The AI Race Is Becoming a Supply Chain Race

For most of the artificial-intelligence boom, we have talked about AI primarily as a technology story. The conversation has centered on better models, more capable agents, faster computing, and new applications. Nvidia’s latest earnings suggest that we may be entering a different phase. The next part of the AI race may be determined as much by supply chains as by software.

Nvidia reported $96.2 billion in quarterly revenue this week, more than double its revenue from the same period a year ago. Data-center revenue alone reached $89 billion, an increase of 117% year over year, and the company expects approximately $108 billion in revenue next quarter.[1] Those numbers are extraordinary, but the demand story may actually be the less interesting part. The bigger question is whether the physical infrastructure supporting AI can expand quickly enough to keep up with it.

That is easy to overlook because AI feels almost weightless when we use it. We open a browser, type a prompt, and receive an answer seconds later. But there is an enormous industrial system sitting behind that interaction. Advanced processors have to be fabricated and packaged with high-bandwidth memory. Servers have to be assembled and connected. Data centers have to be built, powered, and cooled. All of it depends on suppliers, specialized equipment, skilled labor, transportation networks, energy infrastructure, capital investment, and years of planning.

That creates an interesting mismatch. Software can scale remarkably quickly. Factories, electrical grids, semiconductor plants, and data centers cannot. When demand for computing grows faster than the physical system supporting it, having the best technology is no longer enough. Companies also need access to the resources required to actually deploy it.

High-bandwidth memory is a good example. This specialized memory is essential for high-performance AI computing, but production capacity is limited and difficult to expand. SK Hynix, the leading producer of high-bandwidth memory, said this week that it expects the current memory shortage to persist through the end of 2030. The company is investing more than $4 billion in an Indiana facility that will produce next-generation HBM4E chips and serve as a major U.S. production base for advanced AI memory.[3] Nvidia and SK Hynix have also entered a multiyear partnership aimed at developing next-generation memory and expanding supply for the global AI infrastructure buildout.[2]

This is where a very old supply-chain lesson becomes relevant to one of the newest industries in the world. When an input is readily available, purchasing tends to revolve around price, quality, service, and efficiency. When that same input becomes scarce, the equation changes. The priority becomes access.

Companies begin signing longer-term agreements, reserving capacity, investing upstream, building deeper supplier relationships, and sometimes accepting redundancy that would have looked inefficient a few years earlier. The logic is fairly simple: when the cost of not having supply becomes greater than the cost of securing too much of it, resilience starts to look a lot less expensive.

I made a similar argument earlier this summer with my colleague Gerry Goldstein in Commercial UAV News. In “The Drone Supply Chain Is the Next Strategic Battleground,” we argued that technological superiority in autonomous systems means very little if the industrial base cannot produce those systems at scale.[5] AI is beginning to expose the same principle on a much larger stage. Innovation matters enormously, but innovation without the capacity to manufacture, power, connect, and sustain it eventually runs into a wall.

We can already see how far upstream the AI opportunity is spreading. Nvidia and Amazon Web Services announced this week that they plan to deploy an additional two million Nvidia GPUs across AWS infrastructure during 2027 and 2028.[4] Those GPUs do not exist in isolation. They create demand for memory, servers, networking equipment, buildings, cooling, electricity, construction, maintenance, and transportation. The AI economy is creating an industrial ecosystem around the technology itself, and some of the most important companies in that ecosystem may never build an AI model.

There is certainly risk in all of this. Reserving capacity and investing billions of dollars in infrastructure assumes that future demand will materialize. Technology history is full of companies that expanded near the top of a cycle only to discover that demand changed before the capacity arrived. Nvidia and its suppliers are making enormous bets about the future of AI, and those bets are now being embedded in factories, contracts, data centers, and infrastructure that can take years to build.

A company can announce an AI strategy, develop an extraordinary model, or identify an enormous market opportunity. Eventually, somebody has to deliver it. That requires suppliers, capacity, infrastructure, energy, capital, logistics, and coordination. Artificial intelligence does not make those fundamentals obsolete. If anything, its extraordinary growth is reminding us how important they are.

The next phase of AI competition will certainly involve better models and more capable software. Increasingly, though, it will also depend on which organizations secured the memory, manufacturing capacity, data-center space, power, suppliers, and infrastructure before everyone else realized just how valuable those resources would become.

The AI race is becoming a supply chain race.

References

1. NVIDIA. NVIDIA Announces Financial Results for Second Quarter Fiscal 2027. August 26, 2026. NVIDIA Investor Relations. View source ↗
2. NVIDIA. NVIDIA and SK hynix Announce Multiyear Technology Partnership to Advance Memory for AI Factories. June 7, 2026. NVIDIA Newsroom. View source ↗
3. Reuters. SK Hynix to Start AI Chip Output in Indiana in 2029, Sees Memory Shortage Through 2030. August 27, 2026. View source ↗
4. NVIDIA & Amazon Web Services. AWS and NVIDIA to Deliver 2 Million Additional GPUs and Next-Generation Infrastructure for Agentic and Physical AI. August 26, 2026. NVIDIA Investor Relations. View source ↗
5. O’Connor, Kenneth W., and Gerry Goldstein. The Drone Supply Chain Is the Next Strategic Battleground. June 29, 2026. Commercial UAV News. View source ↗

Kenneth W. O’Connor, Ph.D.

Professor, researcher, and business practitioner focused on professional sales, supply chain logistics, transportation, artificial intelligence, defense manufacturing ecosystems, and applied business education.