Marvell closed down 8% on Friday June 5, taking the worst single-day spillover from the AVGO custom-ASIC reset. The broader Nasdaq lost 4% across Thursday and Friday combined, and the semiconductor sector shed roughly $1 trillion in market capitalization across the same two sessions. The MRVL move was sharper than the index because the custom-ASIC business that AVGO's print just questioned is essentially MRVL's entire growth thesis.
Earlier in May, MRVL had been up 21% in a single session on Jensen Huang's Computex keynote call-out of Marvell as a "next trillion-dollar company," which positioned the stock at a multiple that priced in flawless execution on the AWS Trainium, Google TPU, and Meta MTIA programs. Most of that gain has now unwound across two sessions, and the June 18 fiscal Q1 print is the next catalyst that either confirms the digestion-phase read or sets up a relief rally.
Why the Spillover Hit Marvell Harder Than the Index
The simplest read is business-model overlap. Marvell's custom-ASIC programs serve the same hyperscaler customers that buy from Broadcom, and the design-win architecture is structurally similar. AWS Trainium2 silicon, Google's Axion-adjacent TPU iterations, and Meta's MTIA inference accelerators all run through a co-design model where Marvell or Broadcom partners with the customer on the chip architecture and TSMC fabricates the silicon.
When Hock Tan refused to raise the $100 billion 2027 AI SAM target on the AVGO call, the read across the entire custom-ASIC layer was that the hyperscaler order cycle is in a digestion pause. MRVL has a smaller customer concentration than AVGO and is more dependent on each program individually, which means a single-customer slowdown hits the MRVL multiple harder than the same slowdown hits the AVGO multiple. The 8% Friday move priced that asymmetry.
The systematic and momentum flow flushed first. The CME order-flow data tracked by Bloomberg showed the bulk of the MRVL selling concentrated in the opening 90 minutes Friday, consistent with momentum-following desks unwinding the post-Computex long. The Marvell investor relations page lists no pre-print disclosures that would explain the magnitude, which reinforces the read that this was a positioning unwind rather than a fundamental signal. The long-only allocator book mostly held, which suggests the bottom of the move sits closer to the June 18 print than the current spot price.
What the Computex Hype Actually Was Versus What AVGO Just Revealed
The Jensen Huang call-out at Computex was a strategic alignment statement, not an order announcement. Huang named Marvell as a critical optical and networking partner for the next-generation NVDA reference platforms, which signaled that the NVDA-MRVL relationship inside the AI infrastructure stack is intact and growing. Marvell stock priced that signal as a structural bull case and pulled the multiple up 21% in a single session.
The AVGO call delivered a different message about the same end-market. Tan said the custom-ASIC order cycle from hyperscalers is in digestion through fiscal Q3 and possibly into early fiscal Q4, per the Broadcom fiscal Q2 2026 earnings transcript on Broadcom IR, with the next acceleration tied to the spec lock-in cycle for the architecture generation that ships in 2027. That message applies to MRVL's order book as cleanly as it applies to AVGO's, because the customers are the same and the procurement cadence is the same.
The honest read is that the Computex narrative is structurally intact but the timing of the next earnings acceleration is later than the multiple priced. MRVL can still cross $100 billion in market cap on the custom-ASIC story alone if the 2027 program ramps hit on the original timeline, a thesis covered closely in Reuters analyst commentary following the Computex keynote. The question the market is now repricing is if the ramp lands six months later than the original Street model.
Why the June 18 Q1 Print Is the Next Real Catalyst
The MRVL fiscal Q1 print is scheduled for June 18 after the close, per the Marvell IR earnings calendar, and it now carries elevated significance because the AVGO digestion signal needs to be either confirmed or partially reversed by a peer-group data point. There are three specific items the market will read most closely.
The first is the custom-ASIC revenue mix as a percentage of total revenue. MRVL has guided that the AI infrastructure segment will cross 50% of revenue inside fiscal 2027, and the Q1 mix print is the cleanest indicator of if that trajectory is intact. A miss on AI mix confirms the digestion-phase read. An in-line print partially reverses it. A beat reopens the bull case.
