Chip Land Economy and Future of Value Chain Beyond Nvidia
Intel CEO Pat Gelsinger has often been heard saying that Moore’s Law is “alive and well," but recently, he seems to have admitted that the pace of the semiconductor industry has at least slowed down. However, Nvidia's CEO Jensen Huang offers a starkly different view, stating that Moore’s Law has ended. As Nvidia, now three times more valuable than Intel, challenges this decades-old theory, the future of semiconductor innovation hangs precariously on the edge, with Nvidia poised to tilt the scales. This shift underscores a broader trend in the industry: fewer and fewer companies have managed to stay competitive in the race to develop the most cutting-edge chips in recent years. The processes of design and manufacturing have grown increasingly lengthy, intricate, and expensive, necessitating more specialized equipment and expertise. Notably, the smallest transistors are now 10,000 times smaller than a human hair, highlighting the extraordinary precision required in modern chip production.
The semiconductor sector is abuzz with activity, with NVIDIA taking the lead, experiencing a remarkable surge of over 145% this year! This surge can be attributed to the CUDA parallel computing platform and programming model. Introduced in 2007, CUDA revolutionized NVIDIA's GPUs, originally tailored for gaming and 3D applications, by enabling them to outperform CPUs in high-performance computing (HPC) tasks. By breaking down tasks into smaller components and processing them simultaneously, CUDA propelled NVIDIA into the forefront of innovation. But it's not just Nvidia making waves—Taiwan Semiconductor, Arm Holdings, Qualcomm, and EUV lithography monopolies like ASML are also riding the AI wave.
🔍 Now, let's dive into some key highlights:
🎛️ Semiconductors:
TSMC's Capacity Boost: With Nvidia's Blackwell platform, TSMC's CoWoS capacity is set to increase by 150% in 2024 and nearly double by 2025.
AI's Impact on HBM3e: HBM3 is the latest generation of High Bandwidth Memory (HBM). Nvidia and AMD are driving HBM3e into the mainstream, expected to dominate by late 2024. From the standpoint of original manufacturers, the main focus of competition in the HBM3e market centers on Micron, SK Hynix, and Samsung. Analysts report that Micron and SK Hynix have fully booked their HBM capacity for 2024 and H1 of 2025, fueled by skyrocketing demand for AI chipsets, which in turn drives up the need for high-end memory chips.
Samsung has about 40% HBM market share.
SK Hynix has abour 30% HBM market share.
Micron has 26% HBM market share.
Nanya has 2%.
💻 Servers: AI server specialist Super Micro Computer has more than doubled in value, Being a smaller company means that sales of AI servers are moving the needle more significantly for Supermicro. Its modular server rack-scale systems are being used to mount AI-related graphics cards from Nvidia, not to mention the fact that it is benefiting from positive developments in its supply chain and experiencing an uptick in customer acquisition for its air-cooled and liquid-cooled AI server systems. Charles Liang, President and CEO of Super Micro Computer, recently mentioned that the company's liquid cooling solutions ca potentially reduce data center power consumption by up to 40%. Dell Technologies on the other hand may see a surge in growth in the latter part of 2024, propelled by the widespread adoption of AI-enabled PCs. According to market research firm Canalys, shipments of AI-enabled PCs are projected to surge from 48 million units this year to an annual total of 205 million units by 2028.
⚡ Utilities/Power Solutions: As artificial intelligence (AI) advances rapidly, the importance of reliable power sources and infrastructure becomes increasingly evident. Underlining this urgency, the International Energy Agency (IEA) estimates that electricity demand from dedicated AI data centers could grow by over 10 times from 2023 to 2026. Power companies like Vistra Energy and Constellation Energy are thriving, boosted by the AI-driven data center power surge predicted to hit 35 GW by 2030.
❄️ Heat Management/Cooling: Nvidia’s liquid cooling supplier Vertiv Holdings has skyrocketed nearly 500% this year! Eaton and nVent Electric also see significant gains, reflecting the growing need for advanced cooling and power management.
