Global Semiconductor Alliance

"This genie is out of the bottle. AI is out, and nobody can put it back." That's Jitendra Mohan, CEO of Astera Labs, answering a question his own wife had asked him just the week before our conversation: are you okay with all these data centers being built, the cooling, the energy? His answer wasn't dismissive. It was historical. Every revolution, industrial, PC, internet, has eventually made life better on average, even when the disruption along the way was real. Horse carriages were replaced by trains. Jobs changed. People adapted. He doesn't think AI will be any different. He was careful to add the part some skip, that the technology unlocking the potential is only half the story... What a society does with the gains, how it distributes them, is what actually defines the outcome. I talked to all three Astera Labs founders, Jitendra M., Sanjay Gajendra, and Casey Morrison, about building the infrastructure underneath this moment and what they think it actually owes the rest of us. Full episode live now: YouTube: https://lnkd.in/drA4fCrD   Spotify: https://lnkd.in/dnyr_ffy Apple Podcasts: https://lnkd.in/d8dBGwFF Our new newsletter has all of my takeaways and more! Check it out here: https://lnkd.in/dnjbkQxa #ABitPersonal #AsteraLabs #ArtificialIntelligence #Founders #Builders #AI

𝗗𝗿𝘂𝗺𝗿𝗼𝗹𝗹 𝗽𝗹𝗲𝗮𝘀𝗲 𝗳𝗼𝗿 𝗼𝗻𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗯𝗶𝗴𝗴𝗲𝘀𝘁 𝗺𝗼𝗺𝗲𝗻𝘁𝘀 𝗼𝗳 𝘁𝗵𝗲 𝘆𝗲𝗮𝗿….. We are honored to announce this year’s WISH 2026 Dr. Lisa Su Woman of Innovation Award recipient, Debora Shoquist, Executive Vice President of Operations of NVIDIA! She is a true trailblazer and industry leader, breaking barriers and paving the way for others to follow. Through her vision, expertise, and determination, she continues to inspire meaningful change and set a powerful example for the future of semiconductor. Debora is responsible for NVIDIA’s operations and supply chain functions. Debora joined NVIDIA in 2007 as senior vice president of operations and was named to her current role of executive vice president of operations two years later. Previously, Debora held executive roles at JDS Uniphase, Coherent. and Quantum after 10 years at HP. Debora holds degrees in electrical engineering and biology from Kansas State University and Santa Clara University, respectively. The 𝗗𝗿. 𝗟𝗶𝘀𝗮 𝗦𝘂 𝗪𝗼𝗺𝗮𝗻 𝗼𝗳 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗔𝘄𝗮𝗿𝗱 is one of WLI’s most distinguished honors, recognizing an exceptional career defined by visionary leadership, groundbreaking innovation, and lasting impact on the semiconductor industry. Congratulations on this outstanding achievement! We look forward to celebrating Debora and our other award nominees at WISH 2026 on November 10th in Santa Clara, California. Join us at WISH! Register at the link:  https://hubs.la/Q04y3V3N0 #WLI #WISH26 #Leadership #WomenInSTEM

Last week I had the pleasure of sitting with Anirudh Devgan, President & CEO of Cadence, for a fireside chat at the Global Semiconductor Alliance's US Executive Forum where we discussed the interplay between AI systems and the tools used to design them. It was a great and wide-ranging conversation, but in short as the AI buildout unfolds, the needs of AI are making the systems we design far more complex, while at the same time giving engineers far more powerful tools to design them. Which curve moves faster over the next five years will have immeasurable impact on how the industry evolves. Thanks to the Jodi Shelton and the GSA team for hosting, and to Anirudh for a great conversation.

