There is a number that frequently comes up in discussions between chip executives and infrastructure analysts, and it has the power to silence the room. The market for AI servers is expected to grow from about $195 billion to almost $2.85 trillion by 2034, a compound annual growth rate that would have seemed ridiculous five years ago. This pace was unanticipated. Not the analysts. not the producers. Not the supply chains that are currently frantically trying to meet demand, which has, by most honest accounts, completely destroyed their models.
The factory floor, rather than forecast slides, is the most obvious indication that something out of the ordinary is occurring. A Morgan Stanley analysis released this week claims that the cost of memory chips has increased sixfold in the last year alone. A healthy, well-supplied market does not experience that kind of price movement. It occurs when sellers are unable to build quickly enough and buyers are in a desperate situation. The shares of Samsung, SK Hynix, and Micron, which collectively account for almost 90% of the world’s memory production, have more than tripled. In the meantime, their clients are either silently absorbing expenses or transferring them to customers who are unaware of what is about to happen.
Morgan Stanley refers to the phenomenon as “chipflation,” and it’s a term worth considering. What began as a bottleneck in hyperscale data centers is now affecting the cost of cloud subscriptions, laptops, and smartphones. Sony has increased the cost of the PlayStation. Lenovo has changed the price of PCs. The AI infrastructure arms race is already changing how the biggest software company in the world spends money, as Microsoft admitted in April that about $25 billion of its projected $190 billion in spending this year comes directly from higher chip costs. That is no longer a data center issue. It has to do with macroeconomics.
Unexpectedly, Dell Technologies has become one of the best indicators of how this develops for hardware firms that are prepared to act quickly. In fiscal 2026, the company’s AI server revenue increased by about 2.5 times to $25 billion, from its estimated 20 percent market share in 2024. After receiving $64 billion in new orders for AI servers, it ended the year with a $43 billion order backlog. These figures indicate that demand isn’t declining and businesses aren’t waiting to see what happens next. This fiscal year, Dell anticipates that revenue from AI servers will double once more, reaching $50 billion. The direction of travel is clear whether or not that prediction comes to pass.
The sovereign angle is one of the things that sets this moment apart from earlier cycles of technology investment. Governments are now purchasing AI infrastructure in large quantities, not just businesses. Dell has been setting up servers in South Korea, Malaysia, and other places as nations rush to develop language models that are culturally sensitive, maintain sensitive data inside their borders, and develop domestic AI capabilities. By 2034, the sovereign cloud market, which is currently estimated to be worth $155 billion, is expected to grow to $1.13 trillion. Three years ago, that category hardly showed up in the models of most analysts.
Observing all of this, it seems as though the supply chain was not designed for a world in which national governments and Fortune 500 companies are vying for the same chips, servers, and manufacturing capacity. It’s difficult to disagree with Larry Fink of BlackRock’s recent assertion that compute demand is increasing more quickly than any supply chain can keep up. The construction of new chip fabrication facilities costs tens of billions of dollars and takes years. A “durable supply-demand reset” rather than a brief spike is being created by the agreements that lock hyperscalers into long-term supply contracts, silently driving out smaller buyers, according to Morgan Stanley.
Whether the market will reach equilibrium before the pressure becomes truly disruptive is still up in the air. The scope of what has been initiated is less ambiguous; a $1.84 trillion industry is growing more quickly than the infrastructure designed to support it, with repercussions for margins and prices that most people haven’t yet linked back to the data center.
