AI infrastructure in the country. Marvell intends to spend $250M to expand operations in India over the next three years. EuroHPC JU launched an AI Gigafactories call to support up to seven large AI-compute facilities across at least seven EU member states, pairing public funding with an effort to raise about US$23B, in private investment. Americas OECD says Panama could become part of the semiconductor ecosystem by investing in talent and reliable utilities, developing local suppliers and customers, and improving coordination through its Commission for Innovation in Microelectronics and Semiconductors. Omdia said AI demand is driving a 94% jump in its 2026 semiconductor revenue, with memory ICs now expected to account for more than half of chip revenue and bottlenecks in HBM, advanced packaging and leading-edge nodes lasting into at least 2027. Infineon rolled out a solid-state isolation technology built for high-volume industrial automation applications. Also, the company’s SEMPER NOR Flash Memory earned qualification for ASPEED Technology’s baseboard management controller. Nvidia announced AI models, libraries, and a toolkit for agentic RTL coding, physics reasoning, and simulation. Siemens released the latest version of its Simcenter industrial simulation software with expanded AI-driven simulation, GPU acceleration, and new multiphysics workflows. Eliyan debuted a power-reach configurable 224G PAM4 SerDes architecture optimized for chip-to-chip and chip-to-module electrical channels. Openchip announced a Linux-capable 64-bit RISC-V processor built on a 2nm GAA process and intended for industrial and AI workloads. Kioxia released new UFS 5.0 embedded flash memory solutions with MIPI M-PHY v6.0 and the UniPro v3.0 protocol. Research Cornell researchers expanded a low-power microwave neural-network chip with a technique that encodes information directly into RF signals, lowering the bandwidth and energy required for communications. A team from Ghent University integrated solution-processable ZnSe/ZnS quantum dots with a CMOS-compatible silicon nitride photonics platform, producing blue lasing from 425 to 445 nm. Researchers from Shanghai University, TSRI, and others created a “reconfigurable field-effect transistor based on the scalable dielectric oxide-vdW quasi-floating-gate configuration, enabling nonvolatile polarity switching and multi-state programmability.” Electrospray Deposition of Silver Films on Insulators for EMI Shielding (SUNY) Security New security alliance and tools Numerous companies joined an Nvidia-led “Open Secure AI Alliance for AI Safety and Security” in the wake of OpenAI agents breaching parts of Hugging Face’s infrastructure this month. AI leaders OpenAI, Google and Anthropic are not listed among the alliance’s members. Freedom Factory rolled out a post-quantum hardware wallet, claiming it “cannot be cracked by quantum computers and delivers digital sovereignty to every ethereum and EVM user on the planet.” Microsoft introduced its first security-focused AI model, MAI-Cyber-1-Flash. The company said that when unified with MDASH, the new model outperforms Mythos, Gemini, and GPT on CyberGym standards. SEALSQ said its post-quantum security chip is now sampling to customers for use in physical AI applications. Government security updates The FCC updated its list of products considered national security risks to include foreign-made “advanced robotic devices,” defined as mobile robots including humanoids and quadrupeds. NIST is seeking public comments to support a new security analysis covering threats to AI data centers. Comments are due by Sept. 25. The NSF announced its first CyberAICorps Scholarship for Service awards to 14 universities developing programs that combine AI and cybersecurity education. National cybersecurity agencies in Australia, the U.S., Canada and the UK published new joint guidance for critical infrastructure operators to isolate vital systems and navigate cyberattacks and disruptions. Attacks and alerts Lava found more than 36,000 internet-exposed baseboard management controller/IPMI interfaces, including 24,000 that disclosed password-derived hashes and that could enable offline cracking and remote server takeover. More than 30 municipal water systems in Minnesota were targeted in a coordinated cyberattack, with investigators telling media that the breach was likely the work of Iran-affiliated hackers. Cybernews researchers discovered an exposed attacker server tied to an active ransomware-as-a-service campaign targeting V-Silicon. Decentralized Compute on Untrusted HW Using Intel TDX and Encrypted CVMs (Intel, Manifold) Vehicles, Batteries Accellera’s Functional Safety Working Group released its Functional Safety Language 1.0 draft for public review. The draft proposes a common, machine-readable format for exchanging safety data. The