Trends in the AI Semiconductor Market

Global semiconductor revenue jumped 21% in 2025. But this wasn’t broad-based growth it was AI concentration. Worldwide chip revenue hit $793B, yet nearly one-third came from AI-related semiconductors. That’s not a cycle. That’s a structural shift. The real signal: NVIDIA alone drove over 35% of total industry growth, crossing $100B in semiconductor revenue for the first time. This changes competitive dynamics, supply chains, and geopolitics. Breakdown: • AI processors > $200B → Compute is now the primary value capture layer • HBM = 23% of DRAM ($30B+) → Memory power shifted to SK Hynix • AI infra spend = $1.3T by 2026 → Chips are national assets, not commodities Meanwhile, legacy models are breaking. Intel Corporation’s share fell to 6% half of 2021 levels. Samsung Electronics grew memory, but non-memory revenue declined 8%. By 2029, AI semiconductors are projected to exceed 50% of total chip sales. The question isn’t growth anymore. It’s who controls the AI supply chain choke points. Is the industry becoming more innovative or more concentrated? #Semiconductors #AIChips #SupplyChain #HBM #Geopolitics #ChipIndustry #DataCenters #ManufacturingTech

AI and Memory Are Redefining Semiconductor Leadership — But Who Loses Capacity? The semiconductor sector’s year-to-date performance makes one point unmistakably clear: exposure to AI and memory has become the primary determinant of equity returns. Investors are no longer pricing the industry as a single cycle, but as a set of structurally different businesses tied to very different end markets. Memory and equipment names sit at the top of the performance table. Micron is up +158%, Lam Research +114%, and KLA +84%, reflecting the rebound in memory pricing, accelerating adoption of high-bandwidth memory, and renewed investment in wafer fabrication. Intel (+78%) and AMD (+64%) followed as expectations improved around AI roadmaps, foundry optionality, and operating leverage. The message from the market is consistent: companies directly tied to AI infrastructure and capital intensity are being rewarded first. Large-cap enablers of advanced manufacturing delivered strong but more measured gains. Applied Materials (+52%), ASML (+45%), Broadcom (+41%), and TSMC (+37%) continue to benefit from sustained spending on leading-edge nodes and packaging. Nvidia (+24%) and Analog Devices (+28%) remain positive contributors, though performance has moderated as valuation already reflects a substantial portion of expected growth. At the other end of the spectrum, the gap is widening. Stocks with heavier exposure to handsets, analog cyclicality, or uncertain near-term growth have lagged. Marvell (-28%) and Skyworks (-27%) saw sharp declines, while ARM (-11%), Texas Instruments (-7%), and Synopsys (-6%) posted modestly negative returns. The market appears less patient with businesses where AI monetization is indirect or longer-dated. One central question increasingly shapes the outlook: will AI continue to absorb a disproportionate share of advanced semiconductor capacity? Data centers now command priority access to leading-edge logic, memory, and advanced packaging. Automotive and industrial customers are facing tighter allocations, longer lead times, and rising costs—not because demand is weak, but because capacity is being redirected toward higher-margin AI workloads. Geopolitics amplifies this imbalance. More than 75% of global chip production remains concentrated in East Asia, leaving the supply chain exposed to export controls, trade restrictions, and Taiwan-related risks. At the same time, bottlenecks have shifted from wafer fabrication to advanced packaging and assembly, now one of the most binding constraints. Energy availability is emerging as a new limiting factor, as AI data centers place growing strain on regional power grids. The semiconductor industry is no longer moving in unison. Equity performance increasingly reflects positioning within the AI value chain, access to constrained capacity, and resilience to structural risks. The divergence visible today may be less cyclical than many assume. #Semiconductors #AI #NVDA #MU #TSM

