Three AI infrastructure companies, Lumentum, Jabil, and Vertiv, are drawing serious analyst attention as the physical buildout of AI data centers accelerates into 2026. Vertiv’s numbers are the most concrete: roughly $10.2 billion in FY2025 revenue, a $15 billion backlog tied directly to AI data center projects, and 14.4% net margins backed by partnerships with NVIDIA. Jabil’s $15.8 billion intelligent infrastructure segment and Lumentum’s optical components for hyperscale networks round out the picture. The full analysis of these AI infrastructure stocks<\/a> surfaces both the growth cases and the balance sheet risks attached to each.<\/p>\n The companies your AI infrastructure depends on are themselves highly leveraged bets on continued hyperscaler capex. All three carry significant debt loads, which means their capacity to invest in next-generation capabilities is tied to the same AI spending cycle they’re riding. If you’re in contract negotiations or procurement planning for data center expansion, the financial fragility of key suppliers isn’t abstract, it’s a vendor concentration and supply continuity risk sitting inside your own roadmap.<\/p>\n Vertiv’s $15 billion backlog is the most operationally significant number in this piece, and not primarily for the reason investors care about it. A backlog that size, concentrated among a handful of hyperscalers, means Vertiv’s production capacity and engineering attention are largely pre-committed. If your organization is not already in that queue, delivery timelines for power and liquid cooling infrastructure, the thermal management systems that make high-density GPU clusters physically operable, are likely to stretch further than your current planning assumptions reflect. The companies that locked in Vertiv contracts in 2024 and early 2025 made a sequencing decision that looks smarter every quarter.<\/p>\n Lumentum’s transition to profitability alongside a high price-to-earnings ratio tells a specific story about optical interconnects, the fiber-based components that move data between chips and servers inside a data center at speeds electrical wiring can’t match. The valuation implies the market expects photonics to become a genuine bottleneck as GPU cluster density increases, which is plausible. But it also means the optical supply chain is now priced for perfection at exactly the moment new entrants and in-house development programs at hyperscalers could compress margins. A single missed earnings quarter from Lumentum would be an early signal that the optical bottleneck is resolving faster than expected, and worth watching as a leading indicator of broader AI infrastructure spend normalization.<\/p>\n The real budget implication here isn’t whether to buy these stocks. It’s that the infrastructure layer powering AI has moved from a commodity procurement category into a strategic constraint, and the financial pressure on the companies supplying it creates real delivery and pricing risk through at least 2027. I’d revise that view if hyperscaler capex guidance in Q3 and Q4 2026 shows meaningful deceleration, which would ease backlog pressure and shift negotiating leverage back toward buyers.<\/p>\nWhat this means for your business<\/h2>\n