☁️Hyperscaler: Hyperscalers are cloud providers with a large network of data centers and a wide range of services, and are often preferred for end-to-end workload support. Three of Wall Street’s heavyweight technology firms have started spending again on cloud computing after pausing last year to cut costs. AI services have propelled Azure's growth, contributing seven percentage points this quarter, up from six in the last. Microsoft’s cloud is now growing significantly faster than Amazon Web Services, partly thanks to its strong partnership with OpenAI.
Analysts estimate Azure, once half the size of AWS five years ago, is now three-quarters its size.
Azure's revenue surged 30% this quarter, compared to AWS's 13% year-over-year growth.
Across AWS, Microsoft Azure, and Google Cloud, two trends are clear: AI is driving growth, and overall cloud spending is accelerating.
📈 Other Key Players: Semiconductor equipment giants Applied Materials, KLA Corp, and ASML Holding, along with data storage leaders Pure Storage and Western Digital, are all benefiting from this AI-driven market expansion.
🌱 Emerging Trends:
New Entrants: The semiconductor landscape is poised for disruption with the emergence of new chips from both established and up-and-coming AI chipmakers. Additionally, companies traditionally known as chip buyers are venturing into in-house chip production. For instance, Apple Inc.'s (AAPL) installed base of over 2.2 billion active devices is considered the "best" in the world regarding its AI chip opportunity. The outcomes for these new players remain uncertain, adding an element of anticipation to the evolving semiconductor market.
Advancements in Semiconductor Assembly and Test: The increasing complexity of chip and module designs elevates the significance of testing. Both Captive Semiconductor Assembly and Test (AT) and Outsourced Semiconductor Assembly and Test (OSAT) facilities are poised to enhance capabilities in system-level testing, adaptive or dynamic testing, and AI/ML-based bin prediction.
Edge Chips Advancements: Anticipated shifts towards edge computing in 2024 may witness increased processing at the edge, utilizing smaller, cheaper, or different accelerators and AI models compared to those dominant in 2023.
Cybersecurity Prioritization: The global expansion of semiconductor manufacturing and supply facilities necessitates a heightened focus on cybersecurity. Insights from Deloitte’s 2023 global semiconductor industry outlook underscore the critical need for advanced cyber solutions to safeguard digital environments across the industry's spectrum. Research firm McKinsey estimates that the corporate sector is set to invest $213 billion in cybersecurity software in 2024. Despite this substantial expenditure, it falls far short of the necessary defense against cybercrime, which Cybersecurity Ventures predicts will cause $9.5 trillion in damages over the course of the year.
Addressing the Bullwhip Effect: The semiconductor industry's susceptibility to over-ordering and excess inventories during times of undersupply can lead to the bullwhip effect. This phenomenon, expected to occur in the short term by H2 of 2024 or H1 of 2025, underscores the need for strategic inventory management to mitigate fluctuations in sales and prices per unit.
Impact of Export Controls: Advanced node manufacturing, especially below 10 nm, relies heavily on cutting-edge technologies like extreme ultraviolet lithography (EUV). Export controls, particularly on EUV machines and supporting technologies, pose challenges to the production of advanced node chips, impacting costs and yields negatively.
Nevertheless, the semiconductor and tech landscape is evolving rapidly, with AI at the forefront, driving innovations and growth across various sectors. Stay tuned as the future of tech unfolds!
#Semiconductors #AI #TechInnovation #MarketTrends #FutureOfTech #ChipWar #MooresLaw #IntelVsNvidia #ChipInnovation #Hyperscaler
Source:
Worldwide Semiconductor Trade Statistics (WSTS), “WSTS semiconductor market forecast – Fall 2023,” press release, dated November 28, 2023.
Deloitte analysis of public market capitalizations in 2023, 2022, and 2021.