SK Hynix participated in a funding round for Etched, an AI inference startup developing rack-scale systems using Low Voltage Inference (LVI) and Cluster Scale Memory (CSM) to address GPU limitations.
Si2 and industry leaders including Nvidia, Cadence, and Synopsys are establishing standardized benchmarks for LLMs in chip design at DAC 2026. This initiative aims to create consistent evaluation methods for AI agents in Electronic Design Automation (EDA) workflows. The move supports long-term AI adoption in semiconductor design but has no immediate impact on component pricing or supply.
Poolside released Laguna S 2.1, an 118B parameter model matching larger models on coding benchmarks. It runs on a single NVIDIA DGX Spark, addressing the West's need for open-weight models. Bullish for NVIDIA inference hardware demand.
This week's semiconductor industry highlights include Intel expansion, CHIPS Act funding, HBM standardization, and Rapidus collaboration. These events reflect policy support and capacity expansion trends, though specific price data is absent.
Alok Jain has transformed Cadence India into a chip-design powerhouse and is focusing on AI as the next growth area. This highlights the strategic shift of design capabilities to India and Cadence's long-term outlook.
Former NVIDIA researcher Mark Ren leads Agentrys in launching Agentic Design Automation (ADA), a platform utilizing AI agents to automate chip design workflows and engineering tasks.
MooresLabAI launched a platform of AI agents to automate chip design and verification, aiming to solve engineering bottlenecks. This represents a shift toward AI-native engineering but has no immediate impact on component prices.
Cadence and Samsung Foundry are expanding certified 2nm and 3D-IC design capabilities for AI infrastructure systems.
Researchers propose using Quantum Multi-Valued Decision Diagrams (QMDD) to simplify quantum circuit simulation, addressing exponential complexity. This represents a technical advancement in EDA and verification, though it currently has no direct impact on spot prices or supply.
Samsung, Intel, and TSMC have initiated 2nm mass production and are outlining roadmaps for 1.4nm nodes. TSMC focuses on specialized nodes, Samsung on yield, and Intel on aggressive GAA and High-NA EUV adoption.
Rapidus targets mass production of 2nm chips in Hokkaido in H2 2027, with pilot production already underway. The Japanese foundry startup is also developing advanced packaging capabilities to compete with TSMC, Samsung, and Intel. Neutral: This is a long-term R&D milestone with no immediate impact on component supply or pricing.
Analysis suggests major hyperscalers (Microsoft, Google, Meta, Oracle) designing custom AI silicon face a strategic dependency on TSMC, creating a need for a second source. The article frames Intel's potential foundry pivot, highlighted by a recent signal with Elon Musk, as a customer-driven capacity shaping move akin to NVIDIA's historical TSMC commitments. This structural shift in the foundry landscape could alter long-term supply dynamics but lacks immediate trading catalysts.
The UALink Consortium has released version 2.0 of its open GPU interconnect specification, adding support for 200G networks and in-network compute. The group aims to provide a high-performance alternative to Nvidia's proprietary NVLink, but commercial silicon based on the v1.0 spec is not expected until 2027. Neutral: This is a long-term R&D development with no immediate impact on component supply, demand, or pricing.
EDA industry is heavily hyping AI, but major customers like Samsung and Nvidia are building their own internal AI stacks. The article notes current progress focuses on optimizing legacy workflows rather than solving core verification problems. Watch: EDA tool vendors face competitive pressure from internal solutions, but price impact is currently neutral.
Rumors indicate NVIDIA's next-gen GeForce RTX 60-series will use the 'Rubin' architecture on TSMC 3nm nodes, with flagship chips like GR202. The series is expected to stick to mature 3nm FinFET technology rather than sub-2nm nodes. Watch: This confirms NVIDIA's strategy to maintain high clock speeds on established nodes.
Ambient Scientific launched the GPX10 Pro SoC featuring 10 programmable AI cores and in-memory compute for edge AI applications. The company also announced the upcoming GPX64 for robotics and ADAS later this year. Neutral: New product launch with no immediate supply or price impact.
Cadence is expanding its collaboration with NVIDIA to accelerate AI chip and system design solutions. This strategic partnership focuses on Design for AI and AI for Design capabilities, signaling strong demand for advanced semiconductor design tools.
Meta announced four generations of its in-house MTIA inference accelerator chips, developed with Broadcom, to be deployed over the next two years. The chips emphasize memory bandwidth and efficiency over raw compute and are designed for modular upgrades within existing data center infrastructure. Neutral: This represents a long-term, internal demand shift for custom AI silicon but has no immediate, direct impact on the merchant component market.
Competition in high-speed SerDes technology intensifies as AI models drive demand for chip-to-chip interconnect bandwidth. MediaTek reportedly entered Google's TPU ecosystem with its 224G SerDes, while NVIDIA is opening its NVLink Fusion SerDes IP to partners. Neutral: This is a long-term technology landscape shift with no immediate, specific impact on component supply, pricing, or production.
Cadence announced a chiplet platform aimed at supporting physical AI hardware development. This is a design tool and methodology announcement from an EDA software vendor, not a hardware supply chain event. Neutral: No immediate impact on component supply, pricing, or procurement.
Whose chips are really inside the hottest new products? We open them up and show you, one part at a time.
iPhone DuoFirst-pass BOM: 37 parts, 12 brands and 16 public MPNs; A20 Pro, C2, N1 and the power/RF chains are located, while the NAND brand stays unconfirmed.See the full teardownAll teardowns| Analog | $0.68 | −5.9% |
|---|---|---|
| Connector | $0.75 | +0.2% |
| CPU | $386.74 | −21.3% |
| DDR4 | $88.40 | −0.2% |
| DDR52026-W13 | $14.89 | −11.2% |
| GPU | $478.20 | +8.2% |
| MCU | $0.68 | −61.7%Unusual |
| MLCC | $0.01 | +54.0%Unusual |
| MOSFET | $0.12 | −82.8%Unusual |
| PMIC2026-W13 | $0.44 | −12.5% |
| SSD/HDD | $288.43 | n/a |