Samsung LPDDR5X-PIM at Hot Chips 2026

servethehome.com Aug 26, 2026

Samsung LPDDR5X-PIM at Hot Chips 2026

Curated by Shubham · TechCapsules · Aug 26, 2026

AI-summarised brief · reviewed before publication

Samsung unveiled its LPDDR5X‑PIM at Hot Chips 2026, the first processing‑in‑memory solution built on LPDDR5X DRAM. The new memory targets AI inference, offering 16 GB per rank in a 561‑ball JEDEC package and delivering 614 GB/s bandwidth—eight times the 76.8 GB/s of conventional DRAM. It supports 15 multi‑precision data types, enabling up to 2.4 TOPS with SINT4 weights or ~1.2 TFLOPs in FP8. Samsung positioned the technology as a lower‑cost alternative to HBM, which now accounts for 63 % of AI chip component spending. The design uses a conventional DRAM controller with a mode‑change switch for PIM operation.

💡 Why It Matters

· By integrating PIM into LPDDR5X, Samsung offers a cost‑effective, high‑bandwidth AI accelerator that could reduce reliance on expensive HBM stacks, reshaping the memory landscape for data‑center and edge deployments.

Capsules Semiconductors

Read full article on servethehome.com

Apple Explains M6 and M5 Ultra Chips With 2nm Tech and Huge…

Apple unveiled its M6 and M5 Ultra silicon on Tuesday, debuting them with refreshed Mac mini and Mac Studio models. The M6 is Apple’s first 2‑nanometer chip, featuring a 12‑core CPU (2 super, 4 performance, 6 efficiency) and a 12‑core GPU, each with two additional cores versus the M5. Apple claims single‑threaded performance 1.2× faster and multithreaded gains of 1.2× over the M5, with a dual 16‑core Neural Engine and 170 GB/s memory bandwidth. The M5 Ultra is a quad‑die design that fuses two M5 Max dies, delivering up to a 36‑core CPU, 80‑core GPU, and 512 GB unified memory at 1.2 TB/s. Both chips launch September 22, with a 512‑GB Mac Studio variant following in October.

💡 The move to 2‑nm and quad‑die architecture signals Apple’s push to keep pace with silicon leaders, offering unprecedented AI compute and memory bandwidth for demanding professional workloads.

iphoneincanada.ca · Aug 26

OpenAI Releases First Benchmark Results for its Jalapeño AI Chip

OpenAI unveiled benchmark results for Jalapeño, its first custom AI inference chip, showing substantial gains in speed and energy efficiency. Tests using the SemiAnalysis InferenceX suite compared the processor against leading commercial systems on open‑weight models such as GPT‑OSS 120B, DeepSeek R1 670B and Kimi K2.5 1T. Jalapeño delivered 1.7‑3.6× lower latency, 1.5‑1.9× more work per watt, and 2.1‑4.1× faster performance on real‑time tasks. On the trillion‑parameter Kimi model it responded 3.4× faster while drawing under 550 watts, below its 700‑watt design limit. OpenAI accelerated development by employing its own AI tools, shrinking design time to nine months. The company plans to integrate Jalapeño into production data centers by year‑end, supplementing existing Nvidia hardware for handling the growing demand of users.

💡 Lower inference costs let OpenAI expand user capacity while reducing dependence on external chip makers, securing a hardware edge in the AI services market.

semiconductor-digest.com · Aug 26

Celera Semiconductor Announces the Closing of a $30M Series B Financing Round

Celera Semiconductor, an analog integrated‑circuit maker that leverages artificial intelligence to reduce design cost and time, announced the closing of a $30 million Series B financing round fully funded by Maverick Silicon, its largest backer. The capital will support the integration of Silicon Gate, a Portuguese product‑development and design‑automation team recently acquired by Celera, and expand its California‑based design and AI capabilities for power‑management applications. CEO Patrick Brockett highlighted the company’s rapid progress over the past two years, while CTO Dr. Helen Song said the new funds will accelerate Celera’s design‑automation roadmap. Maverick Silicon’s managing director, Kenneth Safar, described the financing as a catalyst for removing schedule bottlenecks that customers face when sourcing custom analog components.

