What's happening
Marvell Technology has confirmed it is in active discussions with Taiwan Semiconductor Manufacturing Company (TSMC) to adopt the foundry's A14 (1.4nm) process node for future AI connectivity chip generations. Chris Koopmans, Marvell's president and chief operating officer, stated that 'the company has already begun discussions with TSMC regarding the use of its A14 node for future-generation products,' according to Nikkei Asia as reported by TrendForce on June 18, 2026. TSMC's A14 process is currently scheduled to enter volume production in 2028, meaning any products built on the node would represent a multi-year development horizon. To underpin its access to leading-edge capacity, Marvell has committed $1 billion in prepayments to secure future manufacturing slots.
The disclosure positions Marvell among a narrow group of fabless semiconductor companies pursuing the most advanced process nodes available from TSMC, a tier historically occupied by the largest chip designers. Marvell's process technology trajectory has been notably aggressive: the company bypassed the 7nm node entirely, moving directly from 16nm to 5nm. The A14 discussions represent a continuation of that strategy, with Marvell targeting the connectivity and data infrastructure segments of the AI supply chain rather than the GPU and accelerator markets dominated by larger rivals.
Why it matters for markets
Marvell's transformation into a data center-centric semiconductor company is reflected in its revenue composition. Data center revenue stood at under $200 million — representing less than 10% of total revenue — as recently as 2016. That figure now exceeds 75% of total revenue, which reached $8.72 billion on a trailing basis according to Marvell's ticker profile. The A14 discussions signal that Marvell is making long-duration capital commitments — including the $1 billion in prepayments — to sustain its position at the leading edge of process technology as AI infrastructure demand continues to expand. Accessing the A14 node would allow Marvell to offer customers connectivity chips with the power efficiency and density improvements that advanced nodes provide, a competitive consideration as hyperscalers push for higher-bandwidth, lower-latency interconnects.
For TSMC, engagement from a company of Marvell's scale in discussions around A14 capacity is relevant to the foundry's long-term utilization planning for its most advanced nodes. TSMC's revenue reached $4.10 trillion (in New Taiwan dollars) on a trailing basis, and its A14 ramp in 2028 will require a broad customer base to achieve economically viable volume. Marvell's participation, alongside the established presence of customers such as NVIDIA and AMD at leading nodes, adds another demand signal for TSMC's next-generation capacity. NVIDIA's strategic relationship with Marvell is also notable in this context: NVIDIA has invested $2 billion in Marvell, creating a financial linkage between the AI accelerator leader and a key connectivity chip supplier.
The competitive landscape in AI connectivity is also evolving at the process technology level. Broadcom, a direct competitor to Marvell in optical and networking silicon, began sampling a 3nm digital signal processor for 1.6T optical transceivers in March 2026. Marvell's pursuit of A14 for its next-generation connectivity products indicates that the race to advanced nodes in the AI infrastructure segment is broadening beyond GPU and CPU manufacturers to encompass the full stack of chips required to move data within and between AI systems.
Sectors and assets to watch
The primary tickers directly implicated in this development are Marvell Technology (MRVL) and TSMC (TSM). Marvell's $1 billion prepayment commitment and A14 discussions represent a concrete capital allocation decision tied to its AI connectivity roadmap, while TSMC stands as the sole foundry capable of delivering A14-class manufacturing at scale on the disclosed timeline. Both companies are central to the story's near-term and medium-term trajectory, with TSMC's 2028 volume production schedule setting the outer boundary of when A14-based Marvell products could realistically reach customers.
NVIDIA (NVDA) and AMD (AMD) are relevant as contextual benchmarks: both companies have historically been among the earliest and largest consumers of TSMC's most advanced nodes, and Marvell's entry into A14 discussions reflects a broadening of the leading-edge customer base beyond GPU and CPU designers. Broadcom, though not a primary ticker in this brief, is a direct competitor in optical DSP and connectivity silicon and has already moved to 3nm sampling for 1.6T transceiver applications, establishing a competitive reference point for Marvell's technology roadmap decisions.
What to watch next
Key developments to monitor include any formal capacity reservation agreements between Marvell and TSMC for A14 production slots, which would move the relationship from discussions to binding commitments. TSMC's progress toward A14 volume production — currently targeted for 2028 — and any updates to that timeline will directly affect Marvell's product roadmap. Marvell's quarterly earnings disclosures will be a source of data on whether data center revenue, already above 75% of total revenue, continues to grow as a proportion of the business. Additionally, announcements from competing connectivity chip suppliers regarding their own advanced-node engagements with TSMC or other foundries will provide context for how broadly the AI infrastructure buildout is pulling demand toward leading-edge process technology.