On October 6, Marvell will hold an Investor Day in New York.
Over the past month, the market has been most focused on the new agreement Marvell signed with Google.
Let’s first clarify the timeline.
On July 29, the two parties signed a commercial agreement. On August 3, Marvell announced that it would hold an Investor Day on October 6. On August 18, Marvell issued warrants to Google, and the following day it formally disclosed details of the partnership in an 8-K filing.
So, this Investor Day had already been planned. After the Google agreement was announced, market interest in the event surged, because the most important revenue details of the agreement have still not been fully disclosed.
First, look at the agreement itself.
Google can subscribe to up to about 59.87 million Marvell shares, with an exercise price of $206.58. About 1.36 million shares vest based on time, while the rest are tied to Google’s actual purchase amount.
For every $500 million of related purchase revenue generated by Google, one tranche of warrants is unlocked, for a total of 240 tranches.
240 multiplied by 500 million dollars equals 120 billion dollars.
This figure is easily interpreted as Google placing a $12 billion order with Marvell. The contract does not make such a commitment.
$12 billion refers to the cumulative revenue required for all purchase-type warrants to fully vest, and the calculation period extends all the way through Marvell’s FY2033.
It is closer to a long-term incentive arrangement. The more Google purchases, the more Marvell shares it can obtain; Marvell gets orders and exercise proceeds, while existing shareholders must consider potential dilution.
The agreement also covers many products, including AI inference accelerators, storage controllers, network interface controllers, memory interface controllers, and near-memory computing.
This shows that the scope of cooperation between the two sides has already gone beyond a single TPU project. As Google expands its own AI infrastructure, Marvell hopes to participate simultaneously in compute, networking, storage, and memory connectivity.
The question is, when can these partnerships turn into revenue?
On August 27, Marvell released its second-quarter results. The company raised its FY2027 revenue outlook to about $12 billion, and also lifted its FY2028 revenue outlook from $16.5 billion to $18 billion.
FY2028 refers to the fiscal year ending in early 2028.
Management also said the contribution from the Google agreement before FY2028 is already included in that $18 billion, and the larger impact from the partnership will only appear in FY2029 and beyond.
The day’s results were slightly above market expectations, and next-quarter guidance was also above consensus, yet the stock still fell more than 6% after hours.
This reaction is highly instructive. The market’s expectations for Marvell are already higher than just a typical quarterly beat.
Investors want to know what level the projects with Google, Microsoft, and AWS can ultimately push FY2029 and FY2030 revenue to.
Management did not go into detail at the time, leaving the long-term growth framework for the October 6 Investor Day.
This is the link between the Google agreement and Investor Day.
Now let’s look at valuation again.
As of October 2, Marvell’s stock price was $272.29, with a market cap of roughly $240 billion to $250 billion, and its year-to-date gain was already close to 200%.
Using the company’s revenue targets, the current market cap is roughly 20x FY2027 revenue and 14x FY2028 revenue.
Wall Street currently expects Marvell’s FY2027, FY2028, and FY2029 non-GAAP EPS to be about $4.21, $6.76, and $10.47, respectively.
Based on a share price of $272, the corresponding P/E ratios are approximately 65x, 40x, and 26x, respectively.
This valuation already includes rapid growth over the next two to three years. The average analyst price target is about $298, leaving less than 10% upside from the current share price.
So even if a large long-term number is announced on October 6, the first thing to check is whether it exceeds what the market has already priced in.
Morgan Stanley expects Marvell may present a framework for FY2030 revenue above $40 billion at Investor Day. Citi’s long-term expectation is for FY2030 EPS to reach $15 to $20, while operating margin continues to improve from the current 38% to 40%.
More bullish sell-side models are even calculating $22 billion in custom-chip revenue, nearly $15 billion in optical business revenue, and more than $40 billion in total company revenue.
These figures are currently analyst estimates. If management only says “FY2030 revenue above $40 billion,” the market may not be surprised.
The first and most important thing to watch on Investor Day is whether the company can break down its long-term targets.
Of the $40 billion, how much comes from custom XPU, how much from XPU Attach, how much from optical interconnect, switching chips, and legacy businesses?
In which years do these revenues begin to scale?
If management mainly presents AI market size and product roadmaps, investors will find it hard to raise earnings forecasts based on that. What the market needs is the revenue change between FY2028 and FY2030.
The second item is the customers and ramp-up progress of the custom-chip business.
AWS is still Marvell’s most mature custom-chip customer. The Microsoft Maia project carries very high market expectations, while the Google partnership provides new long-term upside.
