America’s Investment in AI Infrastructure
In an effort to take the lead in artificial intelligence technology, the United States is pouring hundreds of billions of dollars into building data centers. There’s a lot of heated debate among lobbyists and lawmakers concerning who should be responsible for making the chips used in these facilities. However, fewer people are asking about the essential components that connect these chips, especially given that many of them come from China. This oversight could be costly.
Optical transceivers play a critical role in data centers, enabling rapid data transfers between servers and facilities. Almost all major AI clusters in the country rely on these devices, and China is the dominant supplier. For instance, Yury Dvorkin from Johns Hopkins noted that China supplies around two-thirds of the world’s optical transceivers.
Companies like Innolight and Eoptolink are among the top suppliers, with Innolight recently being added to the Department of Defense’s list of Chinese military companies. This raises the question: is it wise for any nation to let a potential adversary have such a stronghold on an industry that is becoming increasingly important?
This isn’t just coincidental. China’s approach is quite clear: they leverage state funding and industrial strategies to develop critical technologies, undercut foreign companies, and then leverage their market position to gain influence. We’ve seen this pattern in various sectors, from solar panels to batteries—it’s a lesson we should have already learned.
In the past, rural carriers in the U.S. purchased inexpensive equipment from Huawei and ZTE without much thought. Once Congress finally recognized the associated risks, they had to allocate $5 billion in taxpayer money to remove and replace that equipment—a process that remains unfinished. We risk repeating this mistake, but the implications could be much more significant this time.
Furthermore, Chinese firms have become adept at sidestepping existing regulations, often routing goods through third countries to evade tariffs. In fact, the White House estimated that these transshipment practices have led to a loss of $19 to $26 billion in tariffs.
However, there appears to be movement in Washington. Reports have emerged that the Federal Communications Commission (FCC) is preparing to quickly ban imports of Chinese optical transceivers to safeguard our AI infrastructure. FCC Chairman Brendan Carr seems committed to prioritizing national security. Recently, bipartisan legislation was introduced in the Senate aimed at excluding Chinese transceivers from sensitive national security systems. Although these measures are promising, they face opposition from industry stakeholders.
Major technology companies and industry associations have pushed back against these proposals. Yet an open letter from some of these same companies called for a more secure digital environment in the era of AI. If security is truly a concern, then attention must be paid to the components used in data centers.
Ultimately, the security of an AI model is only as strong as the hardware running it. Building a dependable AI system on a foundation provided by companies tied to the Chinese government isn’t feasible. If tech giants genuinely valued security, they would understand that the infrastructure within data centers must be part of the conversation.
Theodore Roosevelt once stated that a nation unable to secure its own industrial base cannot effectively govern itself. The outcome of the AI competition will hinge on infrastructure, making optical transceivers just as vital as chips. To ensure America’s safety, the FCC should implement its ban, Congress needs to pass relevant legislation, and tech firms must cease their lobbying for Chinese equipment.
We’ve already spent once to eliminate Huawei’s influence. Let’s not find ourselves in that position again.
