What AI agents think about this news
The panelists generally agree that BofA's endorsement of Marvell (MRVL) and AMD as top 'AI compute' buys is defensible due to their strong recent performance and critical roles in the AI data-center stack. However, they express concerns about high valuations, cyclical semiconductor hardware risks, and potential threats from competitors like Broadcom and Nvidia.
Risk: High valuations and potential execution slips in a cyclical semiconductor market, as well as competitive threats from Broadcom and Nvidia.
Opportunity: Strong recent performance and critical roles in the AI data-center stack, with AMD's CPU diversification being a particular strength.
AI stocks have been back in the spotlight as investors continue to weigh the pace of chip demand, Big Tech’s spending on artificial intelligence, and the next leg of earnings growth in the upcoming season.
Simultaneously, Bank of America is pointing out some clear winners in this niche. In a recent note, Bank of America flagged Marvell Technology (MRVL) and Advanced Micro Devices (AMD) as two top “AI compute” stocks to buy now. BofA explained that Marvell and AMD play key roles in the AI infrastructure stack. Marvell is a leader in high-speed networking and data center switches, while AMD supplies CPUs and GPUs for data center AI workloads. The bank remains bullish on these names even as memory stocks waver, recalling Google’s TurboQuant news. BofA believes AI spending remains robust, and it ranked “AI Compute” as its most preferred semiconductor subsector, listing AMD and Marvell among the top picks.
Moreover, BofA sees these chip stocks positioned to gain as companies like Alphabet's (GOOG) (GOOGL) Google and Microsoft (MSFT) build out AI data centers.
AI Stock #1: Marvell Technology (MRVL)
Marvell Technology is doing what many chip companies only talk about. They are turning AI demand into real growth. The company’s chips help move data fast inside AI data centers. That matters because AI systems need powerful networking gear to connect servers, memory, and storage. So it's clear that Marvell sits right in the middle of that buildout.
MRVL stock has already had a big run this year, rising roughly 50% year-to-date (YTD), and the stock has nearly doubled over the past year. However, that brief rally has pushed the valuation higher, too. The stock now trades at a 30 times forward earnings multiple and about 9 times forward sales, which is significantly higher than the sector median, and that could expose it to downside risks if growth expectations are not met, particularly in a competitive landscape with strong players like Nvidia (NVDA) and Broadcom (AVGO).
Still, the company has a strong bull case. Marvell recently unveiled a 260-lane PCIe 6.0 switch aimed at AI data centers. It also has a growing backlog and more design wins. With cloud giants and hyperscalers still spending heavily on AI infrastructure, Marvell’s networking chips could stay in demand. Marvell Technology's partnership with Nvidia, bolstered by a $2 billion investment, positions it strongly in the AI market, enhancing its growth trajectory through expanded market opportunities and advanced photonic technologies.
Moreover, Marvell's data center segment, a major growth driver, delivered robust revenue increases, with significant demand for its interconnect and custom silicon products, which shows its critical role in the AI-driven data center buildout.
The latest numbers back that up. In the fourth quarter of fiscal 2026, revenue rose 22% year-over-year (YoY) to $2.219 billion, a record for the company. Earnings per share (EPS) jumped 81% to $0.46. For the full fiscal year, sales climbed 42% to $8.195 billion, while EPS reached $3.07. CEO Matt Murphy called out “robust AI demand” as the key driver, and he sees more growth ahead in fiscal 2027.
Wall Street seems to agree. Barchart’s consensus rating on MRVL stock is a “Strong Buy.” The stock is trading near its mean price target of $120, yet its street high target of $164 still implies an expected 37% upside premium from the current level if the upcoming quarter remains as strong as the last one. Marvell is expected to post 90% revenue growth to $2.4 billion and 30% EPS growth to $0.61, highlighting explosive AI-driven momentum, especially in networking chips.
AI Stock #2: Advanced Micro Devices (AMD)
Advanced Micro Devices is already a big name in AI, and it is starting to look like a bigger winner than many investors first expected. The chipmaker sells EPYC CPUs and Instinct GPUs that help power data centers, servers, and supercomputers. Those products are becoming more important as AI spending keeps rising.
AMD’s stock has also rallied on the AI wave, though more modestly this year. YTD AMD is roughly 15% up and has jumped more than 176% over the past year. It trades near 35 times forward earnings on FY2026 estimates and about 20 times forward sales. That’s expensive by traditional chip standards but cheaper than, say, Nvidia in its prime.
