AI Hardware Boom Needs Factories

The United States is experiencing a nascent manufacturing expansion, with upticks in new orders and production. However, this "boom-let," as some term it, is overwhelmingly concentrated on the infrastructure for artificial intelligence.

While there's increased focus on AI-related parts and machinery, a critical bottleneck exists: the lack of sufficient factories and the machines that build those machines. This "capital stock" problem means imports of industrial goods outpace the domestic capacity to produce them.

Rebuilding this manufacturing base is a formidable challenge, partly due to decades of underinvestment. Capital spending on industrial structures and equipment as a share of GDP has been in steady decline since the early 1980s, shifting focus from tangible production to intellectual property.

AI's Singular Focus

Even within the equipment sector, which has seen investment growth nearing 9%, the boom is narrowly defined. The vast majority of this investment is tied to AI development, particularly in semiconductors.

Investment in core industrial machinery outside of AI is significantly smaller, and much of that is also AI-adjacent. This concentration means current industrial investment aligns more with existing global supply chains than with establishing new domestic ones.

The disparity is stark when looking at structures versus equipment. While equipment spending has seen recent gains, investment in the necessary factory buildings has not kept pace. Without adequate structures, true reindustrialization is impossible.

The CHIPS Act spurred some semiconductor-related construction, but this has cooled considerably. Outside of that sector, manufacturing construction growth remains modest, with the bulk tied to chemicals, not a broad expansion of domestic capacity.

Spending on machinery construction has more than doubled since 2022, but from a very low base. This growth, while positive, is insufficient to meet broader industrial revival needs.

The US was once a net exporter of machinery, but that trend reversed around 2013. While domestic machinery supply has grown, its expansion has tapered off since 2021.

The data clearly indicates a long road ahead for domestic manufacturing. We are buying AI components, but the critical infrastructure, the buildings and machines to produce the machines that make the parts, is not yet being adequately built out, a stark contrast to the US manufacturing investment trends.

This situation highlights a critical gap: while AI investment fuels specific sectors, a comprehensive AI hardware manufacturing boom requires a parallel boom in the foundational elements of production, from chip equipment manufacturers like Lam Research and Intel surge on BofA upgrades as chip equipment recovery accelerates, SOXX +8.4% to the very structures that house them.