AI capex – Big Tech’s AI spending headlines may overstate real capacity gains, as soaring memory prices inflate data center capex.
Big Tech’s AI spending boom looks massive on paper, but Misryoum says a large part of the headline capex surge may be driven by pricier hardware rather than a faster expansion of compute capacity.
When major AI players discuss higher infrastructure spending, they often point to the same bottleneck: components.. Misryoum notes that Meta’s leadership tied much of its infrastructure capex increase to higher memory costs. while Microsoft’s CFO described a significant slice of upcoming capex as stemming from elevated component prices.. Even where companies did not raise forecasts in a given week. Misryoum says their executives still highlighted how memory costs have escalated sharply.
This distinction matters because it changes how investors should interpret “more spending” versus “more output.” If the bill rises mainly because chips cost more, the business may be paying extra without necessarily deploying proportionally more equipment.
In the background is a supply-and-demand mismatch that has turned memory into a critical constraint for data center buildouts. Misryoum reports that expectations for both DRAM and NAND flash pricing have been moving higher, reflecting AI demand straining available supply.
A simple way to understand the impact is to separate quantity from price.. Misryoum describes a scenario where buying the same number of AI components still requires a larger capex payment if each unit becomes more expensive.. In that case, the total spend rises even though the company has not expanded capacity by the same magnitude.
For investors. Misryoum says the challenge is that headline capex growth can blend two different effects: genuine acceleration in infrastructure expansion and an inflationary effect from component pricing.. Untangling those drivers can clarify whether the AI arms race is truly speeding up or simply getting more expensive.
The takeaway is a more nuanced picture of how Big Tech is scaling AI. Misryoum says the core investment story remains strong, but the pace suggested by large capex numbers may be overstated if a substantial portion of the increase reflects a premium paid for scarce memory components.
At the same time, Misryoum argues this is also a signal of what comes next: if memory pricing remains volatile, the next phase of growth may be governed as much by semiconductor supply conditions as by software and model demand.