Marvell’s 2027 AI Vision Ignites Wall Street — But Can the Stock Live Up to It?

Marvell’s latest quarter was one of those classic AI-tape moments where the numbers were “good,” but the stock didn’t really wake up until management started talking about 2027. The company delivered a solid, largely in-line print and modestly better near-term guidance, and the stock initially slipped as investors yawned at what looked like a standard beat-and-raise. The mood flipped on the call when Marvell laid out multi-year data center growth targets that bridge the gap to its long-touted AI revenue ambitions. That forward guide helped fuel a double-digit pre-market pop, but the shares have since slipped back below the psychologically important $100 level as a report from The Information that Microsoft is cutting AI software sales quotas weighed on the broader AI complex.

On the numbers, the quarter itself was respectable rather than explosive. Revenue came in at $2.075 billion, up 37% year over year and 3% sequentially, a slight beat versus guidance and consensus. Data center, now 73% of total sales, reached about $1.52 billion, growing 38% year on year and 2% sequentially, driven by custom AI accelerators and networking products. Communications and other end markets contributed roughly $557 million, up 34% year over year and 8% sequentially, helped by strength in carrier and enterprise networking and the divestiture of the automotive Ethernet business. Non-GAAP gross margin improved to 59.7%, with adjusted EPS of $0.76 versus expectations around $0.73, and management guided Q4 revenue to $2.2 billion at the midpoint with EPS of $0.79, essentially in line with the Street.

Where the story really shifts is in the outlook. CEO Matt Murphy framed the quarter as “record revenue driven by strong demand for our data center products” and then pivoted quickly to the medium-term. Marvell now expects total revenue to approach $10 billion next fiscal year, with data center revenue growing more than 25% and custom chip revenue up around 20%, supported by renewed and expanded orders from its hyperscale customers. Looking out further, management told investors to expect data center growth of at least 25% in calendar 2026 and 40% in 2027, effectively sketching a multi-year AI infrastructure up-cycle and a path toward its previously telegraphed $19 billion data center revenue target by fiscal 2029.

The other big swing was strategy, not spreadsheets. Marvell announced a $3.25 billion cash-and-stock deal to acquire Celestial AI, a photonics specialist whose “photonic fabric” replaces copper with optics for scale-up connectivity inside AI data centers. The company expects Celestial to reach a $500 million annualized revenue run rate by the fourth quarter of fiscal 2028 and $1 billion by late fiscal 2029, with gross margins above the corporate average. Near term, the deal adds about $50 million in annual operating expenses and roughly 5% EPS dilution versus some 2026 estimates, but management and the Street are clearly treating that as acceptable tuition for a differentiated position in next-gen AI fabrics.

Analysts, who had been increasingly skeptical about Marvell’s ability to turn its AI hype into concrete contracts, were impressed by the level of detail and apparent order visibility. Evercore raised its price target to $156 and bumped 2027 revenue and EPS estimates by 17% and 9%, respectively, arguing that the AI product cycle remains under-appreciated and calling out multiple legs of growth across custom compute, interconnect, switches, storage, and Celestial’s photonics platform. Wolfe, Piper, Deutsche Bank, Stifel, and others lifted targets broadly into the $110–$135+ range, with Summit Insights upgrading the stock to Buy, emphasizing that 25% data center growth in FY27 and 40% in FY28 should drive earnings outperformance even with some gross-margin headwinds. The common refrain: the multi-year guide finally “bridges the gap” between current run-rate revenue and Marvell’s long-range AI ambitions.

Management’s tone on the call matched that optimism. Murphy and CFO Willem Meintjes stressed that they have “visibility and orders” backing the out-year guidance, highlighted repeat orders from a “major XPU customer” and growing confidence in a second XPU program ramping in the back half of FY27/FY28, and outlined a roadmap that includes coherent light chips to connect data centers and an enterprise networking business on track for a $1 billion annualized run rate. They also emphasized that non-GAAP opex should grow at roughly half the pace of revenue next year, which, if achieved, would allow operating margins to expand even as Celestial investments ramp. Analysts pressed on the usual pain points—customer concentration, execution risk on Celestial, and the potential for “air pockets” in custom ASIC demand—but the tone of the Q&A was notably more confident than in prior quarters.

The market, as usual, is running a different calculation: what happens if the AI spending curve flattens just as Marvell leans into it? The Information’s report that Microsoft is lowering AI software sales growth targets after sales teams struggled to hit quotas landed just as traders were digesting Marvell’s long-term AI guide. Microsoft’s move, echoed in subsequent coverage from Reuters and others, suggests enterprise AI adoption is progressing more cautiously than the marketing narrative implies, reigniting worry about an AI “expectations bubble” on the software side. For a company like Marvell, which is levered to infrastructure rather than application-layer revenue, that doesn’t break the thesis—but it does raise the question of how smoothly hyperscaler AI capex will translate into sustained chip demand over a multi-year horizon.

Against that backdrop, a few key things to watch from here: • Hyperscaler behavior: The implied wins at major cloud customers (widely assumed to include AWS and likely Microsoft) need to show up in sustained data center growth, not just one-off ramps. Any change in capex commentary from the big cloud players will echo loudly in Marvell’s multiple. • Celestial integration and execution: Investors will want periodic proof points that the photonic fabric roadmap is hitting milestones—design wins, revenue run-rate updates, and whether those promised >corporate-average margins actually materialize. • Margins and opex discipline: With Celestial adding cost ahead of revenue, the Street will scrutinize gross and operating margins to make sure the growth story doesn’t come at the expense of profitability. • Valuation and AI sentiment: After an 80%+ run this year and a cluster of price targets well above spot, expectations are high. Technical levels around $100 and how the stock trades on the next AI-headline scare will be a good barometer of how much patience investors really have for a multi-year build-out.

In short, Marvell finally gave investors the AI infrastructure roadmap they’ve been asking for: measurable data center growth targets out to 2028, a credible photonics asset in Celestial AI, and clear signals that its custom compute business is landing repeat orders. The trade-off is that the story is now firmly a “show-me over several years” proposition, precisely as early signs emerge that AI software monetization is bumpier than hoped. For now, the Street is willing to give Marvell the benefit of the doubt; whether the stock can stay comfortably above that $100 line will depend less on last quarter’s beat and more on how reality lines up with those very ambitious 2027 slides.