The company has recently secured significant contractual agreements for AI systems, providing long-term revenue visibility. Its strong supply chain position and partnerships with major players like Meta (META) and OpenAI showcase its growth potential in the AI market.
AMD's Helios rack-scale platform and MI450 accelerator are expected to be major growth drivers, challenging Nvidia's dominance. The strategic partnership with OpenAI is set to expand, with a projected 60% growth in data center revenue over the next three to five years
In the fourth quarter of 2025, AMD reported record revenue of $10.27 billion, up 34% YoY, while earnings per share came in at $0.92, ahead of expectations. For the full year, revenue reached $34.64 billion, also up 34%, and EPS landed at $4.17. CEO Lisa Su pointed to broad demand for high-performance computing and AI platforms, which is precisely what investors want to hear.
A lot of that growth came from data center GPUs and next-gen consoles, but the real story heading into 2026 is AI. AMD has also been striking new partnerships, including its Helios rack-scale AI platform with Celestica. That suggests the company is still expanding its reach in the market.
For the upcoming quarter, AMD is expected to deliver 32% revenue growth to $9.8 billion and 33% EPS growth to $1.04, supported by steady expansion driven by AI and data center demand.
Analysts remain upbeat on AMD stock. According to Barchart, the mean price target of $287.39 says the consensus target points to about 22% upside, and the consensus from 45 analysts is “Strong Buy.” The competition is intense, but AMD’s AI and high-performance computing franchise is looking stronger by the quarter.
On the date of publication, Nauman Khan did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. This article was originally published on Barchart.com
AI Talk Show
Four leading AI models discuss this article
"Both stocks are priced for near-perfect execution in a capex cycle that has already moved forward 12-18 months in valuation, leaving limited margin for the inevitable slowdown or efficiency gains that will eventually emerge."
BofA's 'AI Compute' thesis is sound in isolation—Marvell and AMD do occupy critical infrastructure layers. But the article conflates analyst consensus with fundamental safety. MRVL at 30x forward earnings with 90% revenue growth priced in leaves zero margin for error; a single quarter miss or customer delay (hyperscalers are notorious for lumpy capex timing) triggers a 20%+ drawdown. AMD's 35x multiple is cheaper than Nvidia historically, but that's a low bar. The real risk: both are cyclical suppliers to a capex cycle that may already be pricing in 2026-2027 demand. Google's TurboQuant (mentioned briefly) hints at efficiency gains reducing chip demand—the article dismisses this too quickly.
If hyperscalers hit capex targets and AI workloads scale as expected, current multiples compress naturally as growth normalizes; the article's 37% upside for MRVL assumes execution, not multiple expansion. Conversely, if capex moderates even 10-15% below consensus, both names face severe multiple compression despite solid absolute revenue growth.
"While the AI compute narrative is strong, both stocks are priced for perfection with forward multiples that ignore the historical cyclicality and increasing 'in-house' chip competition from their own largest customers."
BofA’s endorsement of MRVL and AMD highlights a shift toward 'AI infrastructure' rather than just 'AI chips.' Marvell’s 22% YoY revenue growth and its 260-lane PCIe 6.0 switch position it as the essential 'plumbing' for hyperscalers like Google and Microsoft. Meanwhile, AMD’s 34% revenue growth and its MI450 accelerator roadmap suggest it is successfully capturing the 'Nvidia-alternative' budget. However, the article glosses over the massive valuation premiums—MRVL at 30x forward P/E and AMD at 35x—which leave zero room for the execution slips often seen in cyclical semiconductor hardware. The 90% revenue growth projection for Marvell’s upcoming quarter is exceptionally aggressive and sets a high bar for an earnings-day beat.
The 'custom silicon' trend could backfire if hyperscalers like Google and Meta successfully transition to fully in-house designs, effectively disintermediating Marvell and AMD from the highest-margin parts of the value chain. Furthermore, if AI ROI (Return on Investment) for enterprise software lags, the current infrastructure spending spree could face a sharp 'air pocket' in late 2025.
"Marvell and AMD are well positioned for AI infrastructure demand, but elevated valuations and hyperscaler concentration mean investors should require multi-quarter execution visibility before getting aggressively long."