On Tuesday at the Global Semiconductor Alliance US Executive Forum in Menlo Park, I moderated "Silicon at the Breaking Point: Designing Chips for a World That Can't Wait" with Samsung Semiconductor President Paul Cho, AMD CTO Mark Papermaster, Broadcom Semiconductor Solutions Group President Charlie Kawwas and OpenAI VP of Hardware Richard Ho. Thanks to all four for the candor, and to the GSA team led by Jodi Shelton and Cherie Britt at Samsung for pulling it together. Several points are worth putting on the record. The lead-in was that the transistor is no longer the bottleneck. Power, memory bandwidth and packaging are. I opened with that, and not one of the four pushed back. Each built on it from his first answer. On custom silicon: Richard framed OpenAI's XPU as a full-stack bet on intelligence per watt for a known set of workloads, not a bet against merchant GPUs. The unit of purchase is now the fleet across multiple campuses, not the chip or the rack. I said years ago that 75+% design margins would push the biggest buyers to find another path. They found it. Charlie's outlook is single campuses of 5 to 10 gigawatts by 2031. On power: Mark's line was that power is performance, and data movement is where most of it goes. Chiplets were a power decision at AMD before anything else. Thermal and cooling are now job one. On AI in the design loop: Richard said AI compressed OpenAI's design cycle more than his team expected, and the point is to make engineers faster, not to replace them. Mark's rule at AMD is that agentic AI runs the design space exploration no human could do by hand, and accountability stays with the human engineer. I agree with both. On packaging: Charlie called the substrate the choke point for the next five years, which is why Broadcom is going into substrate production itself. On the OpenAI program, Broadcom's tape-out went from 15 to 18 months to nine. His math: that cadence is an extra generation every three years, and in AI that is the whole game. On memory: Paul had the stat of the night. By his numbers, compute is up roughly 60,000x in 20 years and memory bandwidth roughly 100x. Design memory first and bring the memory partner into architecture, not procurement. He also said supply chain resilience is now a design input decided at architecture definition, and that Samsung's Taylor, Texas fab goes to customer mass production in 2027. The node still delivers, at a price. The advantage now goes to whoever co-designs compute, memory, packaging, networking and power from day one. That is structural, not a phase. #Semiconductors #AI #ChipDesign #AdvancedPackaging #HBM

Great discussions at the Global Semiconductor Alliance U.S. Executive Forum this week on how quickly agentic AI is changing our industry. In my keynote, I shared what we are seeing firsthand at AMD. Agentic workflows are already driving significant productivity gains in chip design, while also changing the infrastructure underneath AI with greater demand for CPUs alongside GPUs, networking and memory and a token router across Frontier models, on premise, and personal AI.   At the panel, we discussed the increased need for heterogeneous and modular compute, an open ecosystem, a relentless focus on power efficiency, and holistic optimization across hardware and software. AI is changing how we build that technology, giving engineers the ability to explore and iterate at a scale that simply wasn’t possible before. The new normal will be an AI first approach.   Thanks to Jodi Shelton and the GSA team for bringing the industry together, to Patrick Moorhead for organizing a great panel discussion, and to Charlie Kawwas, Paul Cho and Richard Ho for the thoughtful exchange.

Another tremendously insightful Global Semiconductor Alliance Executive Forum took place this week in the heart of #SiliconValley. Hosted by the iconic and inimitable Jodi Shelton, it was packed with industry decision-makers who shared perspectives on what comes next, while underscoring the need for collective action across the entire supply chain. The theme was the primacy of #semiconductors, but the conversations, both on and off stage, all seemed to converge on one thing: advanced packaging is becoming a critical enabler as the semiconductor market moves beyond the trillion-dollar threshold. Recognizing the significant opportunity ahead, AMD’s Mark Papermaster urged the audience to “design holistically, move faster, and stay open.” Broadcom’s Charlie Kawwas reinforced just how much that will require beyond continued innovation at the chip level. The next wave of system-level performance will depend on collaboration across materials and #substrates, equipment, manufacturing, #advancedpackaging, and the broader semiconductor ecosystem—turning advances at the silicon level into performance at AI and data-center scale. #VIPack #AI #datacenter #HPC #semiconductors #siliconphotonics #CPO