1 week out from our @EYnews x 8090 Modernization webinar. It will cover: ---> Discovery and Portfolio Analysis ---> Large-scale delivery ---> Practical methods for reverse engineering legacy ---> Where AI agents fit ---> How people approve/audit every step One hour with live Q&A. Reserve your spot:

Bottlenecks keep constraining the AI buildout According to ETNews, some Samsung Electro-Mechanics high-capacitance MLCCs are seeing average lead times of around 40 weeks, while Murata products of 1 microfarad or more reached 30 weeks in July, up from 24 weeks in June, with some distribution channels quoting up to 36 weeks AI servers can require five to 13 times as many high-capacitance, compact, and high-reliability MLCCs as conventional servers. Some capacity expansions previously expected in Q4 2026 have also been pushed into 2027, limiting near-term supply relief The extended lead times are not representative of the broader MLCC market Avnet Abacus' August 2026 guide puts multilayer ceramic capacitors and high-CV multilayer types at 8 to more than 20 weeks, with both categories marked stable. Automotive ceramic capacitors are in a similar range The contrast points to a two-speed market. Mainstream ceramic capacitors remain relatively stable, while specialized products tied to AI server power architectures are facing much longer waits This suggests AI infrastructure bottlenecks are spreading beyond processors and memory into less visible parts of the server supply chain. Longer lead times and growing use of long-term supply agreements also show customers are trying to secure specialized passive components before additional capacity comes online

Samsung is reportedly developing next-gen HBM packaging for mobile on-device AI. According to ETNews, Samsung is working on “Multi Stacked FOWLP,” combining advanced copper-pillar stacking with fan-out wafer-level packaging. Today’s mobile LPDDR still uses copper wire bonding, which limits I/O to roughly 128–256 terminals and creates signal-loss, thermal, and power-efficiency bottlenecks. Samsung’s VCS technology improves this by stacking DRAM dies in a staircase structure and connecting them with copper pillars. The new approach appears to push that further, aiming to bring HBM-like bandwidth closer to mobile devices.

Cyberport, in collaboration with @AITinkerers , Prompt in Progress and @EYnews, will co-organise the event, bringing you insights into enterprise-grade AI innovative practices and future development directions. 🔗 Registration link:

Murata is adding MLCC capacity for AI racks and killing the cheap codes that used to fill the same lines. Etnews says it will spend about 80 billion yen, roughly $540 million, over two years through 2027 and lift total output more than 20%. Land for sites through 2028 is already booked. Vice president Masanori Minamide said it is now looking at new plants in Japan and overseas for demand after 2028, on a three-to-five-year view. It has already told customers that parts of nine MLCC series will go end of life. Consumer, industrial and some auto grades are in that list. Last orders are March 2028. Last shipments are March 2029. Samsung Electro-Mechanics is on the same AI-server book. Long-term deals signed since May on silicon capacitors and AI-server MLCCs total 3.32 trillion won. It claims more than 40% share in AI-server MLCCs. An AI rack can take up to 600,000 capacitors, more than ten times a conventional server. Parts have to live at high temperature and voltage around the clock. Few makers can run that spec, which is why the mix pays. Goldman sees the global AI-server MLCC market rising from $1.4 billion last year to $5.8 billion in 2030, about 34% a year. Commodity MLCCs leave the Japanese line. High-cap, high-reliability parts stay. If Murata drops nine series and still adds 20% capacity, does the extra output refill phones or only the 600,000-piece AI rack?

So just tracking the MLCC cycle: ComponentNews classified server MLCCs as “severe” shortage stage, with ETNews reporting: - Samsung has reached ~40 week lead times for some high capacitance MLCCs per DigiKey’s shipment data. Earlier this year, broader reports were ~20 weeks, so the AI server MLCC bottleneck keeps growing. - Murata was ~24 weeks in June. July was ~30 weeks. And now some at ~36 weeks. This was interesting: “The expansion of new production capacity is being postponed from Q4 2026 to 2027” Doesn’t say which expansion… maybe Murata? But if capacity expansion reportedly gets stalled, the bottleneck should tighten short term. Lead times are a good way to track demand imbalances.