Guidance Beat Forecasts by 19%, Yet the Stock Fell 18.6%

Start with operations, where there is almost nothing to fault. A quick glossary: "consensus" is the average analyst forecast, "guidance" is the company's own forecast, EPS is earnings per share, and "non-GAAP" is management's adjusted version of profit, which strips out items such as stock pay.

One detail is easy to miss behind the headline. Revenue only hit the midpoint of guidance, so the entire upside came from margins. The EPS guidance beat is also more than double the revenue guidance beat, which says Wall Street underestimated operating leverage, not shipment volume. That fits what the company keeps saying about constrained supply: there are only so many units to sell, so the extra profit comes from higher prices and a better product mix.

Gross margin can feel abstract, so here it is in plain terms. A year ago, for every $100 the company took in, it spent just over $60 making the product. Now it spends less than $47.

Here is the five-quarter trend (quarterly company releases; next quarter uses guidance midpoints):

The revenue added each quarter has climbed from $132 million to $143 million, $198 million and then $256 million, and next quarter's guidance lifts it to about $340 million. In management's words, the revenue added this quarter exceeds what the whole company brought in during a single quarter two years ago (8-K 2026-11-02). The profit math is more interesting still. At the guidance midpoints, next quarter's operating profit is about $728 million, up $194 million from this quarter. That means about 57% of the added revenue drops to operating profit, below about 65% this quarter. So the guidance does not assume more aggressive leverage, which is why we think it leaves room to be beaten again.

The company gives no full-year guidance, so we have to work out the bar ourselves. If each quarter in the second half of FY27 keeps adding $350 million, Q3 needs to reach $1.95 billion and Q4 $2.3 billion, for a full year of about $7.1 billion. That is more than double FY26's $3.014 billion. The number to watch in the next report is whether the midpoint of Q3 revenue guidance reaches $1.95 billion, meaning the quarterly increase has not slipped. If the increase narrows, the market will immediately read it as either capacity hitting a wall or demand peaking, and either reading would do real damage to the valuation.

On demand, the hardest evidence this quarter is an order for NPO/CPO, a new packaging approach that places optical parts next to or inside the switch chip. The company calls this its second major design win, backed by purchase commitments of more than $5 billion (company release). Visibility on demand for OCS, optical circuit switches that use mirrors to redirect light directly and skip the conversion between light and electricity, has also widened noticeably, and management describes it as a "multi-billion-dollar" annual revenue source (company release). On capacity, the CEO said on October 9 that AI demand has booked capacity through early 2029 (cited by Trefis), and optoelectronic capacity is effectively sold out (24/7 Wall St.). The term, customer and cancellation provisions behind the $5 billion are all undisclosed, and this quarter's call transcript was not available at publication. So treat it as a number with a clear direction but an unclear timeline.

One cash figure sits awkwardly beside the strong income statement. Cash plus short-term investments ended the quarter at $1.98 billion, down $758.8 million from the prior quarter (company release), even though adjusted EBITDA, a rough measure of operating cash earnings, was $578 million. Why did cash fall so much in such a profitable quarter?

The exact breakdown has to wait for the 10-Q, because this quarter's release has no cash flow statement. There are two visible clues. Net fixed assets rose from $1.159 billion to $1.41 billion, and current debt fell from $1.597 billion to $1.36 billion. Capacity expansion and settling convertible bonds were probably the main uses, but how much went into receivables and inventory cannot be determined yet.

Here are the funding sources and obligations side by side (10-K 2026-08-17, company release):

The debt line needs unpacking. The $1.36 billion of convertible bonds in current liabilities looks alarming, and some read it as "short-term debt surging." In fact, it sits there mainly because the stock is far above the conversion price, so holders can convert at any time. Pressure from contractual maturities is small. Principal must be settled in cash on conversion, though (10-K), which makes this a likely contributor to this quarter's cash decline. The larger cost falls on the share count. A convertible bond exchange once produced a one-time non-cash loss of $7.8 billion, the source of FY26's GAAP net loss of $6.935 billion and Q4 GAAP EPS of -$84.65. That accounting loss took no cash out of the door, but the dilution is real:

The profit adjustments deserve a look too. GAAP and non-GAAP operating profit differed by $174.6 million this quarter. Of that, stock compensation and related taxes were $146.7 million, and $89.4 million of that came from an award modification tied to an executive's retirement, a one-time item (company release). Strip it out and the gap narrows noticeably, so this quarter's GAAP profit quality is better than it looks.

Customer prepayments are the line most worth thinking about. For a company with billions in revenue, year-end customer deposits were only $16.8 million. Expansion is mostly self-funded, and long-term contracts protect the company mainly through take-or-pay terms, where customers pay whether or not they take delivery, and pricing clauses (earnings call 2026-08-11), not through cash up front. The upside is that the company has not traded discounts for prepayments. The cost is that if capex keeps rising, cash burn will look worse than the income statement. This company is not short of money, but it does not have so much cash that it can ignore the cycle.