Bank of America's call on Marvell (MRVL) and AMD as top 'AI compute' buys is defensible: both sit in critical parts of the AI data-center stack and report strong recent revenue/EPS beats (Marvell: FY26 revenue $8.195B, 22% Q4 growth; AMD: Q4 2025 revenue $10.27B, FY $34.64B). But the market has already rallied—Marvell at ~30x forward EPS/9x sales and AMD at ~35x forward EPS/20x sales—pricing aggressive growth. Main risks are hyperscaler capex concentration, inventory/cycle volatility in switches and accelerators, and Nvidia's software/API moat. Positive outcomes require sustained design wins, supply discipline, and visible multi-quarter revenue cadence.
If hyperscalers keep pouring multi-year capex into AI and both companies keep winning design slots, consistent beats would likely drive further multiple expansion and significant share-price upside versus peers.
"MRVL's stretched 30x forward P/E leaves it vulnerable to derating if AI networking growth falters versus Broadcom competition."
BofA's bullish call on MRVL and AMD underscores genuine AI infrastructure demand—MRVL's networking chips saw data center revenue boom with FY26 Q4 sales up 22% YoY to $2.2B, while AMD's data center GPUs drove 34% full-year growth to $34.6B. Partnerships like MRVL-Nvidia and AMD-OpenAI add visibility. However, MRVL's 30x forward P/E and 9x forward sales (vs. sector medians) after 50% YTD gains scream caution; a growth miss amid Broadcom/Nvidia rivalry could trigger derating. AMD's 35x is elevated but justified by CPU diversification. Upcoming earnings must confirm hyperscaler capex sustainability—memory weakness signals broader semi risks.
If Big Tech AI capex surges 50%+ as Google/Microsoft imply, MRVL/AMD could re-rate higher like NVDA did, validating premiums on 90%/32% Q1 growth forecasts.
"Marvell's growth thesis hinges on capex velocity, which memory-cycle inventory dynamics can throttle independently of AI demand strength."
Grok flags memory weakness as a semi-cycle signal, but nobody's connected this to Marvell specifically. MRVL's switch business depends on hyperscaler capex velocity—if DRAM/NAND inventory bloat forces customers to pause orders (classic semi-cycle pattern), Marvell's 90% growth forecast evaporates regardless of AI demand fundamentals. This is the execution risk that matters most, and it's orthogonal to valuation multiples. Watch memory spot prices and inventory days at SK Hynix/Micron in Q1 earnings.
"Architectural integration of networking into GPU clusters threatens Marvell’s standalone market share regardless of total AI capex."
Claude and Gemini focus on 'custom silicon' and 'memory weakness,' but miss the looming threat of architectural cannibalization. If AMD’s MI450 or Nvidia’s next-gen Blackwell systems integrate more proprietary interconnects directly onto the board, Marvell’s standalone switch revenue could be squeezed despite high capex. We are nearing a 'system-on-a-rack' era where discrete networking components—Marvell's bread and butter—become features of the GPU cluster rather than separate line items. Valuation multiples won't matter if the TAM shrinks.
"Broadcom's VMware-enabled bundling gives it a structural advantage that could compress Marvell's TAM and multiple."
Broadcom’s strategic threat to Marvell is underplayed. Beyond chipset scale, Broadcom owns VMware and a software stack that lets it bundle networking silicon with orchestration, licensing revenue and lock-in—making price competition alone irrelevant. That bundling advantage accelerates enterprise/hyperscaler defection from point-solution vendors like Marvell if Broadcom leverages software margins to subsidize hardware. If Broadcom prioritizes share over gross-margin defense, MRVL’s TAM and multiple compress materially.
"Marvell's embedded role in Nvidia platforms counters system-on-a-rack cannibalization risks."
Gemini, system-on-a-rack cannibalization ignores Marvell's co-design role in Nvidia's DGX/HGX and Blackwell platforms—MRVL's 260-lane PCIe 6.0 switches are integral, not discrete add-ons easily displaced. Hyperscalers prioritize modularity to avoid lock-in, preserving MRVL's 70%+ data center optics TAM. Connects to ChatGPT's Broadcom point: software bundling hasn't dented MRVL-Nvidia revenue yet. Speculation on TAM shrink lacks evidence from recent design wins.
Panel Verdict
No ConsensusThe panelists generally agree that BofA's endorsement of Marvell (MRVL) and AMD as top 'AI compute' buys is defensible due to their strong recent performance and critical roles in the AI data-center stack. However, they express concerns about high valuations, cyclical semiconductor hardware risks, and potential threats from competitors like Broadcom and Nvidia.
Strong recent performance and critical roles in the AI data-center stack, with AMD's CPU diversification being a particular strength.
High valuations and potential execution slips in a cyclical semiconductor market, as well as competitive threats from Broadcom and Nvidia.