
Arm, IREN, and Nvidia are highlighted as long-term beneficiaries of the AI data center build-out, with Morgan Stanley estimating nearly $3 trillion in new global data center construction costs through 2028. Arm reported record quarterly revenue of $1.49 billion, up 20% year over year, while management expects it to eventually hold the largest share of data center CPUs by 2030. IREN is cited for 385% share gains, $18 billion market cap, 5 GW of grid-connected power, and expected $4.4 billion in annualized revenue by year-end 2026; Nvidia’s data center revenue nearly doubled last quarter and analysts expect total revenue to rise 81% to $391 billion this year.
The key second-order takeaway is that the AI buildout is shifting from a compute-only trade to a power-and-architecture bottleneck trade. If capex keeps migrating toward energy-efficient CPU/accelerator stacks and vertically integrated data-center operators, the real winners are the toll collectors on scarce inputs: IP licensors, grid-connected power owners, and system integrators with execution leverage. That argues for persistent multiple support in ARM and NVDA, but an even larger relative rerating potential in names like IREN if management can keep converting power access into contracted utilization.
The market may still be underestimating how fast hyperscalers will diversify away from x86 in power-constrained workloads. ARM does not need to win every socket to matter; it only needs to keep gaining share in high-density deployments where watts per unit of compute becomes a board-level KPI. The bigger risk for Intel is not share loss in legacy servers, but being relegated to lower-growth, lower-margin workload tiers while ARM captures the expanding edge of AI-infrastructure demand.
For IREN, the core debate is not whether demand exists, but whether execution can keep up with a multi-year backlog of power conversion, permitting, and buildout. The stock can continue to outperform if management proves it can monetize the power portfolio faster than peers, but the equity remains highly sensitive to delays, financing costs, or any cooling in colocated/AI demand. In other words, this is a duration trade on optionality, not just a simple growth story.
Nvidia remains the highest-quality expression of the theme, but the setup is less asymmetric than six months ago because expectations are now tied to platform transitions rather than a one-time GPU cycle. The contrarian point is that the market may be underpricing how much value accrues to the broader AI stack as customers seek lower total cost of ownership: as utilization rises, buyers will increasingly optimize system-level efficiency, which can widen the addressable market for ARM-based CPU layers and non-Nvidia infrastructure plays. That said, any pause in enterprise spend or supply-chain normalization could compress the whole basket quickly because these names are now crowded consensus longs.
AI-powered research, real-time alerts, and portfolio analytics for institutional investors.
Request DemoOverall Sentiment
strongly positive
Sentiment Score
0.72
Ticker Sentiment