Artificial intelligence can feel almost weightless to the person using it. Someone types a question into a computer, receives an answer within seconds, and never sees the machinery required to make that exchange possible.
Behind the screen is an enormous physical system. Data centers need servers, power distribution, cooling equipment, backup systems, structural supports, enclosures, platforms, piping, and the factories that make those products. Semiconductor manufacturers need substrates, packaging, testing equipment, optical components, and additional production capacity.
The software may attract the attention, but the expansion depends on manufactured hardware.
This matters to executives, purchasing professionals, and suppliers because AI spending is moving well beyond technology companies. It is creating demand throughout the industrial supply chain, including opportunities for equipment builders, HVAC manufacturers, electrical suppliers, contractors, and custom fabricators.
The important question is no longer whether AI will affect manufacturing. It is where the physical spending will land and which companies will be prepared to support it.
On October 21, NVIDIA announced a coordinated expansion involving several manufacturing partners. The company said new substrate agreements would support a 2.5 times increase in capacity over the next three years.
Amkor plans to increase advanced packaging capacity by more than 300 percent by the end of 2028. NVIDIA also expects its semiconductor testing capacity to double between 2026 and 2027. A $1.1 billion capacity reservation advance will support new testing facilities and equipment in the United States, while another supplier is targeting a 15 times increase in optical engine capacity by the middle of 2027. investor.nvidia.com
These are not software development expenses. They are investments in factories, production equipment, testing operations, material handling, and the infrastructure needed to support higher output.
The following day, Carrier announced plans to invest more than $600 million in a new manufacturing and engineering center in Fort Worth, Texas. The facility will exceed one million square feet, create approximately 1,500 jobs, and expand production of commercial chillers and other thermal management systems.
Carrier specifically connected the investment to record backlog and demand across data centers, healthcare, education, and industrial markets. carrier.com
Together, the announcements show how quickly AI demand spreads. More computing capacity requires more semiconductor capacity, but it also requires additional cooling, power, buildings, equipment, and supporting components.
It would be easy to look at AI infrastructure and assume that the opportunity belongs only to semiconductor companies or enormous technology suppliers.
The actual supply chain is much broader.
A commercial chiller manufacturer may need fabricated frames, panels, brackets, tanks, guards, and access components. An electrical equipment manufacturer may need cabinets, enclosures, mounting systems, and welded assemblies. Equipment integrators may need skids, platforms, carts, supports, and protective structures.
The facility producing those products may also need new material handling equipment, machine guarding, storage systems, and production support components as its own volume increases.
Not every fabrication supplier will sell directly to a data center owner or a company such as NVIDIA. Many opportunities will appear two or three levels deeper in the supply chain through the companies manufacturing the equipment that makes AI infrastructure possible.
That distinction matters for business development. Searching only for data center projects can miss the manufacturers adding capacity because their customers serve data centers.
The better signal may be an equipment manufacturer announcing record backlog, a semiconductor supplier reserving production capacity, or an HVAC company building another factory.
Those announcements show where demand is becoming physical.
Carrier did not announce another data center. It announced a manufacturing facility intended partly to satisfy demand created by data centers. NVIDIA did not announce only another computing product. It announced capacity agreements and manufacturing expansions involving substrates, packaging, testing, and optical components.
For industrial suppliers, those second order effects may be more useful than the original technology headline. They identify the companies that need buildings, equipment, materials, labor, and supplier capacity to fulfill the larger investment.
They can also reveal which requirements are likely to become more important. Products supporting data centers and advanced computing may face expectations involving repeatability, documentation, corrosion protection, thermal performance, cleanliness, traceability, or higher production volumes.
A supplier does not need to guess at those requirements. It needs to ask how the component will be used and what the next operation expects from it.
Large investment figures can create the impression that every related opportunity is ready to quote today, although announced capacity does not become available all at once.
Some capacity will come from existing facilities. Some suppliers are upgrading equipment, while others are building factories that will not reach full production for several years. NVIDIA’s announced expansions include different production targets across 2027 and 2028.
This means manufacturers should separate announced demand from released demand. A company may have a strong long term outlook while its individual purchasing packages are still being designed, funded, or approved.
For a fabrication supplier, the most useful questions concern the equipment program, the expected production ramp, and the components that must be ready first. For a buyer, the important questions concern supplier capability, inspection, documentation, finishing, packaging, and the ability to maintain consistency as quantities grow.
The goal is not to chase every headline. It is to understand where the headline is creating a real manufactured product.
When a request is tied to a growing technology market, the first quantity may not represent the complete opportunity.
A small initial order may support a prototype, an equipment qualification, or the first installation. If the program advances, the same component could move into a larger and more consistent production requirement.
That possibility should lead to a better quoting conversation. The buyer and supplier should clarify whether the order is a prototype, a limited build, or the beginning of a volume program. They should identify which tooling, fixtures, inspection methods, and packaging decisions may need to change if demand increases.
This does not mean pricing every first order as though it will become a major program. It means avoiding decisions that make the next stage unnecessarily difficult or expensive.
A modest fabrication request can be connected to a much larger equipment investment. Understanding that connection helps both companies make better choices.
The growth of artificial intelligence is often described through software, computing performance, and investment totals. The physical side of that growth is equally important.
NVIDIA’s supplier expansions and Carrier’s new manufacturing center show that AI infrastructure demand is already reaching factories, equipment manufacturers, and industrial supply chains.
For manufacturing leaders, the opportunity is to look beyond the most visible technology company and understand which products must be built around it. For suppliers, the opportunity is to recognize where their capabilities fit into that larger system without claiming work they are not prepared to support.
AI may begin with software, but its growth depends on cooling equipment, power systems, enclosures, platforms, structural components, and the factories that manufacture them.
For a closer look at how that demand reaches the fabrication floor, read Building the Next Infrastructure Boom and Metal Fabrication Is at the Center of It. The article explains how laser cutting, forming, robotic welding, and other fabrication processes support the physical infrastructure behind the AI economy.