Marvell’s AI power play: acquisitions, Google deal, and stock soar as optical arms race heats up
The gist
Marvell is muscling into the AI hardware big leagues, turbocharging its optical interconnect game with headline acquisitions and a blockbuster Google chip deal that sent its stock to record highs.
What to know
- Marvell snapped up DustPhotonics and Polariton Technologies to leapfrog optical bandwidth bottlenecks and push interconnect speeds past 3.2 Tbps by early 2026.
- A strategic partnership with Google for custom AI NPUs signals a shift away from Broadcom, expanding Marvell’s chip influence beyond Amazon and Microsoft.
- Investor confidence soared as Marvell joined NVIDIA’s NVLink Fusion ecosystem and notched an 80% YTD stock gain, even as a wave of insider sales hints at cautious optimism.
Optical Arms Race Accelerates
Marvell’s dual acquisitions of DustPhotonics and Polariton Technologies fuse next-gen optical engines and plasmonics, vaulting the company to the forefront of ultra-high-speed data center interconnects and Co-Packaged Optics innovation.
By early 2026, Marvell strategically bolstered its optical interconnect capabilities through two pivotal acquisitions that address the escalating bandwidth bottleneck in AI data centers. The $750 million acquisition of DustPhotonics by Credo, a Marvell subsidiary, vertically integrated advanced optical engine design with its existing DSP expertise, positioning the company to lead the transition toward Co-Packaged Optics (CPO)—a critical system-level integration where optical engines are placed directly next to ASICs to overcome AI infrastructure constraints. Shortly thereafter, Marvell acquired Swiss startup Polariton Technologies, whose cutting-edge plasmonics-based silicon photonics devices and modulation technology are designed to push optical interconnect platforms beyond 3.2 terabits per second, further strengthening Marvell’s commitment to innovation in AI and cloud data-center optical infrastructure by integrating specialized engineering talent and next-generation coherent optical solutions.
Google’s AI Hardware Shakeup
Marvell’s custom NPU deal with Google signals a decisive industry pivot toward specialized, multi-vendor AI chip ecosystems, challenging Broadcom’s dominance and redefining hyperscaler supply chains.
By early 2026, Marvell has notably expanded its footprint in the hyperscaler AI chip ecosystem, securing a pivotal partnership with Google to develop custom neural processing units (NPUs). This collaboration marks a strategic shift away from Broadcom, highlighting Marvell's growing influence beyond the traditional top-tier hyperscalers like Amazon and Microsoft. The diversification of Google's AI inference hardware stack, which now includes players such as Mediatek alongside Marvell, underscores a broader industry trend toward multiple specialized suppliers rather than a one-size-fits-all approach.
Marvell’s competitive edge in these partnerships stems from its robust intellectual property portfolio focused on memory connectivity, advanced packaging, and interconnect I/O technologies. These capabilities complement Google's TPU architecture by enhancing data flow efficiency and integration, which not only optimizes performance for next-generation AI workloads but also supports higher gross margins. As analysts note, Marvell’s value-add lies beyond core digital logic, positioning it as a critical enabler in the evolving AI chip supply chain.
Market Mania Meets Caution
Soaring to record highs on AI momentum and NVIDIA partnerships, Marvell’s stock faces insider selling and macro headwinds, highlighting a tension between investor euphoria and executive wariness.
By early 2026, investor sentiment around Marvell surged as the company expanded its footprint in AI infrastructure, notably with Google's strategic pivot from Broadcom to Marvell for next-generation AI chips. Bloomberg Intelligence's Kunjan Sobhani highlighted Marvell's competitive edge through its specialized IP portfolio in memory connectivity, packaging, and interconnects, which complements Google's TPU designs and differentiates it from rivals. This focus on NPUs within the XPU portfolio, crucial for data movement in inference workloads and commanding higher gross margins, positioned Marvell as a key player beyond just the top hyperscalers, signaling broadening market confidence.
Marvell's stock performance mirrored this growing investor enthusiasm, reaching an all-time high above $170 in early May 2026, fueled by excitement over its integration into NVIDIA's NVLink Fusion ecosystem. This partnership underscored a strategic industry shift from compute-centric to connectivity-focused AI data centers, with Marvell's strengths in photonics and custom silicon placing it at the heart of optical interconnects and AI scale-up fabrics. However, despite this bullish momentum and widespread social media chatter, insider trading activity revealed a cautious stance, with 13 insider sales and no purchases over six months, suggesting nuanced internal perspectives on valuation.
Despite Marvell’s impressive 80.3% year-to-date gain by mid-May 2026, the stock experienced a 6% pullback following a hotter-than-expected April CPI report, reflecting broader investor concerns about rising Treasury yields and the diminishing likelihood of rate cuts in 2026. These macroeconomic headwinds have pressured semiconductor valuations industry-wide by increasing discount rates on future earnings and compressing price-to-earnings multiples. Nonetheless, Marvell’s resilience near its 52-week high amid volatile market conditions underscores sustained confidence in its AI infrastructure strategy, even as the sector navigates economic uncertainties and competitive dynamics.
