JBL - one of the more interesting picks and shovels plays on the AI infrastructure buildout

JBL has been in my portfolio for about a month and wanted to share it is one of the more interesting picks and shovels plays on the AI infrastructure buildout.

TLDR - AI data centers are no longer just about GPUs as the demand is shifting toward power, liquid cooling, rack-scale deployment, networking bandwidth, and optical interconnects. Jabil sits right in the middle of those physical infrastructure problems.

The main segment is Intelligent Infrastructure which includes AI infrastructure, cloud/data-center systems, networking, communications, and capital equipment.

It grew 52% YoY in Q2 FY26 and is expected to reach about $16.5B in FY26, up roughly 34%. Management is also guiding to around $13.1B of AI-related revenue in FY26, up roughly 46%.

One thing I think is underappreciated is that growth is happening even while parts of the legacy business are weak. Jabil is guiding to about $34B of FY26 revenue, up roughly 14%, even though Connected Living & Digital Commerce is expected to decline. That tells us the AI infrastructure ramp is strong enough to drive the whole company despite the drag.

The reason JBL has high potential is that AI data centers are getting harder to build. It is not just about GPUs.

The bottlenecks are increasingly around power, cooling, rack-level integration, networking, optics, and physical deployment at scale. Jabil sits in the middle of those problems.

Their acquisitions also fit the strategy.

\- Mikros adds liquid-cooling capability.

\- Hanley Energy adds power-management and data-center energy infrastructure exposure.

Those are exactly the kinds of areas that matter as AI racks get denser and more power-hungry.

The company is guiding to roughly:

$34B FY26 revenue

$12.25 core EPS

\>$1.3B adjusted free cash flow

\~5.7% core operating margin

potential path toward 6%+ margins

The margin point is v important. Intelligent Infrastructure did 5.7% core margin in Q2, compared with 5.3% for the company overall. So the fastest-growing segment is already margin-accretive.

If Intelligent Infrastructure keeps becoming a larger part of the company and consolidated margins move toward 6%+ the EPS leverage can be v impactful.

Also worth noting is that this is not just revenue growth - share count is coming down, free cash flow is strong, and inventory metrics improved YoY.

The other point is that JBL’s AI exposure is broader than just server builds.

Management says AI-related revenue comes from portions of capital equipment, cloud/data-center infrastructure, and networking. Cloud & Data Center Infrastructure is expected to grow strongly in FY26, but Networking & Comms is also expected to grow meaningfully.

I am including this because the CPO/silicon photonics opportunity is ultimately about AI networking bandwidth and power constraints.

Now for the very important part - the CPO / silicon photonics.

CPO stands for co-packaged optics. In simple terms, it means moving optical interconnect technology much closer to the networking chip instead of relying only on traditional pluggable optical modules.

The goal is to move data using light more efficiently, with lower power, higher bandwidth density, and less signal loss.

This is v imporatnt because AI clusters need insane amounts of data moving between GPUs, switches, and racks.

Traditional copper//electrical connections become more power hungry and harder to scale as speeds move to 1.6T, 3.2T, and beyond.

CPO is one possible answer to that problem.

Tradtional architecture looks something like this:

Switch chip → electrical traces → pluggable optical module → fiber

CPO moves toward:

Switch chip + optical engine closer together → fiber

That means lower power, higher bandwidth density, and better performance for massive AI clusters.

Goldlman Sachs called optical networking “the next mega trend in AI infrastructure,” with the total optical networking TAM potentially expanding from around $15B to $154B.

That is exactly where the bottleneck is emerging - as AI systems scale from single racks to multi-rack GPU clusters the constraint shifts from just compute to moving data between GPUs fast enough and efficiently enough.

Copper starts running into power, heat, and distance limits and optical interconnects, silicon photonics, CPO, external laser sources, and optical circuit switches are becoming more important.

NVIDIA’s behavior also confirms the direction as they have committed billions into the optical ecosystem, including Lumentum, Coherent, and Marvell.

That is a pretty clear signal that the next leg of AI infrastructure is not just more GPUs but networking, optics, power efficiency, and rack-scale architecture.

Thisis where JBL gets interesting.

Jabil may not be the core IP owner in CPO, but it plays a very valuable role in manufacturing, integration, testing, optical assemblies, external laser source systems, rack-level deployment, and scaling the physical infrastructure.

JBL has already shown ecosystem involvement with names like Ayar Labs, Sivers, and Marvell.

Jabil and Ayar have demonstrated external laser source arrays for co-packaged optics.

Sivers is tied into Ayar’s laser/photonics supply chain. Jabil has also announced work with Sivers on 1.6T optical transceiver technology.

I would not yet model huge CPO revenue into JBL because it is still early but I think the market may not be fully appreciating the optionality here.

If CPO adoption accelerates into the H2 2026 / 2027 AI networking cycle JBL could end up more embedded in the ecosystem than people expect.

Of course this rmeains a risky setup because the stock is not as cheap as it used to be, hyperscaler capex can be cyclical, and CPO may take longer than expected

But I think the positive case is that JBL is becoming a v critical AI infrastructure integration platform at the exact moment the AI buildout is moving from chips to full systems.

Not financial advice, do your own research. Currently holding long JBL.