Why Silicon Photonics is Critical to Scaling AI Efficiently

semiconductor-digest.com Jun 19, 2026

Why Silicon Photonics is Critical to Scaling AI Efficiently

Curated by Shubham · TechCapsules · Jun 19, 2026

AI-summarised brief · reviewed before publication

Silicon photonics is gaining traction as a solution to the "copper wall" issue in data centers, where electrical interconnects are struggling to meet power efficiency and latency thresholds. As AI data centers scale, east-west traffic increases, placing strain on copper interconnects. Silicon photonics offers a more efficient solution, with co-packaged optics bringing optical and electrical components closer together. The silicon photonics market is expected to grow 25-30% annually, driven by AI and cloud infrastructure demand. This growth is projected to reach $10 billion by the end of the decade, with high-quality substrates key to scaling silicon photonics efficiently.

💡 Why It Matters

· Silicon photonics directly addresses the pressing need for faster data movement in AI and cloud infrastructure.
· By integrating optical capabilities close to compute, silicon photonics enables data centers to overcome the limitations of traditional copper interconnects.

Capsules Semiconductors

Read full article on semiconductor-digest.com

AI’s Next Power Challenge Is Inside the Delivery Path

The semiconductor industry faces a critical challenge in power delivery to AI systems, with traditional lateral power architectures resulting in 15-20% power loss due to resistive losses. As device power levels climb and operating voltages drop, the industry is shifting towards vertical power delivery schemes to reduce losses and improve efficiency. Technologies like Saras Micro Devices' STILE platform embed passive functionality into system PCB and package structures, supporting voltage regulation demands and freeing board-level real estate. This shift is redefining packaging discussions, with packaging becoming part of the power architecture. The industry must evaluate energy efficiency through the lens of power delivery path efficiency, not just processor performance and thermal management, to optimize compute capabilities.

💡 Every watt lost in the power delivery path is a watt unavailable for compute, directly impacting AI system performance. Vertical power delivery schemes can significantly reduce energy waste and infrastructure strain.

memeburn.com · Jun 19

Kirin 9030 Pro Teardown Shows Huawei’s SMIC Chip Beating Intel 18A

A teardown of the Kirin 9030 Pro has revealed that SMIC's N+3 manufacturing process achieves a tighter local metal pitch than Intel's 18A node, with a 32.5 nm metal pitch compared to Intel's 36 nm. However, experts caution that this comparison is incomplete, as Intel's 18A node has a higher transistor density, approximately 38% higher than SMIC's N+3 process. The Kirin 9030 Pro's performance also trails the Snapdragon 8 Elite by a wide margin. Huawei's upcoming LogicFolding design may help close the gap by stacking logic circuits vertically. The finding has reignited the global semiconductor debate, with SMIC's achievement being seen as a significant milestone despite the limitations.

💡 SMIC's ability to achieve a tighter metal pitch than Intel using older equipment underscores the potential for innovation despite US sanctions. Huawei's LogicFolding design could further narrow the gap with industry leaders.

moneycontrol.com · Jun 19

US tells ASML it’s concerned China may have top chip tool

The US Commerce Secretary has expressed concerns to ASML Holding NV that one of its top-of-the-line machines may have been shipped to China, violating US-led export restrictions. ASML has denied the claims, stating that none of its extreme ultraviolet lithography machines are in China. The company has never been allowed to ship these machines to China due to curbs imposed during the Trump administration. The US has evidence indicating ASML is not acting in good faith, but has not provided proof of EUV machine shipments to China. ASML is trying to prove a negative and alleviate US concerns.

💡 The situation strains US-EU relations and pressures ASML to comply with US export regulations. EUV machines are crucial for manufacturing processors for major tech companies.

semiconductor-digest.com · Jun 19

Micron Selects Bechtel as Construction Partner for Historic New York Semiconductor Project

Micron Technology, Inc. has selected Bechtel as its construction partner for the engineering, procurement, and construction (EPC) of the first phase of its leading-edge memory manufacturing complex in Clay, New York. The project, the largest private investment in New York state history, is expected to generate 50,000 jobs, including over 4,500 construction jobs, and add approximately $16.7 billion per year in real economic output in the state. Bechtel will mobilize at the site immediately and deploy an integrated EPC delivery model to support schedule confidence and operational readiness.

💡 The partnership between Micron and Bechtel represents a significant step towards strengthening the resilience of America's technology supply chain and expanding U.S. semiconductor manufacturing capacity. By investing in cutting-edge infrastructure, the project aims to secure America's position in the AI era and create a foundation for long-term technology leadership and economic growth.

semiconductor-digest.com · Jun 19

AI Investment Expected to Drive Semiconductor Industry Past $2 Trillion by 2027

The semiconductor industry is projected to surpass $2 trillion in revenue by 2027, driven by artificial intelligence investment. According to TechInsights, the industry recently crossed the $1 trillion threshold, making this a rapid expansion. AI is creating demand for sophisticated semiconductor technologies, with the four largest US hyperscalers expected to spend $718 billion in capital expenditures in 2026. This growth is fueled by investment in AI infrastructure, with datacenter processors expected to contribute over $100 billion in market value by 2027. The industry's supply chains are also being reshaped, with competition for leading-edge technologies intensifying.

💡 AI's impact on the semiconductor industry is redefining the economics of the sector, with performance and system-level integration becoming key differentiators. Investment in AI infrastructure is driving demand for accelerated computing, which will remain a long-term driver of semiconductor growth.

appleinsider.com · Jun 18

Trump confirms that Apple will buy chips made by Intel

Trump announced that Apple will work with Intel to design and build chips in America, prompting an 8.8% rise in Intel's stock and a 0.6% increase in Apple's. The deal, a substantial multi-year agreement, is believed to concern production of older or lower-end processors, such as M-series chips for iPad Pro and MacBook Air. The partnership follows initial rumors and test production in May 2026. Details on delivery, shipment, and deal worth remain unclear. Intel's entry into limited scale testing of the 18A-P process means full production for Apple won't start until mid-2027. Apple's deal with Intel reduces dependence on Taiwan chip production.

💡 Apple's shift to Intel for older processors frees up TSMC's capacity for newer, high-end chips, ensuring a steady supply for flagship devices. This strategic move helps Apple mitigate the global chip shortage's impact on its product lineup.