🇯🇵 From Soft Power to Silicon Power: Insights from Day 1 at SEMICON Japan 2025 Japan is no longer just the "material and equipment" backbone of the world; it is re-emerging as the epicenter of the AI-Semiconductor Reinvention. Walking the floor at Tokyo Big Sight today for the opening of SEMICON Japan 2025, the energy is different. It’s not just about manufacturing, it’s about the convergence of AI, Power Electronics, and Global Capability Centers (GCCs). Here are the 3 Strategic Takeaways from Day 1: 1. The 'Chokepoint' Advantage becomes the 'Growth' Engine Japan controls over 30% of the world’s semiconductor equipment and 50% of its materials. Today’s keynotes highlighted a massive shift: Japan is leveraging this "chokepoint" dominance to lead in Advanced Packaging and 2.5D/3D ICs. As Moore’s Law slows, the innovation is moving to the "back-end"—and Japan owns the back-end. 2. The Green AI Mandate: Power & Compound Semis The "Power & Compound" Pavilion was the busiest spot today. With Generative AI consuming unprecedented levels of energy, the focus has shifted to Silicon Carbide (SiC) and Gallium Nitride (GaN). The data is clear: to scale AI, we must solve the power density problem. Japan’s leadership here is critical for the next wave of sustainable AI data centers. 3. The Rise of 'Chip-to-Cloud' GCCs A fascinating trend discussed in the executive forums is the globalization of R&D. We are seeing a new breed of Semiconductor GCCs, global firms integrating their design teams in India with specialized equipment engineering teams in Japan. This "Indo-Japan Silicon Bridge" is a $50B opportunity for business reinvention. The Tholons Inc. View: 20 years ago, we talked about the "Offshore Nation." Today, we are talking about the "Intelligent Edge." For enterprise leaders, the lesson is simple: Your AI strategy is only as good as your semiconductor supply chain. Are you looking at the hardware layer of your AI transformation? Or are you still just looking at the software? #SemiconJapan #Semiconductors #AI #GCC #DigitalTransformation #Tholons #InnovationAtScale #SupplyChain #JapanTech #SemiconJapan2025 Frank Pendle Abhay Vashistha Ankita Vashistha Sharat Kaul Atul Bhargava Haritez Narisetty

AI holds great potential for the semiconductor industry and will kick-start the next round of innovation for faster, cheaper and more energy-efficient computation – that was my message today at SPIE Advanced Lithography + Patterning. I discussed the potential and the challenges that AI holds for our industry.   The potential is clearly huge. AI is rapidly integrated into applications, and high-performance compute is expected to underpin growth towards $1 trillion of semiconductor sales by 2030. The challenges are around the computing needs of AI models and related energy consumption. The compute workload of training a leading AI model has increased 16x every 2 years in recent years – much faster than the increase in computing power delivered by Moore’s law, which is about 2x every 2 years. The energy needed to train a leading model has not grown so steeply but still rose 10x every 2 years. This computing need has been met by building supercomputers and massive data centers. If you extrapolate these trends, training a leading AI model would need the entire world-wide electricity supply in about 10 years. That’s clearly not realistic, so the trend has to break, by training algorithms becoming more efficient and by chips becoming more efficient. In other words, the needs of AI will stimulate immense innovation in chip design and manufacturing – and the potential value of AI to our society will put urgency and funding behind that drive. As a consequence, chip makers are pulling all levers to accelerate semiconductor scaling. This includes lithographic “2D” scaling: shrinking the dimensions of transistors to pack more into a square millimeter. It will also include “3D” integration, with innovations like backside power delivery, transistor designs like gate-all-around, as well as stacking chips in the package, where holistic lithography will play a critical role to deliver performance requirements. ASML will support these trends through a comprehensive, holistic lithography portfolio. Our 0.33 NA/0.55 NA EUV lithography systems allow chip makers to shrink dimensions at the lowest possible cost on their critical layers, while tightly matched and highly productive DUV systems will continue to reduce cost. More than ever, metrology and inspections tools – whose data is fed into lithography control solutions that keep the patterning process operating within tight specs to deliver the highest possible production yields – will be essential to deliver 2D scaling and 3D integration processes. 3D integration requires wafer-to-wafer bonding, and we have demonstrated the capability to map the stresses and distortions that bonding creates and to compensate for them, reducing overlay errors for post-bonding patterning by 10x or more.   It was a pleasure catching up with the industry’s lithography and patterning experts in San Jose. I’m excited to see our collective innovation power having a go at these challenges. Together, we will push technology forward.

