I'm LongbridgeAI, I can summarize articles.RBC analyst Srini Pajjuri recommends buying the dip in Marvell over Broadcom. While both companies beat earnings estimates, Pajjuri favors Marvell due to its strong custom XPU pipeline, optical leadership, and justified premium valuation with 45%+ growth. He maintains an Outperform rating on Marvell with a $360 price target, citing expected upside from Google program ramps.
Marvell (NASDAQ:MRVL) and Broadcom (NASDAQ:AVGO) are two major players in the semiconductor industry and key rivals in the race to supply the custom chips and networking technology powering AI data centers.
Broadcom remains the much larger of the two, with a market value of roughly $1.8 trillion versus around $180 billion for Marvell, and it has a substantially broader business spanning semiconductors and infrastructure software. Marvell is a more focused semiconductor company, but it has been gaining ground in custom AI silicon and going head-to-head with Broadcom for business from the hyperscalers building their own AI chips.
Both also disappointed investors with their latest quarterly readouts. The headline numbers came in ahead of Wall Street’s estimates, but they still fell short of the sky-high expectations investors had set.
So, with both stocks now trading some distance below recent highs, investors have an opportunity to buy these AI stalwarts at a discount.
However, taking the measure of the pair, RBC’s Srini Pajjuri, an analyst ranked among the top 2% on Wall Street, thinks one stock’s case is currently stronger than the other.
Starting with Marvell, the company reported second-quarter fiscal 2027 revenue and non-GAAP EPS of $2.74 billion and $0.94, respectively, ahead of the analyst’s estimates of $2.70 billion and $0.93 and consensus expectations of $2.71 billion and $0.93. Its third-quarter outlook of $3.15 billion in revenue and $1.10 in EPS also topped the analyst’s estimates of $3.0 billion and $1.05 and consensus of $3.0 billion and $1.08.
The company raised its fiscal 2027 revenue forecast from $11.5 billion to $12.0 billion and its fiscal 2028 outlook from $16.5 billion to $18.0 billion, implying annual growth of 45% and 50%, respectively, compared with previous expectations of 40% and 45%.
The strength was broad-based, with Marvell’s optical business benefiting from robust demand for 800G and 1.6T products and growing adoption of scale-up architectures.
Meanwhile, management expects its custom business to accelerate in the second half of fiscal 2027 and more than double in fiscal 2028, driven by progress with AWS Trainium and Microsoft’s MAIA XPU program. It also expects “significant acceleration” in fiscal 2029. XPU-attach programs, particularly CXL and memory expansion, are seeing strong adoption among hyperscalers as inference workloads drive demand and memory remains scarce.
Marvell provided limited details on its recently announced Google agreement, with more information expected at its October 6 analyst day. However, management reiterated its projection for more than $10 billion in custom revenue in fiscal 2029 and indicated a “significant upside bias” to previous estimates. The lack of detail might have upset investors somewhat but is not a big deal to Pajjuri. “We are not overly concerned and fully expect Google program ramps to inflect in FY29, which should drive sizable upside to current $10B+ custom revenue target,” the 5-star analyst said. “We raise estimates and believe premium valuation is justified given 45%+ topline growth, optical connectivity leadership, and strong custom XPU pipeline.”
Accordingly, Pajjuri raised his fiscal 2027 EPS estimate from $4.02 to $4.19 and his fiscal 2028 estimate from $6.01 to $6.51. He maintained an Outperform rating on the stock along with a $360 price target, a figure pointing toward 12-month returns of 72%. (To watch Pajjuri’s track record, click here)
Turning to Broadcom, its fiscal third-quarter results were also ahead of expectations, with revenue of $29.6 billion and non-GAAP EPS of $3.32, compared with the analyst’s estimates of $29.4 billion and $3.22 and the Street’s estimates of $29.2 billion and $3.22. AI revenue surged 221% year over year to $16.7 billion, exceeding the company’s previous guidance of $16 billion, while a 75% gross margin was about 150 basis points above consensus despite a higher XPU mix.
Broadcom’s fiscal fourth-quarter revenue forecast of $34.8 billion was broadly in line with consensus of $34.7 billion, while its $21.7 billion AI revenue forecast was slightly above the analyst’s roughly $21 billion estimate.
The company raised its fiscal 2027 AI revenue forecast to $115 billion from more than $100 billion, generally matching Pajjuri’s expectations but falling short of consensus at $123 billion. Management also pointed to strong demand from its six XPU customers and introduced a fiscal 2028 AI revenue forecast of $230 billion, driven mostly by its engagements with Anthropic and OpenAI.
Pajjuri said the extended visibility was encouraging but highlighted component supply and infrastructure readiness as “potential limiting factors.” He also noted that LLM companies have supply and financing agreements with multiple chip suppliers and cloud-service providers, creating uncertainty around silicon allocation. At roughly 18.5 times estimated 2027 earnings, Broadcom is trading at a premium of more than 30% to Nvidia on a stock-compensation-adjusted basis. “We see balanced risk/reward as solid customer engagements and multi-year visibility are outweighed by supply/LPS limitations and what we consider is a high bar for FY27/28,” he summed up.
To this end, Pajjuri rates AVGO as Sector Perform (i.e., Neutral). His $400 price target suggests shares will gain 12% over the next year.
Pajjuri’s colleagues seem to agree with half of his thesis, as the analyst consensus rates both stocks Strong Buys. MRVL’s average price target lands at $302.33, a figure factoring in one-year returns of 45%. AVGO shares are also expected to rise by 45%, given the average target clocks in at $518.81. (See MRVL stock forecast or AVGO stock forecast)
