Marvell Doubles Down on AI Connectivity with $540 Million XConn Acquisition

January 6, 2026 · by Fintool Agent

MRVLNVDAAVGO

Marvell Technology-0.28% is making its second major AI connectivity acquisition in roughly six weeks, agreeing to pay approximately $540 million for XConn Technologies, a provider of advanced PCIe and CXL switching silicon. The deal, announced Tuesday, adds critical switching capabilities as Marvell races to build an end-to-end connectivity platform for next-generation AI data centers.

The acquisition comes just over a month after Marvell announced it would acquire Celestial AI for $3.25 billion (up to $5.5 billion with earnouts), signaling an aggressive push into the AI infrastructure market where it competes against semiconductor giants Nvidia-0.12% and Broadcom+3.76%. Together, the two deals represent nearly $4 billion in committed capital to AI connectivity in less than two months.

The Deal

Marvell will pay approximately $540 million for XConn in a mix of 60% cash and 40% stock, with the stock portion valued using Marvell's 20-day volume-weighted average price. The stock consideration represents approximately 2.5 million Marvell shares.

The transaction is expected to close in early calendar 2026, subject to customary closing conditions and regulatory approvals. Wilson Sonsini Goodrich & Rosati is serving as Marvell's legal advisor, while Evercore acted as XConn's financial advisor.

Ask Fintool AI Agent

Show me Marvell's acquisition history over the past 5 years and how much they've spent on M&ACompare Marvell's data center revenue growth to Nvidia and Broadcom over the last 8 quartersWhat is the total addressable market for CXL switches and how fast is it growing?

Why XConn Matters

XConn, founded in 2020, has positioned itself as an innovation leader in next-generation interconnect technology. The company was the first to deliver a hybrid switch supporting both CXL (Compute Express Link) and PCIe on a single chip—a critical capability as AI data centers evolve.

The acquisition brings Marvell:

Production-ready products: XConn's PCIe 5 and CXL 2.0 switches are already in production, while its next-generation PCIe 6 and CXL 3.1 switches are currently sampling with customers.

Deep customer relationships: XConn is engaged with more than 20 customers, including multiple hyperscalers planning to use its technology in next-generation scale-up architectures.

Engineering talent: Marvell gains highly experienced engineers with deep domain expertise in high-performance switching to expand its Ultra Accelerator Link (UALink) scale-up switch team.

"This combination creates a compelling switching platform for accelerated infrastructure, advancing Marvell's connectivity strategy for next-generation AI and cloud data centers," said Matt Murphy, Marvell's chairman and CEO. "With XConn, we add proven PCIe and CXL switch products, IP, and engineering talent to expand our UALink scale-up switch team."

Financial Impact

Marvell expects XConn's CXL and PCIe switching products to begin contributing revenue in the second half of fiscal year 2027 (calendar H2 2026), at which point the acquisition will become accretive to non-GAAP earnings. The company projects XConn will ramp to approximately $100 million in revenue by fiscal 2028.

For context, Marvell's data center segment—its largest business—generated $2.07 billion in Q3 FY2026 revenue, growing 37% year-over-year. 4 The company has seen eight consecutive quarters of revenue growth as AI demand accelerates:

Source: S&P Global

The Strategic Picture: Building an AI Connectivity Empire

The XConn deal must be understood in the context of Marvell's broader strategy to become the go-to provider of AI data center connectivity. CEO Matt Murphy has repeatedly emphasized the scale-up opportunity—the high-bandwidth, low-latency interconnects needed to link AI accelerators within racks and across data centers. 5

"As next-generation AI data centers evolve, scale-up networks are becoming essential to tightly interconnect tens, hundreds and eventually thousands of XPUs within and across racks," Murphy explained on the company's August 2025 earnings call. "These require ultra-low latency and multi-terabit bandwidth to meet the demands of training and inference workloads." 6

The back-to-back acquisitions create a portfolio that spans:

