UBS Raises Marvell Price Target to $340, CXL Revenue Expected to Reach $1 Billion by 2027
Alina Collins
UBS lifted its Marvell price target from $230 to $340, driven by the company's lead in CXL — a fast-growing protocol that widens the data highway between chips inside data centers.
What is CXL, and why does it suddenly matter?
CXL — Compute Express Link — is a connection standard that lets CPUs, GPUs, and other chips share memory and talk at high speed. It is becoming core plumbing for data centers.
UBS estimates the total CXL market will reach $4.5 billion by 2027 and expand to $7–10 billion by 2030.
This means → CXL is not a niche accessory play; it is an infrastructure layer tied directly to the AI compute build-out.
Where does Marvell's $1 billion come from?
Analyst Timothy Arcuri breaks CXL revenue into three buckets: interconnect (traditional expanders), XPU-attach (custom designs for hyperscalers), and CXL switches.
Nearly all of the ~$1 billion target comes from XPU-attach sockets. In plain terms = Marvell's money rides on making the custom key that fits each cloud giant's AI chip lock, not on selling generic parts.
Marvell currently has five XPU-attach programs with two major U.S. hyperscalers, including Google's TPU (per UBS's assessment).
How much did the revenue forecasts move?
UBS raised Marvell's full-year revenue estimates: FY2027 from $16.5B to $16.8B, FY2028 from $21.9B to $23.9B.
The 2028 uplift — +$2 billion — is almost entirely driven by higher CXL assumptions.
This reflects a broader Wall Street shift: CXL is being re-priced from "concept narrative" to "visible revenue."
Who else stands to benefit?
Astera Labs (ALAB): its Leo CXL expander is expected to see Microsoft order ramp in H2 2026, followed by new hyperscaler design wins in 2027. Per-CPU expander content value could rise significantly from the prior $100–$200 range.
UBS raised Astera's target from $205 to $400 but kept a Neutral rating. This means → UBS sees the growth but believes the stock already prices in much of it.
Broadcom (AVGO) was also flagged as a CXL beneficiary, though UBS provided no specific figures.
What needs to be proven next?
Whether Marvell can expand beyond XPU-attach into CPU-attach and switch share is the key checkpoint for the revenue path.
In plain terms = the $1 billion story stands on one leg — custom chip bundling. Getting to $2 billion requires the second and third legs to show up.
Agentic AI — a paradigm where AI agents autonomously execute multi-step tasks — and its incremental CPU demand form another pillar of UBS's long-term CXL thesis.
Content is for reference only, not financial advice.