The merger itself needs no financing because it is paid entirely in stock. The real issue for Lumentum shareholders is the terms. The exchange ratio is fixed at 4.500 with no upper or lower protection (the 8-K summary of 2026-11-01 mentions no collar), so every move in Marvell's stock passes straight through. The breakup fees are not soft. Marvell pays $5.8 billion if its board changes its recommendation or accepts a better offer, Lumentum pays $2.9 billion in the equivalent case, and Marvell pays $3.2 billion if regulators block the deal, equal to about 5% of Lumentum's current market value (8-K). That regulatory fee cushions part of the blow, but falls far short of covering the valuation pullback if the deal collapses.

The tightest upstream link is InP, or indium phosphide, the base material for high-speed laser chips. Reuters reports that about 90% of global InP substrates are concentrated in three companies, and China added the material to its export control list in February 2025. Lumentum's response is a primary arrangement with a Japanese supplier plus a new supply agreement with AXT, but management has also said that substrates may need more help if demand keeps accelerating (earnings call 2026-08-11).

At the wafer level, the company is expanding two InP fabs in Japan and converting its Greensboro plant from gallium arsenide to InP. First revenue from Greensboro comes in early 2028, with full output from late 2028 to 2029 (earnings call). That means growth in 2027 must come mainly from squeezing more out of existing plants, with no new factory to the rescue. In July the CEO said the supply-demand gap for EMLs (electro-absorption modulated lasers, the main light source for 800G and 1.6T optical modules) exceeded 30% (cited by TechNews). The Q4 call also said EML unit output would grow more than 50% year on year by the December quarter, and supply still cannot keep up.

Customer concentration is the other side of the picture. The largest customer made up 26.6% of FY26 revenue, sharply up from 15.4% the year before, and the second largest 15.0% (10-K). Together they exceed 40%. The company does not name them, so we will not guess here. By shipment destination, 79.2% of revenue is outside the US. Asia-Pacific is the largest region at about $1.75 billion in FY26, followed by the Americas at about $1.08 billion and EMEA at about $177 million (10-K). This only reflects where goods land at contract manufacturers, however, and does not mean end demand is in Asia.

The contract claims should be sorted by how firm they are:

We judge the most fragile link to be the handoff between product cycles. Pump lasers have an estimated market share of about 70%–80% and are sold out, and EMLs are controlled by a few players. These moats are deep: TrendForce data shows Lumentum, Broadcom and Sumitomo Electric together hold about 55% of monthly EML and CW-DFB capacity. But large-scale CPO volume has to wait until the second half of 2027 through 2028, NPO until late 2027 to 2028, and Greensboro until 2028. If cloud companies slow capital spending before then, a gap opens between today's fast growth and tomorrow's ramp. One customer also has an in-house OCS source (earnings call).

Another risk comes from the manufacturing footprint. The company has plants in the US, Thailand, China, the UK, Slovenia and Japan (10-K), and TrendForce notes that China holds 56% of global optical module manufacturing capacity (2026-08-05). If the US imposes restrictions, a quick decoupling would be hard, and this issue could become entangled with regulatory review of the merger.

No comparable optical company has reported this quarter yet, so there is no read on how the market priced a peer's report. Arista reports on November 3, Coherent on November 4 and Applied Optoelectronics on November 5, so that comparison arrives within a week. For now, here are the latest public numbers side by side:

Coherent is the largest by revenue, but Lumentum leads on growth and gross margin, with a margin gap of more than 10 percentage points. The source of the gap is clear. Components make up nearly 60% of Lumentum's revenue, and those are laser chips in short supply. Fabrinet is a contract manufacturer, while Applied Optoelectronics and Coherent sell more modules. Whoever controls the scarce upstream step takes the profit. Coherent has set a target of more than $3 billion in quarterly revenue by the end of fiscal 2027 and says outright that laser output caps its data center revenue (TIKR summary), which supports the same point.

Demand direction is consistent across peers. Lumentum and Applied Optoelectronics both describe AI-related demand as a tailwind, although the latter's wording also covers cloud infrastructure and CATV (peer disclosures, as of 2026-11-02). On EPS, Lumentum's $5.90–$6.20 is next-quarter guidance, while Fabrinet's GAAP EPS of $3.83 and Coherent's $1.19 are FY26 Q4 actuals. One is a forecast and the others are realized, so they cannot be ranked directly.

Coherent's report on November 4 bears directly on Lumentum. If it also says laser supply remains the bottleneck, Lumentum's shortage story gains a second witness. If it is first to say supply is easing, we need to rethink how long a 53% gross margin can last.

The usual yardsticks have stopped working here. FY26's GAAP net loss of $6.935 billion came entirely from a non-cash debt loss, so the trailing P/E on quote screens shows either 0 or several hundred times. The PEG of 1.1 rests on the same distorted earnings base and growth rate, so it is unusable too. A better ruler is to annualize guided next-quarter earnings, then cross-check using the exchange ratio.