• Optical interconnects (Celestial AI): Photonic Fabric technology delivering 16 terabits per second of bandwidth per chiplet for scale-up connectivity
• PCIe/CXL switching (XConn): Production-ready switches supporting both open standard Ethernet and UALink fabrics
• Electro-optics (existing): Market-leading PAM DSPs and DCI products for scale-out networking

Ask Fintool AI Agent

Explain UALink and how it compares to Nvidia's NVLink for connecting AI acceleratorsWhat is CXL 3.0 and why is it important for AI data centers?Show me which hyperscalers are working on custom AI chips that would need these interconnects

Why CXL Is the Next Big Thing

CXL (Compute Express Link) has emerged as a critical technology for next-generation AI infrastructure. The protocol enables memory pooling and disaggregation—allowing multiple CPUs and accelerators to share memory resources efficiently rather than being constrained to fixed, on-board memory.

For AI workloads, this is transformative. Large language models require massive memory capacity for storing model weights, KV caches, and intermediate activations. CXL enables:

• Dynamic memory pooling: Flexibly allocate memory across compute nodes based on workload needs
• Lower latency: CXL delivers sub-microsecond latencies compared to microseconds for traditional Ethernet-based sharing
• Cost efficiency: Avoid over-provisioning memory on every server by sharing pools

XConn's "Ultra IO Transformer" technology is particularly notable—it enables PCIe-based GPUs (which don't natively support CXL) to directly access CXL memory pools. This bridges the gap between current GPU architectures and emerging CXL infrastructure.

Market Reaction and Competitive Landscape

Marvell shares fell 2.2% on Tuesday to $88.23, though this appeared more related to broader semiconductor weakness following Nvidia's CES announcements about new cooling technologies that rattled HVAC stocks. The stock opened at $92.70 before sliding through the session.

Marvell has underperformed its larger rivals over the past year, with shares up roughly 82% compared to Broadcom's 149% gain. The company faces intense competition in AI data center connectivity:

• Nvidia dominates with its proprietary NVLink scale-up fabric and has been expanding into networking with InfiniBand and Ethernet switches
• Broadcom leads in Ethernet switching and has its own co-packaged optics roadmap with the Tomahawk 6 platform

Marvell's strategy is to differentiate through open standards (UALink, Ethernet) and tight integration with hyperscaler custom silicon programs. The company has secured 18+ custom AI design wins and claims multi-billion dollar lifetime revenue potential from these engagements. 1

"We are investing in developing scale-up switches tailored to each customer's protocol of choice," Murphy noted. "Beyond switching, our interconnect portfolio extends the opportunity. While copper dominates the scale-up links today, as networks expand and bandwidth grows, optics adoption will follow." 2

Ask Fintool AI Agent

Compare Marvell's custom silicon business to Broadcom's ASIC offerings for hyperscalersBuild a DCF model for Marvell assuming they hit their AI connectivity revenue targetsWhat portion of Marvell's revenue comes from custom AI silicon vs. standard products?

What to Watch

Near-term catalysts:

• Celestial AI deal closure (expected Q1 2026)
• XConn deal closure (expected early 2026)
• Q3 FY2026 earnings in early December

Longer-term milestones:

• XConn revenue contribution beginning H2 FY2027
• Celestial AI revenue reaching $500M annualized run rate by Q4 FY2028
• Progress on 18+ custom AI silicon programs ramping through 2026-2027

The acquisitions represent a significant bet that the AI infrastructure buildout will require purpose-built connectivity solutions beyond what incumbent players offer. With nearly $4 billion committed to M&A in less than two months, Marvell is clearly playing offense—but execution risk has increased alongside the opportunity.

"Combined with our pending acquisition of Celestial AI, we will be well positioned to deliver customers the performance, flexibility, and architectural choice they need as AI systems grow in size and complexity," Murphy concluded.

Related Companies: Marvell Technology-0.28% · Nvidia-0.12% · Broadcom+3.76%