Optics Shortage Creates AI Hardware Bottleneck
A significant shortage of co-packaged optics (CPO), High Bandwidth Memory (HBM), and 3D NAND is becoming a major topic of discussion in the AI and robotics community. The scarcity of these components is seen as a potential bottleneck for developing next-generation AI systems. This has sparked conversations about investment opportunities in companies that can address the complex manufacturing challenges and supply chain constraints.
- The co-packaged optics (CPO) market is projected to grow from an estimated $161.43 million in 2026 to $748.62 million by 2031, driven by key players like Broadcom, Marvell, Intel, and the startup Ayar Labs. NVIDIA's decision to integrate CPO into its GPU networking stack has signaled to the industry that older board-level designs will not meet future latency targets for AI. - The pivot to AI has made High-Bandwidth Memory (HBM) a systemic bottleneck, with manufacturing capacity now dictating the pace of AI hardware growth. Micron reported its entire 2025 HBM production was sold out before the year began and forecasts that the memory shortage will extend through 2026. - This HBM scarcity is intensified because major memory manufacturers like Samsung, SK Hynix, and Micron are diverting manufacturing capacity away from standard DRAM to produce the more profitable, multi-layered HBM chips, causing shortages and price hikes for components used in consumer electronics. - The demand for HBM is now driven by its status as a mandatory, performance-defining component for flagship AI accelerators like NVIDIA's Blackwell B200 and AMD's MI300X, a shift from its prior status as a specialized, optional component for high-performance computing. - In the 3D NAND market, major suppliers like Samsung, SK Hynix, Micron, and Kioxia hold over 85% of the market share and are competing by increasing storage density through higher layer counts, with a rapid transition from 100-layer chips to over 200 layers. - To meet the exploding demand for AI-related hardware, global investment in new semiconductor fabrication facilities is projected to exceed $1.5 trillion between 2024 and 2030. In the third quarter of 2025 alone, worldwide billings for semiconductor equipment reached a record $33.66 billion. - A critical barrier to scaling CPO production is a shortage of skilled professionals in photonics. This talent gap in manufacturing, testing, and assembly is cited by industry leaders as a primary obstacle to innovation and meeting future demand. - The transition to optics is necessary because traditional copper-based interconnects are failing to meet the bandwidth, power, and latency requirements of multi-terabit AI clusters. Marvell has demonstrated that its CPO technology can achieve data transfer rates and distances up to 100 times greater than electrical cabling.