Investors need confirmation of the shipment scale of the Microsoft project in 2027, which Google products have already completed design, which are still in development, and the approximate timing for each project to enter mass production.
Whether FY2029 custom business can exceed $10 billion has now become an important threshold. If management’s target is only $10 billion to $12 billion, it broadly matches sell-side expectations.
To drive a meaningful upward revision to earnings forecasts, higher revenue or better margins are needed.
The third item is XPU Attach.
This part includes network interface cards, CXL controllers, storage controllers, memory interfaces, and other supporting chips.
Its importance comes from the number of products. Once a cloud vendor develops a custom XPU, it still needs a variety of other chips around it to handle transmission, storage, and memory connectivity. If Marvell participates in these products at the same time, it can generate more revenue from the same system.
Investor Day needs to provide several concrete figures: on average, how much Attach revenue corresponds to each XPU, how many design wins have been secured, when mass production is expected, and how gross margin differs from that of custom XPU.
If this part can only provide market size, it will still be very hard to model the revenue.
The fourth item is the optical interconnect business.
Marvell completed its acquisition of Celestial AI this February. The company previously expected this business to start contributing revenue in the second half of FY2028, reach an annualized revenue run rate of $500 million in FY2028 Q4, and $1 billion in FY2029 Q4.
Investors need to watch whether management maintains this set of targets and how many customers have already entered design validation.
800G, 1.6T, NPO, CPO, and in-rack optical interconnect will all be mentioned. What truly affects valuation are customer count, mass-production dates, and revenue scale.
The fifth item is margins.
Marvell’s current third-quarter non-GAAP gross margin guidance is 57.5% to 58.5%. Higher volumes in custom chips will increase revenue, but may also drag down overall gross margin, because some products include higher-cost steps such as manufacturing, packaging, and HBM.
Investor Day needs to explain how high operating margins can rise after rapid revenue growth. Citi mentioned market expectations for long-term operating margin to improve by another 100 to 200 basis points from the current 38% to 40% base.
If revenue reaches $40 billion but margins do not improve meaningfully, the upside revision to EPS will be limited.
Finally, let’s discuss how the stock price might react.
I would break it into three scenarios. The price range given here is suitable for observing the next one to three months; intraday volatility on Investor Day will also be affected by the broader market and U.S. Treasury yields.
In the first scenario, management gives FY2030 revenue above $40 billion, FY2029 custom business reaches $10 billion to $12 billion, and operating margin stays around 40%.
This set of data basically matches current market expectations. The stock may fluctuate in the $250 to $300 range, and short-term profit taking may also occur.
In the second scenario, FY2030 revenue reaches above $45 billion, FY2029 custom business exceeds $12 billion, XPU Attach and optical interconnect both provide clear customer and mass-production plans, and long-term operating margin reaches above 42%.
In that case, analysts will raise both revenue and EPS. The stock has a chance to be re-priced first toward $325 to $350.
Using the current FY2029 consensus EPS of $10.47, $272 implies a P/E of about 26x. If that multiple is maintained, $350 would require FY2029 EPS to rise to about $13.5, nearly 30% above current consensus.
This is also a condition that the stock price re-rating must meet. The company needs to show the market a mid-term EPS path of $13 to $15; simply increasing revenue scale is not enough.
As for $400 to $450, the bar is even higher.
Even at $450 and using a 30x P/E, that would require $15 in EPS, about 43% above current FY2029 consensus.
Management needs to provide a revenue target above $40 billion, stronger margin improvement, and sufficiently specific order-based support.
In the third scenario, management does not provide clear FY2029 and FY2030 targets, the ramp-up timing of the Google and Microsoft projects is pushed back further, the revenue plan for Celestial AI is delayed, and custom chips weigh on gross margin.
In that case, the market may re-evaluate the certainty of earnings over the next two years. $220 to $240 would become a more reasonable valuation range.
The current interest-rate environment also raises the bar for a re-rating. With the 10-year U.S. Treasury yield still above 5.2%, the market is unlikely to keep expanding valuation multiples based solely on revenue visions many years out.
To sum up,
How much can FY2029 and FY2030 actually earn?
When will Google, Microsoft, and AWS start contributing revenue, respectively?
How much accompanying chip revenue can each custom XPU generate?
Has the optical interconnect business entered mass production yet?
After revenue grows, how high can margins and EPS rise?
If all of these issues come with specific numbers, Marvell’s earnings expectations may continue to be revised upward.
If the answer still centers on market size and product roadmaps, the current $250 billion market cap has already priced in quite a lot of growth.