Marvell's Google Deal Is Real. The Cash Flow Still Has to Catch Up to the Price.

Marvell's Google Deal Is Real. The Cash Flow Still Has to Catch Up to the Price.

The market stamped Marvell Technology a first-tier AI chip company this week, and the shares responded with a roughly 6% pop that put the close near $251. The trigger was a custom-chip pact with Google that hands the search giant a warrant to buy up to 58.97 million Marvell shares — a stake worth about $12.2 billion — with the deal tied largely to how much silicon Google actually purchases through fiscal 2033.

I have no quarrel with the business logic. Google has largely been working with Broadcom on custom chips, and hyperscalers diversify their silicon suppliers exactly the way Google just did: once a design house proves it can handle the volume, a second source inevitably follows. Marvell now counts Amazon and Microsoft as custom-accelerator customers too, and Google is the biggest single prize left. The report on Google "landing the white whale" is not hyperbole; it is a genuine step change in the company's ceiling.

But the price has already read that memo. At $251, Marvell carries a market value near $220 billion, up roughly 195% this year and more than 200% over the past four months. This is no longer an inflection stock. It is a stock whose next twelve months are being shipped to the buyer in advance.

The cash finally showed up

For most of this bull market, Marvell was the AI trade's second cousin — a company with the right addresses in networking, optics, and custom silicon, but financials that always seemed one quarter behind the narrative. That gap is closing. In its first quarter of fiscal 2027, which ended in early May, Marvell posted record revenue of $2.418 billion, up 28% year over year, and a record $638.8 million of cash flow from operations. On a non-GAAP basis — the measure Marvell and the Street use because it strips out acquisition amortization and stock compensation — first-quarter earnings came to $0.80 a share. On a GAAP basis, $0.04. That gap tells you how much of the current income statement is still accounting noise from deals like the Celestial AI and XConn acquisitions the company closed in February.

The trajectory is the point. Management guided the current quarter to about $2.70 billion of revenue, roughly 35% higher than a year earlier, and said it expects growth to accelerate every quarter through fiscal 2027, which is projected to come in around $11.5 billion, up about 40%. More tellingly, the company said it is significantly raising its revenue outlook for both fiscal 2027 and fiscal 2028. The old objection — Marvell's AI story never translated into money — is now stale.

What $220 billion is actually paying for

The year-over-year math is easy. Revenue accelerates, non-GAAP earnings accelerate, and free cash flow is climbing. The multiple in the price is the hard part.

Free cash flow is operating cash flow minus the modest capital spending that comes with being fabless — Marvell designs chips but does not own the factories — and over the trailing twelve months Marvell generated about $1.66 billion of it, at a roughly 19% free-cash-flow margin. Set that against a $220 billion market value and you are paying roughly 130 times trailing free cash flow. That is not a typo.

Run my rough bridge forward, using the company's own numbers. At $11.5 billion of revenue and the roughly 26% operating-cash-flow margin Marvell just delivered, free cash flow lands near $2.5 billion this fiscal year — still about 88 times that cash flow. Slide to fiscal 2028, assume revenue reaches roughly $16 billion and free cash flow pushes toward $3.5 billion to $4 billion, and the multiple only falls to the high-50s or low-60s. In other words, the current price assumes Marvell roughly triples free cash flow over two years, holds a valuation reserved for scarcity, and then keeps growing. That can happen. It is not an entry; it is an assumption.

The sell-side calendar shows how far ahead the tape has run. In late May, Oppenheimer raised its target to $200 from $170 on an Outperform rating; the same week, Stifel pushed its target to $210 from $140. Both calls were already behind the stock when they were written. The shares kept climbing to a 52-week peak of $329.88 on June 18, then slipped about 28% into mid-August as the AI trade got jittery about multiples. The operating path never worsened during that slide — it improved. But the pullback resets nothing now, because the Google deal just re-inflated expectations overnight.

The warrant is the fine print

The structure of the Google deal tells you how long a leash the market is granting. Google can buy those shares at a strike of roughly $206, already below the market, so its option is in the money. The clever part is that the warrant converts into value only as Google hits purchasing milestones that run through fiscal 2033 — effectively a seven-year revenue contract dressed up as equity. Best case, that dilution (nearly 59 million shares, about 7% of the company) is a rounding error next to the orders that trigger it. But the fine print is a reminder that every part of this rally now depends on Google's buying actually materializing on schedule.

The balance sheet, for the record, is clean: about $3.8 billion of cash, $8.7 billion of total debt, and only $1.1 billion of net debt after all the acquisition spending. The financial engineering is not the problem. The multiple is doing all the work.

The two dates that matter

The next evidence point is Thursday, August 27, when Marvell reports its second quarter of fiscal 2027. The bar is set at $2.70 billion of revenue and $0.93 of non-GAAP earnings. What I will watch is not the beat — the company has made a habit of those — but cash flow: whether the operating-cash-flow margin holds near the 26% a record March-May quarter produced while revenue scales, and whether management adds specific numbers to that fiscal 2028 outlook it has only promised to raise. The second date is October 6, an investor day where the Street will demand a free-cash-flow bridge it has never actually been given.

The break condition is equally specific. If a guide points down, or the cash-flow margin slips as the custom business scales, the summer's correction resumes, and this time the warrant does not soften it because the milestone structure only pays off if the buying happens. That risk is concentrated by design: three hyperscalers are both the upside and the vulnerability.

I am not chasing this at $251, and I am not shorting it either. The trade that paid — buying the disbelieved AI supplier while its cash-flow path still lagged the story — was the trade from under $100 to the June peak. What is left from here is a bet that the multiple holds while the numbers sprint to catch it. That is a reasonable multiyear position for someone who already owns it. For a new buyer, the honest move is to wait for one of two things: August 27 showing cash flow is inflecting faster than my bridge, or a shakeout that brings the price back to where the cash flow does the talking. The company earned its celebration. The price has not earned a chase.