Marvell Technology's Dance With AI: A Reality Check On Hype And Valuation

Last month, we published a neutral article on Marvell Technology (NASDAQ:MRVL) highlighting its competitive advantage and near-term challenges, including unfavorable product mix and inventory management issues. Since we published the article, the Artificial Intelligence revolution has created a frenzy in the semiconductor industry, with Marvell becoming the latest participant in this wave. Recent optimistic management projections during their earnings call has caused a buzz in the investment community. But are these projections indicative of a bright future or are they simply amplifying the hype?

In this article, we aim to provide a comprehensive analysis of Marvell's potential in the AI landscape, an exploration of its recent financial performance, and an in-depth evaluation of the company's valuation. This analysis is intended to equip investors with the necessary insights to make informed decisions and shield themselves from potential market over-enthusiasm.

FYQ1 Earnings: Lots of Sizzle, Not Much Steak, Yet.

Our analysis of Marvell's latest earnings and financial performance draws attention to a few striking patterns. The FY Q1 earnings announcement on May 26, 2023, led to an astonishing 32.4% uptick in stock price. However, this positive reaction is in sharp contrast to the company's declining revenue, which has contracted by 8.7% year-over-year to $1,322 million. Even though this figure was in line with consensus estimates, we believe the celebration is somewhat premature, especially given the significant deterioration in operating margin from 35.4% to 25.3% year over year.

From our perspective, the reason for the market's positive response was the management's adeptness in presenting the company as an AI beneficiary in the earnings call. Notably, management projected AI-related revenues to increase significantly:

In fiscal 2023, we estimate that our AI revenue was approximately $200 million, up dramatically from the prior year. This revenue was primarily from our PAM4 optics and 400ZR BCI products. Since our last earnings call in early March, bookings for these products have increased very significantly. As a result, we expect Marvell's overall AI revenue to at least double in fiscal 2024.

Looking to fiscal 2025, we expect robust growth to continue from AI for network connectivity. Layering on top of this is the growth we expect from the ramp of the cloud optimized programs we discussed earlier in the call. In aggregate, we foresee our overall AI revenue to at least double again next year. In other words, we are forecasting an AI revenue growth CAGR of over 100% for the fiscal 2023 to 2025 time frame.

If MRVL achieves $1 billion in AI-related revenue by FY25, by our calculations, this would imply that the company could generate ~15% of its total revenue from AI. While this is exciting, we believe investors could get more exposure to AI at a much lower valuation through Broadcom (AVGO), which we believe could generate ~20% of its total FY24 revenue from AI-related revenues. For more information, see the Financial & Valuation section below, and read our latest article on Broadcom.

How MRVL Benefits From The AI Boom

Marvell Technology stands to greatly benefit from the surge in Artificial Intelligence due to several key reasons. First, the company has recognized the rising importance of AI in cloud applications and has positioned itself to meet the changing needs of its cloud customers. In essence, Marvell is capitalizing on the shift in data center architecture from dual-socket servers, common in standard cloud infrastructure, to systems that are primarily built around multiple accelerators such as GPUs for AI applications.

A standout trend in AI data centers is the massive increase in the amount of data being processed, leading to a significantly higher requirement for bandwidth. The network bandwidth required to interconnect the systems in AI data centers is markedly higher than that in standard cloud infrastructure. Marvell has responded to this by focusing on high-bandwidth connectivity solutions, such as its PAM4 optical DSP platform, that offer ultra-low latency and high reliability. The company is also prepared for the growing trend of more advanced AI systems requiring even higher bandwidth. Their PAM4 DSP platform is poised to support this increasing demand, further driving its adoption.

Moreover, as AI technology rapidly advances, the technology refresh rate is speeding up significantly in AI infrastructure compared to standard infrastructure. This faster refresh rate is expected to be another major demand driver for Marvell's PAM4 DSP platform. The company is also leading the charge with their industry-first 1.6 terabit PAM4 PSP platform, offering double the throughput of the current generation.

In addition to the intra-data center connectivity, inter-data center connectivity, referred to as Data Center Interconnect (DCI), is crucial for AI cloud data centers. As these centers grow and become regionally distributed, they need to be connected via high-bandwidth, low-latency optical links over substantial distances. Marvell has been a pioneer in the DCI market, offering 400 gigabit-per-second solutions in its latest DCI product line. The company is already seeing a significant increase in demand for these solutions due to the rise of AI.

