3 stocks profiting from Google 2026 Capex spending

Alphabet (GOOGL) just put a very loud number on the table for 2026.

On February 4, 2026 the company said capital spending could “as much as double” as it ramps AI compute and data center capacity.

Other reports pegged the 2026 capex outlook at roughly $175 billion to $185 billion, well above what Wall Street had estimated.

That kind of capex isn’t abstract. It will turn into purchase orders for silicon, networking gear, and the unglamorous infrastructure that keeps megawatt-scale GPU and TPU clusters alive. If you want to know “who gets paid” during Google’s 2026 buildout, here are three stocks with the most direct line of sight.

1. Broadcom (AVGO), the pick and shovel behind Google’s custom AI silicon.

If Alphabet is going to spend like that, a big chunk flows into accelerators, and not all of them are Nvidia.

Google’s competitive angle is its TPU roadmap.

Broadcom makes next-gen custom AI chips for hyperscalers that are widely considered to include Google.

Why this matters in 2026. Custom silicon isn’t a one-time event. Each generation means years of design services, IP, packaging choices, memory interfaces, and then a volume ramp through foundries and advanced packaging. A capex jump tends to show up as “more wafers, more networking, more power,” and Broadcom sits right in the “more wafers” lane for Google’s in-house accelerators.

The disruption angle.

Hyperscalers are trying to break the Nvidia tax by owning more of the stack. Broadcom is one of the few companies that can reliably co-design these data center grade chips at scale. That’s a bottleneck skill set, and bottlenecks get priced in.

Watch-outs.

Google has explored additional partners for TPU development, which can dilute exclusivity. Google was preparing to partner with MediaTek on a next TPU generation while not ending the Broadcom relationship. So the bull case is “bigger pie,” not “sole supplier.”

2. Arista Networks (ANET), because AI clusters are networking problems in disguise.

Once you buy the compute, you still have to wire it into something that doesn’t melt down under training traffic.

AI training is brutal on networks.

It’s a constant firehose of data moving between accelerators. That’s why 400G and 800G Ethernet switching has become a core part of AI capex, not an afterthought.

Arista is one of the cleanest public-market ways to play that. It’s built its business selling high-performance data center switching to large cloud operators, and it’s explicitly positioned around AI data center networking.

Why tie it to Google specifically. Even if Google mixes vendors, it’s still an Ethernet buyer at massive scale. Nvidia itself highlighted Google as a new customer for its Spectrum-X Ethernet networking push, which is useful signal that Google’s AI networking spend is real and rising. When a hyperscaler scales AI clusters, it can’t avoid buying a lot of switches, optics, and network software.

The disruption angle.

There’s a real architectural fight happening. Nvidia wants to bundle accelerators plus networking into a tight ecosystem. Ethernet vendors like Arista benefit when hyperscalers keep networks more open and mix-and-match. If Google keeps leaning into open Ethernet fabrics, Arista has a path to win spend even in a noisy competitive field.

Watch-outs.

Competition is not theoretical. Nvidia is pushing hard into Ethernet, and that can pressure pricing and share. So you’re paying for execution, not just “AI happens.”

3. Vertiv (VRT), the “boring” infrastructure that becomes critical at Google scale.

When capex jumps, people focus on chips.

But data center builds are increasingly constrained by power delivery and heat removal.

That’s where Vertiv shows up.

It sells the gear that keeps compute running, including power management and thermal systems, and it has been expanding deeper into liquid cooling as AI rack densities rise. Vertiv’s move to strengthen its liquid cooling services portfolio, is explicitly tied to AI-driven power strain on data centers.

The Google link is unusually concrete here. Vertiv publicly points to work connected to Google’s Open Compute Project efforts, including Google’s “Project Deschutes CDU” for liquid-cooled systems. That matters because OCP designs tend to propagate. When a hyperscaler standardizes a cooling approach, suppliers that are already inside that ecosystem can ride the standard into broader deployments.

The disruption angle.

AI is turning data centers into energy infrastructure. The winners aren’t only model makers. They’re the companies that can ship reliable power and cooling systems fast, at hyperscale volumes, with service capacity. If Alphabet is really pushing 2026 capex into the $175 billion to $185 billion range, the enabling constraint is increasingly physical, not digital.

Watch-outs.

Vertiv is more economically sensitive than chips. If build schedules slip, revenue can also arrive later.

The clean way to think about this trade

Google’s 2026 capex surge is a supply chain event.

Broadcom maps to custom AI silicon. Arista maps to the networks that make clusters usable. And Vertiv maps to power and cooling that make clusters possible. If you buy only one idea, buy this. AI capex turns into bottleneck spending. And bottlenecks tend to be where the profit pools in disruptive technologies like AI form.

Google (GOOGL) has a Disruption Score of 4. Click here to learn how we calculate the Disruption Score. 

Google is also part of the Disruption Aristocrats, our quarterly list of the world’s top disruptive stocks.