CONNECT WITH US
Aug 12
OCP photonics panel warns CPO packaging is still unsettled as pluggables fade

Pluggable front-panel optics still own the market, but Arista co-founder Andy Bechtolsheim gave them a deadline at an OCP APAC panel here: the form factor reaches its physical limit at the 49.6T switch generation, roughly 18 months out.

Speaking at the OCP APAC Summit 2026, Ayar Labs CEO Mark Wade argued that copper interconnects have become one of the biggest obstacles to realizing the vision of open computing.
As AI workloads push data center interconnects from 800G toward 1.6T, indium phosphide (InP) has become a critical material for high-speed optical links, particularly for longer-distance transmission. But crystal growth challenges, supply constraints and high costs are prompting the industry to explore gallium arsenide (GaAs) as a partial alternative for shorter-reach connections, with the technology expected to become clearer by 2028.

Intel is signaling a broader push into next-generation memory technologies, with CEO Lip-Bu Tan revealing that the company is exploring new memory architectures after hiring former SK Hynix CEO Seok-Hee Lee.

South Korea's Hanmi Semiconductor is pulling ahead of rival Hanwha Semitech in thermal compression (TC) bonders for high-bandwidth memory (HBM), as HBM4 equipment orders rebound and Hanmi's growing business with Micron helps reduce its reliance on SK Hynix, according to Chosun Biz.

Odisha's state cabinet has approved a third amendment to its Semiconductor Manufacturing and Fabless Policy, adding fresh capital support to attract projects backed by India's national chip-making push, according to state officials cited by regional outlets on August 12, 2026.

Bijan Nowroozi, Head of Ecosystem Development at Lightmatter, announced at the OCP APAC Summit 2026 that the company has released an infrastructure white paper aimed at building a more unified, less fragmented supply chain for photonic interconnects.
Siemens EDA has introduced an AI-Native Design strategy in Seoul, aiming to help chipmakers and system designers manage rising complexity while reducing the risk of AI errors. The move could shape how semiconductor tools evolve worldwide, as companies seek faster, more reliable automation across increasingly global supply chains.
Nvidia, AMD and Broadcom are moving beyond chip-level design into subsystems and complete systems, a shift that is increasing the need to account for thermal, electrical, mechanical and other physical effects earlier in the design process, according to Synopsys executives.

NXP Semiconductors has begun expanding its Petaling Jaya assembly and test site in Malaysia, a move that could support semiconductor supply chains worldwide by adding capacity, automation, and geographic diversification. The project is part of a broader effort to improve manufacturing resilience as global demand and supply risks continue to shape the industry.

Nvidia's plan to assemble more than US$500 billion in third-party capital for AI infrastructure could reshape how the industry pays for compute. By bringing major Wall Street firms into the deal, the chipmaker is effectively turning AI capacity into a financeable asset class, with implications for cloud providers, investors, and credit markets.
Intel sees agentic AI changing the balance of AI servers, as CPU orchestration, memory capacity and data movement emerge as performance constraints even when more GPU capacity is available.