Celestica poised for growth as Google and OpenAI push data center demand

The AI boom gets described in terms of chips. Bernstein thinks the more interesting story is the plumbing that connects them.

The research firm argues that a critical bottleneck inside data centers is opening up a large growth runway for networking stocks. It names Celestica as a prime example, with an Outperform rating and a $520 price target.

What Celestica actually builds

Celestica, which trades as CLS on both the NYSE and the TSX, is a Toronto-based electronics manufacturing services provider.

The company supplies custom data center hardware to hyperscalers. That includes servers and Ethernet switches running at 400G, 800G and 1.6T speeds.

Google has identified Celestica as a primary collaborator on its Tensor Processing Unit systems and upgraded networking. TPUs are Google’s in-house AI chips, its alternative to buying everything from outside vendors.

The partnership comes as Celestica expands US manufacturing capacity, specifically in Texas. Its long-term projects are closely tied to Google’s AI buildout.

OpenAI is the other name in the frame. Growing demand from OpenAI and its large-scale projects adds potential upside as Celestica ramps production on its 1.6T switch programs.

The numbers behind the bull case

Celestica forecasts roughly $20.5 billion in revenue for 2026. That would mark year-over-year growth of 65%.

The engine is the Connectivity & Cloud Solutions segment, which has posted triple-digit growth. That is the part of the business tied to high-bandwidth networking.

Projected data center capex across Alphabet, Amazon, Meta and Microsoft is expected to reach $650 billion or higher in 2026.

Celestica recently completed a $3 billion equity offering aimed at fueling growth. Capacity expansions are internally funded as well.

It has also restructured its leadership to support operations at larger scale.

How Wall Street sees it

More than 22 analyst firms rate Celestica a Buy or better, with an average 12-month price target of around $447.

Bernstein’s $520 target sits well above that consensus.

The stock has already delivered. Shares have gained more than 10x over two years, with increases of more than 100% within a single year.

Why networking became the choke point

Training large AI models requires enormous clusters of processors exchanging data constantly. If the network cannot keep up, expensive chips sit idle.

That dynamic is pushing hyperscalers to upgrade from 400G to 800G, and now toward 1.6T switching. Celestica has positioned itself to supply each of those steps, often as a custom builder for a specific customer rather than a seller of off-the-shelf gear.

What this means for investors and the AI supply chain

The core bet on Celestica is really a bet on hyperscaler spending staying elevated. The company’s growth forecast leans heavily on a small group of very large customers, led by Google.

Google’s TPU program is a signal that hyperscalers want more control over their hardware stacks. Firms that can build bespoke systems at scale, rather than just reselling standard components, stand to benefit from that trend.

The $3 billion equity raise gives Celestica fuel for capacity, including the Texas expansion. It also dilutes existing shareholders, a trade the market has so far tolerated because growth has outpaced it.

Ramping 1.6T switch production is technically demanding. Manufacturing delays or yield problems on a new product generation could dent the 65% growth forecast.

The gap between Bernstein’s $520 target and the roughly $447 consensus average shows analysts agree on direction but not on how much upside remains.

The key things to watch are hyperscaler capex updates, progress on the 1.6T ramp, and whether demand from OpenAI turns into meaningful new business.