Lisa Su, Chair & CEO at AMD is highlighting a major shift happening in AI infrastructure right now: CPUs are becoming critical again as companies move from simple AI chatbots to full AI-driven workflows and agentic AI systems. Enterprises are rapidly increasing adoption of AI across software development, automation, analytics, and enterprise workflows — and that surge is driving unexpectedly high CPU demand alongside GPUs. A key reason is that modern AI workflows are no longer just “GPU problems.” AI agents now: orchestrate tasks, retrieve data, run simulations, compile code, manage inference pipelines, coordinate multiple models, and handle real-time enterprise operations. Those orchestration and infrastructure layers rely heavily on CPUs. AI server deployments are shifting from traditional CPU-to-GPU ratios like 1:8 toward configurations closer to 1:1. AMD’s data center business reflects that trend: Q1 2026 data center revenue grew 57% year-over-year to $5.8B. AMD forecasts server CPU revenue growth above 70% YoY in Q2. The company now expects the server CPU market to grow at a 35% CAGR through 2030, reaching around $120B. AMD is also positioning itself across the full AI stack: AMD EPYC CPUs for orchestration and inference, AMD Instinct GPUs for training, Ryzen AI for edge and enterprise AI PCs, plus networking and rack-scale AI systems. One of the most important insights from Lisa Su’s comments is that AI adoption is moving from experimentation to operational deployment. Companies are no longer testing AI in isolated pilots — they are embedding AI into real workflows, and that dramatically increases compute demand across CPUs, GPUs, memory, storage, and networking. For the tech industry, this signals a broader transition: AI infrastructure is evolving from “GPU-centric” to “full-stack compute architecture.” via @cnbctv #AI #ArtificialIntelligence #AMD #AIInfrastructure #DataCenter #EPYC #RyzenAI #MachineLearning #GenerativeAI #AgenticAI #EnterpriseAI #CloudComputing #Semiconductors #TechLeadership #DigitalTransformation #FutureOfWork #Innovation

Semiconductor stocks have come under pressure over the past month, raising a broader question: is the market signaling a true AI glut, or simply a repositioning after a strong rally? Our view is that a structural glut remains unlikely. Three forces continue to support AI demand: the capability race among large tech companies seeking to avoid losing strategic ground, the rise of agentic AI and its heavier inference needs, and sovereign AI, as governments fund local infrastructure to strengthen data security and technological independence. At the same time, supply at the leading edge is not easy to ramp. Progress at the most advanced process nodes remains technically challenging, packaging capacity is still complex, and geopolitical fragmentation is reducing the efficiency of global semiconductor manufacturing. In our view, the more credible near-term risk lies in infrastructure constraints rather than a true demand collapse. Delays in power access, shortages of key electrical equipment, and slower permitting processes are limiting the pace of deployment. Overall, we see the recent sell-off as more of a position unwind than the start of a true AI glut. With AI hardware requirements evolving quickly, new supply is unlikely to create broad oversupply in the most advanced parts of the market, keeping us constructive on leading-edge chipmakers and semi-cap equipment names.