standard could be particularly useful in automotive chip development, where ISO 26262 compliance requires safety information to remain consistent and traceable across increasingly complex supply chains. Comments are due August 14. Auto industry news BMW reportedly will cut up to 8,000 jobs globally. Qualcomm will be BMW’s lead chip provider for the digital cockpit and ADAS/AD systems over the next 10 years. AI-driven memory shortages have carmakers weighing price-hike decisions, with GM executives saying that the company’s memory-related material and other costs this year will grow by $1.5B to $2B on a recent earnings call. The comments come after GM, Ford and several other auto industry firms inked long-term memory supply agreements with Micron earlier this month. San Francisco Bay Area political leaders are calling for stricter federal regulations on autonomous vehicles following a series of reported safety and traffic congestion issues linked to Waymo vehicles during the region’s Fourth of July holiday celebrations this month. Verifying Networks On Chip: More data, more processing elements, and more memories are making it much harder to maintain consistent throughput in a multi-die assembly. There is more concurrent traffic, more potential interactions, and more protocol layers. Ashish Darbari, CEO of Axiomise, talks about the role of coherency in serialization bugs, how to ensure data is secure, and how to ferret out the root cause of problems like silent data corruption using formal technology. Quantum IBM claims it and collaborators have demonstrated quantum advantage in three tasks: random-circuit sampling with error detection with the University of Chicago; modeling Floquet dynamics with Qedma, RIKEN, and BlueQubit; and estimating how information spreads through heterogeneous quantum systems with Algorithmiq. IonQ expects to complete its acquisition of foundry SkyWater Technology on July 31, which will continue to serve a full range of customers. The Pittsburgh Supercomputing Center will build a hybrid quantum-classical computer with help from HPE, Rigetti Computing, and a $5M NSF grant. HRL says its quantum processor can run itself by replacing “racks with a custom control chip that sits near the qubits, in the extreme cold.” Delft researchers demonstrated a 5-qubit silicon processor that coherently transported spin qubits along a shuttling bus to four interaction sites. The architecture performed 4-qubit parity checks and generated entanglement across all five qubits, addressing connectivity and wiring challenges in scaling semiconductor quantum processors. Qnity and the University of Delaware formed a semiconductor workforce and manufacturing-innovation partnership to give engineering students hands-on experience through projects, an extended work-study pilot, automation efforts, and semiconductor materials research. Karlsruhe Institute of Technology (KIT) and SUSS MicroTec are planning a semiconductor partnership that will create a new KIT professorship in advanced packaging and heterogeneous integration, backed by German government funding. SUSS also plans to invest about $52M in a Karlsruhe application and development center. InP and SiPho join CMOS as critical technologies. Lasers, CPO and OCS will be everywhere (indium phosphide, silicon photonics, co-packaged optics, optical circuit switch). Reducing variation in manufacturing, monitoring behavior over time, and targeting specific workloads can have a big impact on power, performance, and area/cost. This website uses cookies to improve your experience while you navigate through the website. The cookies that are categorized as necessary are stored on your browser as they are essential for the working of basic functionalities of the website. We also use third-party cookies that help us analyze and understand how you use this website. We do not sell any personal information. By continuing to use our website, you consent to our Privacy Policy. If you access other websites using the links provided, please be aware they may have their own privacy policies, and we do not accept any responsibility or liability for these policies or for any personal data which may be collected through these sites. Please check these policies before you submit any personal information to these sites. Necessary cookies are absolutely essential for the website to function properly. This category only includes cookies that ensures basic functionalities and security features of the website. These cookies do not store any personal information. Any cookies that may not be particularly necessary for the website to function and is used specifically to collect user personal data via analytics, ads, other embedded contents are termed as non-necessary cookies. It is mandatory to procure user consent prior to running these cookies on your website. Leave a Reply