Marvell Technology Group Ltd (MRVL) Report Interpretation

Overview

Morgan Stanley assesses Marvell's expanded Google partnership as a material positive for its long-term AI-infrastructure opportunity, particularly in TPU-adjacent connectivity, memory and custom-silicon content. The institution remains Equal-weight, citing elevated expectations, valuation, potential warrant dilution, and the need for more evidence on estimate revisions, share gains and program timing.

Core views

Google expanded its relationship with Marvell across custom-silicon programs connected to the TPU ecosystem, including inference accelerators, storage controllers, NICs, memory-interface controllers and near-memory compute. Google received warrants for up to roughly 59 million Marvell shares at a $206.58 strike. Nearly all vest according to Google purchases, with incremental shares earned for every $500 million of qualifying revenue through early 2033. Full vesting would require $120 billion of cumulative qualifying revenue. Morgan Stanley considers that figure the upper end of the commercial framework rather than a committed spending amount, but believes the scale signals that the relationship could become very large. The report considers the announcement clearly positive for Marvell because it broadens the company's role across the AI-infrastructure stack. Morgan Stanley believes Marvell likely already has meaningful Google exposure through optics and connectivity, while any accelerator-ASIC role is probably narrower. The agreement's wording—programs that “attach to the TPU ecosystem”—is interpreted as evidence of incremental content surrounding TPUs rather than replacement of the core TPU itself. The report flags “Kestrel,” referenced in the warrant agreement, as an unresolved project. It notes reported work on disaggregated inference and memory solutions and Marvell's recent emphasis on CXL, near-memory acceleration, memory expansion and optical shared memory. Product qualification appears contemplated around late 2027, although Morgan Stanley cannot confirm Kestrel's exact nature. For Broadcom, Morgan Stanley does not regard Marvell's agreement as evidence of displacement from Google's core TPU programs. Broadcom signed a long-term Google agreement earlier in 2026 covering future TPU generations through 2031, and the report argues that replacing an incumbent designer in a scaled TPU platform is more difficult than investors may assume. While recent reports of AMD involvement in TPUv10 add uncertainty and Broadcom's Google position remains noisy, Morgan Stanley interprets the development as Google broadening its supplier base and adding adjacent content. It remains positive on Broadcom and believes its TPU moat is underappreciated. For Marvell shares, the institution says the Google deal raises the potential size of XPU-attach and inference opportunities beyond prior expectations. It remains constructive on Marvell's interconnect portfolio—especially optics and scale-up connectivity—and sees long-term upside in adjacent networking, memory and custom-silicon programs. However, expectations have already risen. Morgan Stanley wants sustained estimate revisions, evidence of durable networking share and content gains, and clearer timing and scale for Google programs before becoming more constructive at current levels. The warrants also create incremental dilution if the relationship reaches the higher end of the framework; the report expects issuance to be expensed, although dilution would accompany substantial revenue generation. The standing valuation framework supports Equal-weight. Morgan Stanley's base case is $195, based on roughly 40x CY27e ModelWare EPS of $4.98, described as in line with high-growth AI semiconductors. The bull case is $286 on 50x CY27e ModelWare EPS of $5.72, assuming CY27 revenue exceeds $17 billion, AI opportunity is larger than expected, and multiple cloud-growth opportunities lift the valuation multiple. The bear case is $130 on 30x CY27e ModelWare EPS of $4.33, with weak storage and enterprise data-center demand, continuing networking inventory headwinds, and downside toward tangible book value. The report expects AI-related growth, especially in higher-speed Inphi optical products such as PAM4 and 400ZR, to allow Marvell to outgrow overall data-center spending, but cites valuation and high stock-compensation expense as reasons to remain Equal-weight.

Analysis framework

Morgan Stanley evaluates the commercial terms and vesting thresholds of the Google warrants, maps the disclosed programs to Marvell's AI-infrastructure products, and distinguishes adjacent TPU content from core TPU design. It then tests the share implications through bull, base and bear EPS-multiple scenarios, with scenario probabilities informed by options-market implied volatility.

