Marvell's CXL Re-Rating Settles Just Under UBS's $340 Bull Case

The Catalyst: UBS Rebuilds the Bull Case Around CXL

There was no earnings report, no contract announcement, and no product launch inside the 24-hour window — this move is analyst-driven. On June 29, UBS analyst Tim Arcuri lifted his Marvell target to $340 from $230 and kept a Buy, arguing the CXL opportunity is "inflecting." The number that actually matters is the revenue model: UBS now expects roughly $1 billion of Marvell CXL revenue in 2027, scaling toward $2 billion in 2028, largely from XPU-attach programs at two U.S. hyperscalers, which feeds higher 2027/2028 estimates of about $16.8B/$23.9B in revenue and $6.23/$9.62 in EPS.

Arcuri wasn't alone. Cantor Fitzgerald raised its target to $300 from $220 over the same stretch, though it stayed Neutral. And the call wasn't Marvell-specific: UBS also lifted Astera Labs to $400 from $205 on the same CXL thesis, and ALAB ran alongside MRVL. When two names move on one note, you're trading a theme — memory interconnect for AI clusters — not company news.

Why the Thesis Has Teeth This Time

Analyst CXL revenue models have floated around Marvell for a while. What changed recently is that Marvell put hardware behind the story. On June 24, Marvell closed acquisitions of CXL-switch maker XConn and optical-compute startup Celestial AI, bolting the exact switching IP the UBS model leans on directly into its portfolio. The sequencing is the tell: the Street started putting CXL dollars in its numbers days after Marvell bought the silicon to earn them.

It also stacks on top of a deeper Nvidia relationship. Back on March 31, Nvidia took a $2 billion equity stake in Marvell and tied it to an NVLink Fusion interconnect partnership, with Marvell supplying custom XPUs and scale-out networking. CXL switching is the memory-pooling layer that sits underneath those clusters. The bull case is that Marvell is no longer just a custom-silicon vendor levered to one or two AI ASIC programs — it's positioning as the interconnect and memory-fabric supplier for them too.

Where the Re-Rating Sits Now

The stock has already priced a lot of this. MRVL pushed near $296 earlier before settling at $293.80, leaving the 24h gain at 7.34% after intraday prints ran hotter. That puts it above the roughly $249 average analyst target — the re-rating has carried price past the Street's mean and into territory only the loudest bulls (UBS at $340, Cantor at $300) still defend. It also follows the June 22 S&P 500 inclusion, which forced index buying and can leave the float jumpy once that mechanical demand clears.

The risk is plain: this is a momentum re-rating built on a forward model, not a beat. The CXL revenue is a 2027–2028 story, and after a months-long run the valuation leaves little room for the timeline to slip. Continuation depends on the next print validating XPU-attach traction; the $282 area from the prior session's base is the obvious shelf if the move unwinds. On the HIP-3 perp specifically, the $16.46M of 24h volume is thin relative to the underlying, so funding and book depth here will swing more on local flow than on anything happening at Marvell.