Celestica's revenue jumped 62% last quarter as its role in AI infrastructure and AMD's Helios platform expanded rapidly.
Growing from nearly nothing, Marvell's custom AI chip business now makes up roughly a quarter of its data center revenue.
Celestica offers steadier infrastructure exposure, while Marvell carries higher risk and higher reward potential tied to demand for custom chips.
Palantir (NASDAQ: PLTR) has become one of the biggest names in artificial intelligence (AI), but the boom is still in its early innings, and there's plenty of time for new leaders to rise. Over the next five years, the biggest winners may be the companies quietly building the infrastructure that makes the AI revolution possible. Celestica (NYSE: CLS) and Marvell Technology (NASDAQ: MRVL) fit that description, thanks to their cutting-edge tech.
If the physical end of the AI boom continues to shift toward entire AI factories, Celestica could be sitting in one of the most important parts of the supply chain. That's because the company is increasingly involved in designing and manufacturing the data center infrastructure that connects AI compute.
Missed AI’s "Act 1"? Act 2 Could Be 15x Bigger. Most investors think they missed the AI boat because they didn't buy Nvidia in 2005. But according to our analysts, we’re only at the end of "Act 1"—the R&D phase. "Act 2" is the global rollout. Continue »
That transformation is already showing up in the business. In the second quarter, Celestica's revenue jumped 62% year over year, and management raised its full-year outlook. It also said it expects its growth to accelerate in 2027, and that adjusted earnings growth would outpace revenue growth. Management pointed to new program wins and improving customer forecasts as drivers of that upbeat guidance.
Plus, it's not just about selling more servers. Celestica is moving deeper into advanced AI systems. For example, the company announced a collaboration with AMD to bring the Helios rack-scale AI platform to market, with Celestica responsible for the research and development (R&D), design, and manufacturing of scale-up networking switches that connect clusters of AMD's AI accelerators.
That matters because AI infrastructure is becoming even more complex. As successive generations of processors increase their compute power, the challenge becomes more about moving data between accelerators, memory, and different parts of the data center. Celestica can help solve those connectivity issues and benefit no matter what type of AI chips a data center operator is using.
Celestica is also attracting attention because of its relationships with major AI players. For example, the company has worked with Google on its custom processors and with OpenAI on its next-generation AI systems.
Marvell, on the other hand, is tackling the same opportunity from a different angle: designing custom chips tailored to specific AI workloads.
Image source: Getty Images.
Marvell has made a name for itself as one of the biggest suppliers to AI giants, especially as hyperscalers look beyond off-the-shelf chips to more customized AI infrastructure. In just a few years, its custom silicon business has grown from essentially nothing to roughly a quarter of its data center revenue, while its optical interconnect segment has grown rapidly as AI systems require more bandwidth.
Meanwhile, management said that its AI-related bookings were exceptionally strong and that demand was broad across high-speed optics, Ethernet switching, and data center interconnect. That means Marvell isn't betting on the success of a single software platform like Palantir, but is benefiting from several layers of the physical AI build-out, including accelerators and memory interfaces, networking, optical connectivity, and advanced packaging.
It even expanded its relationship with Google, extending its commercial agreement to cover custom silicon integrated into Google's Tensor Processing Unit (TPU) ecosystem, including AI inference chips, storage controllers, and near-memory compute.
Then there's its relationship with Nvidia. Marvell Technology and Nvidia announced a partnership around NVLink Fusion, giving customers building on Nvidia's infrastructure access to Marvell's custom chips and scale-up networking. As a sweetener, Nvidia invested $2 billion in Marvell, too.
If AI infrastructure continues to become more customized, Marvell could be one of the biggest winners in the next phase of the AI race.
Market cap tells the real story here. Palantir sits at around $462 billion today, Marvell at about $252 billion, and Celestica at about $45 billion. That means Marvell would need to almost double its market cap to catch up to the data analytics giant, while Celestica would need roughly tenfold growth, a much taller order by any means.
Let's assume Palantir's earnings multiple keeps compressing toward 20, roughly where its own forward estimates already point. Its earnings growth would largely offset that compression, leaving its market cap near $460 billion, where it sits today, rather than pushing it higher.
For Marvell, EPS is projected to climb more than sixfold from $4.22 in its fiscal 2027, which just began, to $27.77 by its fiscal 2031. That means even with a modest rerating downward, let alone a flat earnings multiple, the stock could more than double based on earnings growth alone.
But for Celestica, a lot more would need to go right. Consensus EPS estimates climb from $11.21 this year to north of $41 by 2030, but for the stock to 10x, its growth would have to keep compounding, and the stock's P/E multiple would have to expand rather than contract the way a static forward table implies.
Marvell's growth trajectory could easily turn it into a larger company than Palantir in five years. Celestica can reach that size too, but it will need both its earnings growth and its P/E multiple to break in its favor.
Before you buy stock in Celestica, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Celestica wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $370,440!* Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,470,022!*
Now, it’s worth noting Stock Advisor’s total average return is 955% — a market-crushing outperformance compared to 215% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
See the 10 stocks »
*Stock Advisor returns as of October 8, 2026.
Rick Orford has positions in Alphabet. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Celestica, Marvell Technology, Nvidia, and Palantir Technologies. The Motley Fool has a disclosure policy.