The global artificial intelligence boom is straining supply chains in unexpected ways, with optical components and laser materials emerging as critical bottlenecks that few in the industry anticipated. While most attention focuses on processor shortages, the real squeeze is happening in the unglamorous world of fiber optics, laser sources, and specialized substrates needed to connect AI chips together. China's export controls on indium, a key material for making lasers, have further complicated an already fragile supply situation.

What Are Co-Packaged Optics and Why Do They Matter?

The AI infrastructure world is undergoing a fundamental shift in how data moves between chips and servers. Traditionally, AI systems have relied on copper wiring and pluggable optical transceivers, small devices about the size of a lipstick that convert electrical signals into light for faster communication. But industry leaders from Taiwan Semiconductor Manufacturing Company (TSMC), Intel, Nvidia, and Broadcom increasingly see a new approach called co-packaged optics (CPO) as the future.

CPO places optical engines directly next to the main processor, dramatically speeding up communication between chips and servers while saving space and reducing power consumption. TSMC has developed a technology called COUPE, short for compact universal photonic engine, which integrates photonic and electronic chips into a single package. Industry executives are calling 2026 the dawn of the CPO era, though adoption will initially be limited to the most advanced AI systems.

Where Are the Supply Chain Bottlenecks Hitting Hardest?

The shift toward optical-heavy AI infrastructure is creating shortages in places most people have never heard of. Optical fiber has become a strategic resource in the AI data center buildout. Nvidia has signed a long-term agreement with Corning to expand the latter's U.S. fiber capacity tenfold, signaling just how critical this material has become.

But the real crisis is in laser sources and the materials needed to make them. Indium phosphide substrates, a key material for manufacturing lasers used in optical transmission, are increasingly constrained. This shortage caught many in the industry off guard. China plays a dominant role in the global market for pluggable optical transceivers, with companies like Zhongji Innolight and Eoptolink serving major customers including Google, Nvidia, and Amazon. Both companies have seen their profits skyrocket by over 800% and 900%, respectively, since the AI boom began in 2022.

How Are Supply Chain Pressures Reshaping the Industry?

• Fiber Expansion: Major chipmakers are locking in long-term supply agreements for optical fiber, with Nvidia committing to expand Corning's U.S. capacity by a factor of ten to secure future supplies.
• Substrate Scarcity: Indium phosphide substrates for laser production have emerged as an unexpected chokepoint, with China's export controls on indium further limiting availability and driving up costs.
• Chinese Market Dominance: Chinese optical component manufacturers are capturing massive profits from the AI boom, with some seeing profit increases exceeding 800% since 2022 as global demand outpaces supply.

The convergence of optics and electronics is reshaping supply chains in ways that go far beyond simple processor availability. Companies are committing to supply agreements longer than ever before. According to industry insiders, customers are willing to lock in memory and component capacity for five years in advance, a sign of just how desperate the scramble for resources has become.

What Does Huawei's Chip Comeback Mean for Export Controls?

While optical shortages are creating supply chain chaos, another major shift is underway on the geopolitical front. Six years after U.S. export controls cut Huawei off from global semiconductor production, the Chinese tech giant has outlined an ambitious chip comeback strategy. Huawei's semiconductor chief, He Tingbo, unveiled a new design approach and technology roadmap aimed at overcoming severe U.S. export restrictions.

The challenge Huawei faces is stark: China still lacks access to advanced equipment such as extreme ultraviolet (EUV) lithography machines needed for cutting-edge chip production. Despite these constraints, Huawei is pursuing new Kirin smartphone processors as part of its strategy to reduce dependence on foreign semiconductors. The company is also grappling with the challenge of boosting the quality and volume of chip production to meet massive local demand.

How Are Global Investors Repositioning Away From China?

As political and regulatory pressures mount, global investment firms are quietly exiting China's data center sector. Princeton Digital Group, backed by Warburg Pincus, is launching a sale of its China assets that could fetch as much as $1 billion. The company owns data centers across six Chinese cities. This sale would mark the near end of a decade-long push by global private equity firms to invest directly into China's AI infrastructure.

Similar exits are happening across the industry. Bain Capital and Carlyle have already begun divesting from Chinese data centers as foreign ownership of critical digital infrastructure becomes increasingly fraught in the country. From 2017 onward, buyout groups poured billions into the Chinese data center sector, drawn by booming demand from cloud providers owned by Alibaba, Tencent, and ByteDance. But rising AI-driven demand has lifted valuations, giving international investors an opportunity to sell to domestic buyers and redeploy capital into politically safer markets elsewhere in Asia.

Why Is Memory Becoming as Important as Computing Power?

While the industry obsesses over processor power, a different vision of the AI future is emerging. The chief technology officer of SanDisk, one of the world's leading NAND flash memory and storage solution makers, offered a striking perspective: the AI race could become increasingly "memory-centric" rather than focused solely on computing power.

The executive argued that humans function more as "a collection of memories" than simply as raw computing brains, and said future AI systems may evolve in a similar way. This shift in thinking reflects a broader industry trend. Leading memory chipmakers Micron and SK Hynix have both joined the trillion-dollar market valuation club, a milestone that underscores the growing importance of memory in the AI era. The fact that customers are willing to commit to five-year supply agreements just to lock in memory capacity early demonstrates how critical this resource has become.

The convergence of optical bottlenecks, geopolitical tensions, and shifting investment patterns reveals that the AI boom's supply chain challenges run far deeper than chip availability alone. From fiber optics to laser substrates to memory capacity, every layer of the infrastructure stack is under pressure, forcing companies to rethink their strategies and lock in supplies years in advance.