Jabil's third hyperscaler win is the most important manufacturing story of 2026
Count that. Amazon, Google, Microsoft, Meta (the hyperscalers building the physical AI world) and now Jabil has three of them paying for manufacturing services. That is not a diversification story. That is a moat forming in real time.
Here is the chain of logic the market is probably not running. Hyperscalers are building at a pace where their own supply chains are becoming the constraint. Custom silicon, liquid cooling assemblies, high-density power distribution hardware. None of it gets made at the scale they need by sitting on a single contract manufacturer. So they add a second, a third. Each new win for Jabil means a competitor gets squeezed out of that slot. And once a hyperscaler certifies your facility and integrates your quality management into their build cycle, switching costs are brutal. These are not annual contracts you can renegotiate. They are multi-year entanglements.
CEO Mike Destor ran the earnings call with the controlled confidence of someone who knows the numbers are compounding and is trying not to say too much before September's investor day. His prepared remarks were clean, zero fillers, deliberate pace. Q&A was different. Filler density tripled when analysts pushed on FY27 margin shape and the Adani alliance in India. That kind of vocal shift usually means the executive knows more than they can say, not that they know less.
The Adani discussion is where the second-order story lives. Jabil is positioning itself as the manufacturing operating system for India's AI infrastructure buildout. India is running a parallel AI arms race. Adani's data center ambitions alone would require the kind of manufacturing depth that almost no company outside China currently has at scale in that geography. If Jabil gets that framework signed, the FY28 revenue addition would not fit inside the current analyst model.
The second-order read for investors watching adjacent companies: whoever supplies liquid cooling components, high-density interconnect PCBs, and power distribution hardware into the India AI buildout is about to have a very interesting few years. Jabil's win validates that the geography is real and the capital is moving.
The bear case is worth naming honestly. Destor deferred specifics on margin shape, customer concentration risk, and Adani structure to September. That is a reasonable choice, but it means you are buying a thesis that has not been fully argued yet. Leverage is slightly elevated. And a third hyperscaler win, while directionally great, is still only three. If one of them pulls volume in a down-capex cycle, the revenue concentration hits differently.
But if you believe the AI buildout is a decade-long capital cycle, and I do, then what Jabil is building in customer relationships right now is the kind of foundation that takes a competitor five years to replicate. The market is pricing it as a cyclical manufacturer. That gap is the opportunity.
Three more stories worth knowing about
CarMax's pricing algorithm is eating its own lenders
SpaceX is public. Three stocks reprice next.
Anthropic's IPO will reprice AI-native software before it lists
Credo Technology Group Holding Ltd
The setup is straightforward but underappreciated. As hyperscalers push rack densities higher and switch from copper to higher-bandwidth interconnects, the components Credo makes go from optional to mandatory. HiWire active electrical cables in particular solve a power-per-bit problem that passive copper cannot. The customer list includes hyperscalers and major OEMs. Revenue has been growing fast.
What makes it worth a deeper look rather than a quick pass: Credo is at the intersection of the bandwidth explosion and the power constraint. Those two forces are not going away. The stock is not cheap, so this is not a value pitch. But for anyone building a position in the physical AI infrastructure layer, Credo is a name worth understanding before the next earnings cycle.
The Golden Dome defence stocks trade is backwards
The more interesting position is that the Golden Dome buildout, if it actually happens at the scale Washington is describing, requires a massive volume of lower-cost autonomous and semi-autonomous systems to make layered defence economically viable. Interceptors at $3 million each cannot be the only answer to a $50,000 drone swarm. That arithmetic pushes the program toward exactly the kind of attritable, distributed hardware that Kratos (NASDAQ:KTOS) builds.
The incumbents get the headline contracts. Kratos gets the units. If the program scales the way proponents claim, the relative repricing in KTOS versus LMT is likely to be larger, not smaller. That is where the asymmetry lives, not in the names everyone is already buying.