The second is the forward guide on the Trainium, TPU, and MTIA program ramps specifically. Matt Murphy has historically been explicit about the program-level visibility on the call, and the language he uses to frame the next-quarter and full-year ramps will set the multiple. Constructive language reopens the upside. Cautious language confirms the AVGO read.
The third is the operating margin trajectory. Custom-ASIC programs scale at lower margins than merchant-silicon products during the ramp phase, and MRVL has been guiding margin expansion as the programs reach steady-state volume. A miss on margin expansion is the worst possible signal because it would suggest the customer-pricing power has shifted against MRVL.
The print sits inside the broader Phemex AI agents context, where the AI infrastructure layer is the dominant capital-intensity story across the semi sector.
Where MRVL Support Sits and Why the Custom-ASIC Story Is Still Intact
MRVL closed Friday around $70. The support stack from here is well-defined.
$66 is the immediate test. That is the early-May consolidation zone before the Computex spike and the level where most of the post-Computex long entries sit. A hold at $66 with a recovery back above $74 inside the print window is the constructive setup.
$58 is the structural level. That is the February swing low and the level the post-DeepSeek volatility resolved through. A break of $66 likely tests $58 quickly, and that level is where the long-term thesis either holds or breaks. A clean hold at $58 with positive flow on a recovery rally is the bull setup. A close below $58 invalidates the multi-month base and opens $48 as the next major level.
The custom-ASIC story is structurally intact even if the multiple compresses. The hyperscaler customers are committed to in-house silicon as the strategic path forward, the design-win architecture cannot be unwound inside a single product cycle, and Marvell sits on the critical path of three of the four major hyperscaler programs. The question the market is repricing is the multiple, not the business model. A digestion phase produces a lower multiple. A demand collapse produces no business. The current setup is the first, not the second.
Frequently Asked Questions
Is MRVL a buy at the current level?
The cleanest entry waits for the June 18 print to clear and the $66 level to hold on retest. Buying ahead of the print is a directional bet on Matt Murphy delivering constructive language on the program ramps. The risk-reward improves materially if the print delivers an in-line custom-ASIC mix and the post-print drawdown can be bought near the $58 structural level.
How does the AVGO digestion phase actually affect MRVL revenue?
The digestion phase delays the next order-cycle acceleration by an estimated two to three quarters. That translates to MRVL revenue running slightly below the original Street model through fiscal 2027 even if the underlying programs are healthy. The total addressable market does not shrink. The timing of when MRVL captures the next leg of the TAM shifts right.
Why is MRVL exposed to four hyperscalers when AVGO is only exposed to three?
Because MRVL has won design slots at AWS, Google, Meta, and Microsoft (the rumored Maia silicon partnership), while AVGO's three named customers are Google, Meta, and ByteDance. The customer overlap is significant but not identical, and the MRVL exposure is structurally broader. That helps in a normal demand environment and partially offsets the digestion-phase pressure.
What would invalidate the bull case entirely?
A miss on the June 18 print combined with a guide-down on the AI infrastructure mix would be the clearest signal that the digestion phase is deeper than the Street is modeling. Combined with a break of $58, that scenario opens a path to $48 and removes the multi-month base structure. Short of that combination, the digestion-phase read keeps the bull case alive on a longer timeline.
Bottom Line
MRVL lost 8% Friday on the AVGO custom-ASIC spillover and most of the post-Computex Jensen-call gain unwound across two sessions. The structural story is intact. The customer commitments at AWS, Google, Meta, and Microsoft are not in question. The timing of the next earnings acceleration is what the market is repricing. The June 18 fiscal Q1 print is the catalyst that either confirms the digestion-phase read with a soft AI infrastructure mix or partially reverses it with constructive program ramp language. Support sits at $66 and $58. The first level is the trade. The second level is the thesis. The multiple compresses on a digestion phase but the business does not break.
This article is for informational purposes only and does not constitute financial or investment advice. Cryptocurrency trading involves substantial risk. Always conduct your own research before making trading decisions.