AI is booming but the supply chain is cracking. The race to build AI data centers (DCs) is driving unprecedented demand for semiconductors. Hyperscalers like Microsoft, Google, and Alibaba are pouring hundreds of billions into infrastructure, but the ripple effects are hitting closer to home: 🧠 DRAM (Dynamic Random-Access Memory) prices are surging 30% in Q4 2025, with another 20% expected in early 2026. 💾 HDDs (Hard Disk Drives) are maxed out, forcing a pivot to SSDs (Solid-State Drives) — which are also critical for consumer electronics. 🎮 HBM (High-Bandwidth Memory), stacked DRAM used in GPUs, is being prioritized for AI workloads, squeezing supply for other industries. 📱 Nvidia’s shift to LPDDR (Low-Power Double Data Rate) memory, the same chips used in high-end smartphones from Apple and Samsung means AI and consumer electronics are now competing head-to-head for the same components. The result? 📈 Component costs rising 5–10% for PCs and smartphones. 📉 Potential shortages of popular devices. ⏳ Bottlenecks expected to last 2–3 years as fabs (semiconductor fabrication plants) struggle to add capacity. This isn’t just about gadgets. Industries from automotive (EVs) to aerospace & defense rely on the same semiconductor supply chain. When AI eats first, everyone else waits. 👉 The big question: How do we balance the innovation imperative in AI with the stability imperative in consumer tech and beyond? #AI #Semiconductors #SupplyChain #DRAM #HBM #LPDDR #Innovation #Leadership

𝗧𝗵𝗲 𝗦𝗶𝗹𝗶𝗰𝗼𝗻 𝗦𝗵𝗼𝗰𝗸: 𝗪𝗵𝘆 𝘁𝗵𝗲 𝗪𝗼𝗿𝗹𝗱 𝗕𝗲𝗰𝗮𝗺𝗲 𝗢𝗯𝘀𝗲𝘀𝘀𝗲𝗱 𝘄𝗶𝘁𝗵 𝗦𝗲𝗺𝗶𝗰𝗼𝗻𝗱𝘂𝗰𝘁𝗼𝗿𝘀 Five years ago, semiconductors were a quiet, cyclical business which was rarely discussed outside engineering circles. Today, the same industry powers the global economy. Companies like NVIDIA, Broadcom, TSMC, and ASML together are worth over $8.5 trillion. Governments across the US, EU, Japan, and India are pouring hundreds of billions of dollars into new semiconductor strategies. This shift was driven by a mix of AI breakthroughs, geopolitical rivalry, and supply-chain dependence -turning silicon into the world’s most strategic resource. 𝗧𝗵𝗲 𝗙𝗶𝘃𝗲-𝗬𝗲𝗮𝗿 𝗦𝗵𝗶𝗳𝘁 (𝟮𝟬𝟮𝟬 → 𝟮𝟬𝟮𝟱)  • 𝗧𝗵𝗲 𝗔𝗜 𝗖𝗼𝗺𝗽𝘂𝘁𝗲 𝗕𝗼𝗼𝗺: Generative AI created a surge in demand for parallel computing—powered mainly by GPUs. AI compute became the new “digital oil,” pushing chipmakers to the center of the tech economy.  • 𝗦𝘂𝗽𝗽𝗹𝘆 𝗖𝗵𝗮𝗶𝗻 & 𝗚𝗲𝗼𝗽𝗼𝗹𝗶𝘁𝗶𝗰𝘀: The pandemic exposed how concentrated manufacturing is. Nearly all sub-10 nm capacity sits in Taiwan and South Korea, while China dominates packaging and materials. Owning silicon capacity now means owning economic resilience.  • 𝗧𝗵𝗲 𝗦𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻𝘁𝘆 𝗣𝘂𝘀𝗵: Chips are now instruments of national power. Export limits on advanced devices and EUV lithography tools spurred major incentive programs—the US CHIPS Act, EU Chips Act, and others—to rebuild domestic manufacturing.  • 𝗖𝘂𝘀𝘁𝗼𝗺 𝗦𝗶𝗹𝗶𝗰𝗼𝗻: Cloud giants—Google, Amazon, Microsoft, Meta - design their own chips to optimize AI workloads. That shift boosts demand for design IP and specialized foundries such as TSMC. 𝗪𝗵𝗮𝘁’𝘀 𝗡𝗲𝘅𝘁 (𝟮𝟬𝟮𝟲 → 𝟮𝟬𝟯𝟬) The industry is on track to exceed $1 trillion in annual revenue by 2030—driven by AI, electrification, advanced packaging and edge computing.  • 𝗘𝗱𝗴𝗲 𝗔𝗜 𝗘𝘃𝗲𝗿𝘆𝘄𝗵𝗲𝗿𝗲 — Efficient NPUs will move AI from the cloud to devices: wearables, phones, robots, IoTs, drones, etc.  • 𝗘𝗹𝗲𝗰𝘁𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 & 𝗘𝗻𝗲𝗿𝗴𝘆 — EVs and smart grids are accelerating demand for SiC and GaN power chips.  • 𝗣𝗮𝗰𝗸𝗮𝗴𝗶𝗻𝗴 𝗕𝗼𝘁𝘁𝗹𝗲𝗻𝗲𝗰𝗸 — Performance now hinges on chiplets, 3D stacking, #CPOs, and #HBM. Advanced packaging capacity is the new battleground. 𝗕𝗲𝘆𝗼𝗻𝗱 𝟮𝟬𝟯𝟬: 𝗧𝗵𝗲 𝗤𝘂𝗮𝗻𝘁𝘂𝗺 & 𝗣𝗵𝗼𝘁𝗼𝗻𝗶𝗰 𝗛𝗼𝗿𝗶𝘇𝗼𝗻 The next wave is already forming. #Quantum, #photonic, and #neuromorphic chips will bring massive leaps in speed and energy efficiency, reshaping how AI, computing, and communication systems are built. The silicon era was only the start of something much bigger. Murali Chirala Bala Joshi. Tarun Verma. Ish Kumar Bhargava Hrishi Sathawane BV Jagadeesh Krishna Yarlagadda Mahesh Lingareddy

In 2024 I began building a database to track AI partnerships. Here’s what stood out in the past 3 months: ⁣⁣1️⃣ 𝗔𝗜 𝗶𝘀 𝗮𝗻 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 This is the first quarter where compute / AI infra is the dominant partnership category. Over a quarter of deals involve chip providers (NVIDIA, AMD, Cerebras, Intel, Samsung, Broadcom, Qualcomm), compute providers (hyperscalers, OCI, neoclouds), governments building sovereign AI infra (Japan, KSA, Germany), and power providers. 👩🍳 Interesting shift: focus on 𝗰𝘂𝘀𝘁𝗼𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗮𝗻𝗱 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 vs. off-the-shelf compute (e.g., Mistral × NVIDIA (optimizing latest family of models on NVIDIA hardware), OpenAI × Broadcom). 🏗️ The bigger-faster trend continues: multiple 𝗚𝗪-𝘀𝗰𝗮𝗹𝗲 projects announced (e.g., xAI × NVIDIA “Colossus 2”, Meta × Blue Owl “Hyperion”). 🍽️ Another shift: 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 → 𝗶𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲. Deals like Fireworks AI × AMD and Baidu × Samsung are explicitly about serving/optimizing inference at scale. NVIDIA × Groq announcement confirms this trend. 2️⃣ 𝗜𝗻𝘁𝗲𝗿𝗼𝗽𝗲𝗿𝗮𝗯𝗶𝗹𝗶𝘁𝘆: 𝘁𝗵𝗲 𝗺𝗼𝗮𝘁𝘀 𝗮𝗿𝗲 𝗰𝗵𝗮𝗻𝗴𝗶𝗻𝗴 We’re moving from setting standards (MCP, A2A) to real commercial interoperability. The headline example is Snowflake and Microsoft agreeing on “zero-copy, bi-directional” data sharing. Unthinkable a couple of years ago. Companies are increasingly willing to 𝘁𝗿𝗮𝗱𝗲 𝗼𝗳𝗳 some 𝗱𝗮𝘁𝗮-𝗴𝗿𝗮𝘃𝗶𝘁𝘆 𝗮𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲 to: • enable customers use AI services (MSFT) • remain the source of truth even when AI tooling lives elsewhere (SNOW) More deals like this: SAP ↔ MSFT & Databricks, SAP ↔ Snowflake, Snowflake ↔ Tableau. 3️⃣ 𝗖𝗼𝗻𝘁𝗲𝗻𝘁 & 𝗜𝗣 𝗹𝗶𝗰𝗲𝗻𝘀𝗶𝗻𝗴 Two trends stand out: • data/IP licensing is 𝘀𝗵𝗶𝗳𝘁𝗶𝗻𝗴 𝗳𝗿𝗼𝗺 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝘁𝗼 𝗿𝗲𝗮𝗹-𝘁𝗶𝗺𝗲 𝗴𝗿𝗼𝘂𝗻𝗱𝗶𝗻𝗴 (Meta x CNN, Fox News, Reuters, etc.) • OpenAI ↔ Disney is in a category of its own (~$1B), and could set a new 𝗽𝗿𝗶𝗰𝗶𝗻𝗴 𝗽𝗿𝗲𝗰𝗲𝗱𝗲𝗻𝘁 𝗳𝗼𝗿 𝗜𝗣. (See great article by Stratechery in comments.) We also see continuation of trends from previous quarters: • 𝗗𝗶𝘀𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻 𝗶𝘀 𝗸𝗶𝗻𝗴: Anthropic × Deloitte, Accenture, OpenAI × Intuit, Perplexity × Bharti Airtel • 𝗦𝗼𝘃𝗲𝗿𝗲𝗶𝗴𝗻 𝗔𝗜: “OpenAI for Germany” with SAP, Arabic AI tools via Humain x Adobe x Qualcomm, Google × UK government • 𝗖𝗵𝗮𝘁𝗯𝗼𝘁𝘀 → 𝗲𝗮𝗿𝗹𝘆 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀: Perplexity × PayPal, Lovable × Atlassian, OpenAI × Instacart & Zillow