Methodology notes

• Valuation methodsP/E and PEG Valuation

  Forward EPS multiple valuation

  Morgan Stanley values Marvell through CY27e ModelWare EPS multiplied by 30x, 40x or 50x in its bear, base and bull scenarios.

• Other

  Options-implied risk-reward scenario probabilities

  The report uses implied volatility data as of August 19, 2026 to estimate approximate risk-neutral probabilities of the stock moving beyond scenario prices over three months and one year.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Marvell Technology Group Ltd (MRVL.O)

  Primary covered company; Google partnership may expand long-term AI-infrastructure revenue through TPU-adjacent programs.

  Strengths

  Broad exposure to optics, connectivity, networking, memory and custom-silicon opportunities; potential XPU-attach and inference upside.

  Weaknesses

  Valuation and high stock-compensation expense weigh on enthusiasm.

  Comparison

  The report expects Marvell to add adjacent TPU content rather than displace Broadcom's core TPU design role.

  Risks

  AI opportunity may be smaller than expected; storage, enterprise data center and networking could remain weak; warrants may dilute shareholders if the relationship scales.

• Broadcom Inc. (AVGO.O)

  Incumbent Google TPU designer and comparison company in assessing whether Marvell's opportunity displaces existing suppliers.

  Strengths

  Long-term Google agreement covers future TPU generations through 2031; Morgan Stanley considers its TPU moat greater than appreciated.

  Weaknesses

  Its Google position remains noisy amid reports of additional suppliers, including possible AMD involvement in TPUv10.

  Comparison

  Marvell is viewed as gaining adjacent content while Broadcom retains the core TPU role.

  Risks

  Potential networking-share loss, uncompetitive ASICs or customer losses, and VMware-acquisition execution.

Key data

• Google warrant share capacityUp to roughly 59 million MRVL sharesIssued to Google at a $206.58 strike; nearly all vest based on Google purchases.
• Qualifying-revenue threshold for full vesting$120 billion cumulative revenueThrough early 2033; described as an upper-end framework rather than a committed purchase amount.
• Marvell price target$195.00Morgan Stanley base-case price target.
• Marvell closing share price$237.27As of August 19, 2026.
• Base-case CY27e ModelWare EPS$4.98Valued at roughly 40x in the $195 base case.
• Bull-case price and EPS$286.00; $5.72 CY27e ModelWare EPSBased on a 50x multiple and CY27 revenue above $17 billion.
• Bear-case price and EPS$130.00; $4.33 CY27e ModelWare EPSBased on a 30x multiple.
• Projected GAAP revenue$11,498 million in Jan 2027e; $16,161 million in Jan 2028e; $21,394 million in Jan 2029eMorgan Stanley estimate table.

Impact & implications

The report says the agreement improves confidence in Marvell's long-term AI opportunity and could add meaningful TPU-adjacent revenue, without establishing that Marvell will replace Broadcom in core TPU design. For Marvell's rating, the positive strategic signal is offset by already-higher expectations, valuation, stock-compensation expense, possible warrant dilution and incomplete visibility into execution.

Risks

• Marvell's AI opportunities could be smaller than expected.
• Storage and enterprise data-center markets may remain weak, while networking inventory at customers and in the channel could continue to offset custom-ASIC tailwinds.
• The warrant structure could create incremental dilution and expense if Google purchases scale toward the framework's upper end.
• Broadcom's Google position remains uncertain amid reports of a broader supplier base.

What to watch

• Additional details on the Google agreement at Marvell's upcoming earnings release.
• Evidence of sustained estimate revisions and durable networking share and content gains for Marvell.
• Clarification of Kestrel and the timing and magnitude of Google programs, including potential qualification around late 2027.
• Whether Google broadens suppliers only for adjacent content or changes the role of incumbent core TPU designers.