💡 The infusion enables Celera to scale AI‑driven automation, potentially reshaping how engineers source and develop custom analog chips, a traditionally time‑intensive segment.

taipeitimes.com · Aug 26

Manufacturing confidence rises: TIER

Taiwan’s manufacturing confidence surged to a four-year high last month, reaching 101.09, driven by robust global demand for artificial intelligence chips and related hardware. The Taiwan Institute of Economic Research reported that semiconductor shipments, exports, and overseas orders all improved, boosting optimism among electronics and machinery firms. High capacity utilization was noted in advanced-process fabs and packaging facilities. Conversely, the services sector weakened, with its climate index falling to 98.2 due to financial market volatility affecting securities firms, though banks remained steady. The construction sector also strengthened, rising to 110.55 for the fourth consecutive month, supported by infrastructure needs from AI and ICT companies. Overall, the manufacturing rebound highlights the significant economic momentum generated by the AI boom, contrasting sharply with the softer sentiment in service industries amid broader market fluctuations.

💡 The divergence between manufacturing strength and service sector weakness reveals an economy increasingly dependent on specific high-tech export cycles rather than broad-based domestic growth. This structural imbalance suggests that Taiwan's economic resilience is tightly coupled with global AI infrastructure spending, leaving it vulnerable to shifts in international tech demand.

taipeitimes.com · Aug 26

EDITORIAL: A window of opportunity

SEMI Taiwan launched the SEMI Taiwan Materials Alliance last Friday to bolster local supply chain resilience amid geopolitical risks and AI-driven demand. Taiwan, the world’s largest semiconductor materials market for 16 consecutive years, recorded US$21.7 billion in sales last year. Despite this dominance, local firms rely heavily on imported raw materials like specialty chemicals and photoresists. The alliance, comprising over 30 domestic and foreign companies including ASE Technology and GlobalWafers, aims to help smaller local suppliers overcome scale disadvantages and limited patents. Members like LCY Chemical and Sunlit leverage geographic proximity and operational agility to serve clients like TSMC more quickly than foreign rivals. While global giants like Japan’s JSR and Germany’s Merck remain dominant, the alliance seeks strategic collaborations to enhance technical capabilities. Experts suggest local firms should initially focus on supplementary roles or consumables rather than displacing established first-tier suppliers, capitalizing on the current AI boom to secure a foothold in the critical materials market.

💡 Local agility offers a distinct competitive edge against entrenched global giants in the high-stakes race for AI chip production. By formalizing these partnerships, Taiwan transforms its geographic advantage into a strategic buffer against supply chain disruptions.

semiconductor-digest.com · Aug 25

Qnity Expands AI-Driven Advanced Packaging Capabilities with End-to-End Process Solutions

Qnity Electronics unveiled a scalable materials solution platform that integrates metallization, bumping, dielectric, and fine‑line patterning technologies for AI‑driven advanced packaging. The platform supports high‑density, 3D integrated systems, offering precise plating height control, superior coplanarity, and low defectivity for 2 µm line/space metallization. It includes SnAg micro‑bump plating, Cu‑Cu bonding, and high‑resolution dielectric processes, enabling Chip‑on‑Wafer‑on‑Substrate, EMIB, and HBM applications. Qnity will present the platform at SEMICON Taiwan 2026, inviting industry partners to discuss future packaging trends.

💡 By unifying critical process steps, Qnity reduces integration complexity for vertical architectures, accelerating the adoption of high‑bandwidth memory and chiplet designs. The platform’s precise control over critical dimensions directly addresses yield challenges that have stalled scaling in next‑generation semiconductor systems.