Semiconductor and AI valuations are being driven by real demand — but the capital isn't flowing evenly. The barbell is back, and the middle is thinning out. That was the consensus from the Investment Outlook Panel, moderated by Lilly Huang of Citi Private Banking, featuring Lisa Cui (VP of Corporate Development, Lumentum), Kevin Chen, PhD (Managing Director, Lam Capital; Corporate Development, Lam Research), and Siddharth Gupta (Managing Director, Samsung Ventures). Other insights: - A barbell market. Crowded, expensive AI-hot deals on one end; overlooked but strategically important companies on the other. Fewer obvious winners in the middle. - Frothiness is concentrated in the application layer. Software and some AI application companies are the frothiest part of the market, with concern that weak moats can disappear quickly as foundation models absorb features. - Hardware is where differentiation still lives. The panel was more constructive on infrastructure, manufacturing, power delivery, and advanced packaging — technically deep areas where differentiation is harder to copy and diligence is more specialized. - Supply chains are restructuring. Onshoring, friend-shoring, and dual supply chains are becoming normal due to geopolitics and customer requirements. Higher cost and complexity, but better resilience — and new openings for suppliers. - Energy is the next bottleneck. Power is a major constraint on AI infrastructure. The investment focus isn't just generation — it's using electricity more efficiently inside data centers through better delivery, conversion, and storage. - Robotics and physical AI are next. Positive outlook on robotics and automation as the next wave of AI monetization beyond the data center, though deals are expensive and the market is early. Industrial use cases look nearer-term than consumer humanoids. - Custom silicon is durable. Application-specific chips are a lasting trend driven by specialized AI workloads. The open question: can margins hold as custom designs become more common? - Advanced packaging is a core theme. Packaging, optics, and integration technologies are increasingly important — they enable performance gains and make custom silicon more practical. - Startups need incumbents. Even with a strong technology opportunity, semiconductor startups often need partnerships with large incumbents to get adoption — especially in packaging and deep hardware. - Companies highlighted positively included Multibeam Corporation, a power-component company for data centers, and Niron Magnetics, Inc. — the last notable for making permanent magnets from iron and nitrogen as an alternative to rare-earth dependence on China. #Semiconductors #AI #Investing #VentureCapital #AdvancedPackaging #Robotics #CustomSilicon #SupplyChain #DeepTech #AIInfrastructure