Marvell's Andrew Yick in PIC Magazine Cover Story

🚀 Cover story from #PhotonicIntegratedCircuitsMagazine - Issue 2! 'Why scale-up AI networks demand scalable optical test' by Andrew Yick, Marvell Technology Scale-up AI networks are pushing optics directly onto switches and AI accelerators. To avoid a manufacturing bottleneck, the industry is shifting optical test from lowparallelism, custom configurations to scalable, automated, IC-style methods. Read below or click the link 🔗 https://lnkd.in/g2Tssre #PICMagazine #PIC #photonicintegratedcircuits

Perfect article from Andrew Yick! This is why it’s very exciting working in the photonics industry at this stage from the tester side as well as from DUT side. For example, a uniformed superset configuration of standardized optical instrumentation across the photonics industry will help vendors create native ATE instruments within the test head :)

𝗘𝘅𝗽𝗲𝗿𝘁 𝗽𝗲𝗿𝘀𝗽𝗲𝗰𝘁𝗶𝘃𝗲 🔹Addressing AI and Advanced Packaging Challenges with Synopsys 3DIO PHY By Lakshmi Jain, Wei-Yu Ma, Synopsys Inc https://lnkd.in/eYrVE5TU Learn how Synopsys 3DIO PHY delivers predictable timing, low power, and scalable die to die connectivity for AI and hybrid bonding. Read more at https://lnkd.in/eYrVE5TU #chiplet #3DIC #AdvancedPackaging #MultiDie #semiconductor

Today we introduce to you to Kinex, the hybrid bonding foreman within Applied Materials semiconductor chip city. 🧱 Stacking chiplets with atomic precision and building the backbone of tomorrow’s AI, Kinex is Applied’s fully integrated hybrid bonding system, purpose built for high volume manufacturing environments. Learn more about Kinex: https://bit.ly/46CDkP0 We deliver material innovation that changes the world. #MakeItMaterial

The Celestica 1.6T series is now ready to order. Powered by Broadcom Tomahawk 6 silicon, the DS6000 (air cooled) and DS6001 (hybrid cooled) switches are built to break AI/ML bottlenecks and scale the world’s most massive AI clusters. Get the full technical breakdown in our latest blog, found in comments. 👇 #16TNetworking #AI #OCPEMEA #Celestica

Ultimate Guide to AI Chip Processors "From Moore’s groundbreaking prediction of chip components doubling every 24 months to AI revolutionizing industries from healthcare to space tech is evolving faster than ever. AI chips are the powerhouses driving this change, fueling the future of machine learning and next-level data efficiency". The future is now! Please click on the link in the comments to read the full blog post Please Like and Comment if you find this post useful #MooresLaw #TechInnovation #FutureTech #ChipTechnology #GordonMoore

AI doesn’t exist in isolation. It depends on advanced chips, and those chips depend on lithography. This Bloomberg video with our CEO Christophe Fouquet explains that connection clearly. Proud to work on technology that sits so close to the foundation of AI progress! Check it out: http://ms.spr.ly/6043vJGSB

Meet the Partners... Get to know Cognichip, an official AI Infra Summit 2026 Platinum Sponsor. Cognichip is applying AI directly to semiconductor design through its Artificial Chip Intelligence platform, built around a physics-informed foundation model tailored for chip design. As demand for more capable infrastructure grows, that kind of design acceleration could play a major role in how the next generation of AI hardware gets built. Explore partnership opportunities with AI Infra Summit: https://lnkd.in/eeUvE7Tz #AIInfraSummit #AIInfrastructure #AI #Innovation #Cognichip

Can AI agents reshape global power? This week: the race for AI dominance and how optical computing could slash energy costs for massive models. Watch the breakthroughs. #DeepTech #AIpolicy https://lnkd.in/eb6TnuW4

2. In-house mixed-precision NPU: Axera develops the mixed precision NPU (Axera Neutron) in-house for smart vehicles and edge AI applications, with its AI algorithms and underlying hardware. To deploy the AI models in edge or on-device, management notes the company breaks through the data and memory bottlenecks, adopts more concise data representations and lower the redundancy of weights/ feature values to improve the computational efficiency. We see the company developing solutions with enhanced processing speed, and clients could select different numerical precision based on their demand.