KOSUNI CRYPTO
Korea's Crypto Pulse, in English
$BTC $ETH $XRP $SOL $DOGE
Market Analysis
Neutral

AI Chip Demand Surge... Development Time and Yield Improvement Key

Published October 19, 2026 12:39 PM · 5 views
AI Chip Demand Surge... Development Time and Yield Improvement Key

As AI technology advances faster than semiconductor production capabilities, reducing chip R&D timelines and improving yield rates have become critical challenges, according to industry analyst Stephen Sorko. The competition for integrated data center design is also intensifying.

In a Schwab Network interview on October 17 (local time), Sorko noted that while AI develops annually, the semiconductor industry struggles to scale technology and production capacity accordingly. Applied Materials aims to reduce semiconductor material and equipment R&D periods from current 10-15 years to 6-8 years.

"AI is now advancing on an annual basis, and it doesn't require a long time for the semiconductor industry to catch up," Sorko stated. Limited production capacity was also cited as a major constraint. Sorko referenced ASML's recent earnings report highlighting production capacity limitations, predicting that Applied Materials, ASML, and Lam Research will struggle to keep up with demand.

Improving yield rates is another solution for expanding supply. Producing modern semiconductors requires 500-1000 manufacturing steps, and even when processes run smoothly, some chips may be discarded. Sorko emphasized that reducing discarded chips would allow more AI chips to be secured.

Brocade was noted for simultaneously pursuing multiple initiatives including customized semiconductors for large cloud providers and optical communication technologies to improve data center system efficiency. While the company's broad scope makes it difficult to determine which segment drives overall performance, Sorko viewed its outlook as highly positive. AMD's newly announced server rack was described as an attempt to more efficiently configure data center systems.

Early AI training environments saw one CPU managing 8-24 GPUs, but in autonomous AI, CPUs are taking on leadership roles directing other semiconductors, gradually equalizing the importance of CPUs and GPUs. This shift has led the semiconductor industry to return CPUs to central positions in data center design while building integrated systems that encompass networking and data flow.

"Now what's important is not just semiconductors but ensuring all elements work smoothly within an end-to-end system," Sorko said. Intel was noted for improved execution over the past 18 months, with strong capabilities in packaging complex semiconductors domestically in the U.S. A card-based product designed to enhance AI computing power when installed in existing air-cooled data center servers supports companies in utilizing AI within their current facilities without building new data centers.

Sorko stressed that both large-scale new data center construction and optimizing existing infrastructure for AI demand are important, with leveraging established infrastructure being key to reducing costs and power grid burdens for the semiconductor industry.

Korean Source

This article is an English localization of a Korean-language crypto news report. Original headline: AI 반도체 수요 폭증…칩 개발 기간 단축·수율 개선이 관건