Furthermore, Marvell is poised to leverage the opportunities in the growing market for low latency, high capacity switches, which are essential for AI data centers. The company's recently announced HaraLynx-10 Ethernet switching platform is built on Marvell's leading 5-nanometer technology platform and is expected to help lead the industry's transition from 12.8T switches to 51.2T switches, quadrupling network bandwidth.

Marvell's foray into custom silicon solutions is another major avenue through which the company can benefit from the AI boom.

AI data centers have different computational demands compared to traditional cloud infrastructures. In standard cloud infrastructure, Central Processing Units (CPUs) handle most of the compute tasks, with accelerators being used for offload tasks. However, in AI infrastructures, this architecture is flipped - accelerators, like GPUs, become the primary compute engines, while CPUs are used for control functions. This fundamental shift opens up enormous opportunities for Marvell.

Many cloud customers are now building custom accelerators designed to address their specific AI needs. Marvell's cloud-optimized silicon strategy is uniquely tailored to cater to these requirements. This approach allows Marvell to meet the rising demand for custom compute solutions in AI infrastructure.

Financial & Valuation

Note: All historical data in this section comes from the company's 10-K filings, and all consensus numbers come from FactSet.

Looking at the company's financial trends, on the surface, MRVL's has a solid track record in revenue growth, with a three-year CAGR of 29.9%. However, much of this growth comes from a number of acquisitions, making it very difficult to assess its true organic growth. Sell-side consensus anticipates a 6.7% contraction in revenue this fiscal year, reducing it to $5.5 billion. While an expected rebound of 16.9% in the following fiscal year might offer some comfort, it doesn't mask the disturbing downward trend we've witnessed this year.

Although MRVL's EBIT margin grew impressively by 15.3 percentage points over the past three fiscal years, the forecasted contraction of 617 basis points this fiscal year to 29.3% again raises concerns. It should be noted that the operating numbers have been adjusted to account for the negative impact of M&A and other corporate activities on earnings.

In terms of cash flow, the company shows promise. Consensus estimates indicate a healthy free cash flow margin of 23.9% for this fiscal year, up significantly from a margin of 16.4% four years ago. MRVL's average FCF margin over the past four fiscal years has also been robust at 18.7%. Furthermore, the firm's capital expenditures amount to 3.6% of revenue, demonstrating moderate capital intensity, which in our view underscores the efficient use of resources.

However, not all metrics look optimistic. Looking at GAAP EPS, MRVL has not been posted positive earnings since 2019. If we do not exclude various M&A, corporate, and stock-based compensation expenses, MRVL's return on invested capital is rather dismal -0.8%. In addition, the stock's current dividend yield of 0.4% is 110 bps below that of the S&P 500.

Concerning stock valuations, MRVL's FY2 PEG ratio stands at a premium of 38.2% relative to the S&P 500's PEG ratio, which could deter some investors. Moreover, the stock's current forward 12-month P/E of 33.1 is significantly above its 5-year mean of 25.4, suggesting a historically high valuation, though still within its 2-standard deviation range. When compared to its peers Broadcom (AVGO), NXP (NXPI), and AMD (AMD) trading at 18.3, 13.3, and 36.8, respectively, MRVL does seem to carry a higher valuation.

Conclusion

Marvell's positioning in the AI arena is indeed promising, given their strategic approach to capitalizing on changes in data center architecture and their customized silicon solutions. Their significant growth projection in AI revenue, while attractive, does warrant cautious optimism, given the firm's recent financial performance and declining revenue. While there's potential for a bright AI future, the valuation appears excessive.

Comparatively, Broadcom's lower valuation and substantial AI-related revenue stream might provide a more attractive AI investment opportunity. Furthermore, for those seeking more AI-dedicated play, Nvidia (NVDA) presents a faster-growing alternative. With AI's continued growth, investors have a range of options, and while Marvell's plans are encouraging, its current valuation makes it a less appealing choice.

As always, a balanced portfolio that includes a diverse set of AI stocks might be the best approach for long-term growth and risk management. The key is to look beyond the hype and make decisions based on thorough research and careful consideration of the financials and valuations of these companies.