Nvidia’s Wednesday-night earnings call delivered the headline the market had been waiting for: $96.2 billion in Q2 revenue, up 106% year over year, followed by a Q3 guide of $108 billion, plus or minus 2%. Cloud industry backlog now sits above $2 trillion, and management confirmed Vera Rubin is ramping into full production, with first shipments expected in Q3 and a volume ramp through Q4 and into 2027. The Vera CPU simultaneously opens what management has described as a roughly $200 billion CPU addressable market.
But buried inside that triumphant guide is a confession: Nvidia told investors its gross margin bottoms at 71-72% in Q4 because component costs have risen significantly. Samsung and SK Hynix reportedly raised HBM3E pricing by nearly 20% for 2026 deliveries, an unusual move for a generation that would normally be getting cheaper. The direction is not debatable: memory input costs are rising while Nvidia is putting more memory into each new platform. Every dollar compressing Nvidia’s margin this quarter is revenue flowing somewhere upstream. The real question for long-term investors is who in that chain has genuine pricing authority, and who is simply renting the boom.
The Component Monopolists
Three businesses hold positions that look structurally irreplaceable at this moment in the cycle.
Memory (Micron). Only three companies, SK Hynix, Samsung, and Micron, produce HBM, and capacity expansion requires meaningful lead time. That oligopoly structure has produced pricing behavior rarely seen in a business historically defined by oversupply. Micron said it is in high-volume production of HBM4 designed for NVIDIA Vera Rubin. And each Rubin GPU is expected to ship with up to 288GB of HBM4. More memory per platform times rising per-unit prices is a compounding revenue tailwind, not a one-quarter event.
Foundry and packaging (TSMC). TSMC holds a dominant position in leading-edge chip fabrication, and customers can be locked in for long qualification cycles when they choose advanced nodes. The advanced packaging side is particularly significant: industry reporting continues to describe CoWoS as tight, with bookings extending into 2027 for parts of the supply chain. Nvidia’s Rubin GPU has been widely reported as a 3nm-class TSMC product paired with HBM4 memory, which means every Vera Rubin rack that ships depends on both TSMC’s leading-edge logic and advanced packaging capacity.
Power and cooling (Vertiv). The Vera Rubin NVL72 rack is fully liquid cooled, and that is not a preference, it is a physics requirement. Vertiv is benefiting from significant operating margin expansion, reporting an adjusted operating margin of 20.8% in Q1 2026, up 430 basis points year over year and 180 basis points above guidance. Management raised full-year guidance, with adjusted operating margin now expected at 23.3% at the midpoint, and it has said it expects to remain price/cost positive for the year, including mitigation of current tariff impacts through countermeasures. Vertiv passes costs through; it does not absorb them.
The Boom Renters
The distinction matters most for Celestica. Celestica sits at the final integration layer. While enterprises buy servers from Dell or HPE, hyperscalers design their own and hire Celestica to build them, integrating the GPU, custom silicon, and networking gear into a finished, tested rack. Revenue is growing rapidly, and management previously raised 2026 revenue guidance to $17 billion. But contract manufacturing economics are inherently thinner than component monopoly economics. Volume is real; the durability of the margin advantage depends on how much proprietary engineering content Celestica can embed over time.
Amphenol, which has been tied to high-speed interconnect and signal integrity content in AI systems, has reported exceptional growth in IT datacom. That custom-engineering involvement is the distinguishing factor: where Amphenol has designed itself into a specific architecture rather than simply assembling commodity connectors, it holds something closer to a component monopolist’s position. Arista’s non-GAAP operating margin came in at 49.9% in Q2 2026, reflecting a networking business where software and differentiated silicon command real pricing.
The Screening Principle
Nvidia CFO Colette Kress said on Wednesday’s call that executed price increases will land in FY28 Q1, allowing margins to recover. That timeline is the tell. A business that must wait for a future price hike to recapture margin has already demonstrated that its input suppliers have leverage. The component monopolists, those controlling qualified HBM capacity, advanced packaging, and the power infrastructure that liquid-cooled AI racks cannot function without, are collecting that leverage as revenue today. Investors willing to look one step upstream from the headline number will find the durable compounders that Nvidia’s own margin trajectory is pointing toward.