What does 28 times mean? For a company more than doubling revenue, it can even look cheap, and analysts agree: of 36, 31 rate it positively, 5 say hold, and none is bearish. But this multiple pays for something very specific. It assumes an operating margin above 45% is the new normal, when a year ago the figure was 18.7%. Cyclical chip companies look cheapest when margins are at historic highs, which is why P/E often misleads during a shortage. The 0.5% free cash flow yield is a reminder that profits have not yet fully turned into cash.

The merger rewrites the valuation question again. Based on the October 28 close, the announcement implied an EV of $115 billion, a premium of about 23.9% to the unaffected share price (Marvell press release). Because the ratio is fixed, though, Marvell's roughly 32% drop from then to the November 2 close shrank the implied value by about 32%. By Motley Fool's math, it was down to about $81.2 billion by midday Monday, and lower still at the close.

The figure that really matters is a ratio. Lumentum's share price divided by Marvell's was about 3.63 on October 28 and about 3.99 at the close on November 2. That is 11.3% below the 4.5 agreed ratio, while last Friday the discount was only 4.9% (Motley Fool). If the deal closes on schedule in the first half of 2028, that 12.7% upside works out to roughly 7%–9% a year, paid in Marvell stock, not dollars. Against a 10-year Treasury yield of 4.94% (2026-11-02), the spread is not generous, though it does show the market wants compensation for a year and a half of regulatory and shareholder-vote uncertainty. Lumentum's own fast growth has moved to second place. First place now belongs to Marvell's share price, and Marvell's valuation depends on delivering its revenue targets of $20 billion in fiscal 2028 and $70–$90 billion in fiscal 2031 (Motley Fool). The companies say the deal adds to non-GAAP EPS in its first year and target more than $1 billion in cost synergies. Rosenblatt, however, estimates it dilutes Marvell's EPS by 5.0% in 2028 and 2.4% in 2029 (cited by BigGo). With two sets of math this far apart, the market is unlikely to award a high valuation until the disagreement is settled.

Our view: the upstream optical shortage is in the middle of its upswing and still some way from a peak, but the easiest stretch of valuation expansion has passed.

On selling prices, the shortage-era bargaining power already shows in the financials. Double-digit price increases for 200G EMLs are an analyst expectation (TechTimes citing Nikkei). The company's own wording is more cautious: CW lasers currently carry a premium, long-term contract pricing is mostly forward-looking, and the margin benefits are still arriving (earnings call 2026-08-11). InP has no transparent contract price index, so this can only be read indirectly through supplier margins, and five straight quarters of rising gross margin is the closest proxy.

On capacity and tightness, pump lasers are sold out, the EML supply-demand gap exceeds 30%, 1.6T is just starting to ramp, 200G EMLs exceed 25% of revenue, and the company expects them to be more than half of EML units by mid-2027 (earnings call). On demand, TrendForce estimates the AI optical transceiver market at $26 billion in 2026, up 57% year on year, with VCSEL demand reaching 650 million units by 2031 (2026-10-08). In January this year, LightCounting forecast that the InP laser shortage would begin easing in mid-2026, a call that has yet to show up in the company's data.

On payback and inventory, three-year take-or-pay contracts offer some assurance that expansion will pay off. But the $2.35 billion of purchase commitments and Greensboro's first revenue in 2028 show that the heavy investment is stacked into 2027 and 2028. In past optical cycles, the most dangerous moment has been when capacity arrives all at once just as customers begin working through inventory. On the current timetable, that window opens around 2028, right when the deal is due to close.

Two leading signals of a peak can be tracked:

This section is a weekly signal for timing over the next few weeks. Readers holding for a year or more can skip it.

Money is indeed leaving, and not just this stock. As of 2026-11-02, the optical module sub-sector has seen 14 straight days of net outflows. Its sector money flow reading is -62, and it has spent 12 days in the weakest tier of the whole market, having dropped from the middle tier only on October 16. Net flow over the last 5 days was -771.9 million shares and over the last 20 days -1.5 billion shares (traded shares estimated from price and volume, not dollars), and the sub-sector fell 10.3% in 5 days. On OBV, a volume-momentum gauge, it ranks in only the 8th percentile of the market. Has the stock price run ahead of the money? It seems not. Money and price are moving together, and this kind of joint decline usually has not reached the "selling is over" stage.

The wider backdrop is a market on defense. On 2026-11-02 the market regime reading was -35, in bear-market territory. SPY sat 2.24% below its 20-day average, and asset rotation leaned risk-off. Money moved from XLK into real estate XLRE (strength 4.44) and materials XLB (strength 4.27). The semiconductor index SMH was among the heaviest outflow sub-sectors at -70, while the largest inflows went to IT services, chemicals, and exchanges and financial data. Macro offered no help either: net liquidity was $5.75 trillion, down 1.0% on the week, the 10-year Treasury yield was 4.94%, and liquidity conditions were cautious (FRED, 2026-11-02).

Two readings should not be left out. The sub-sector appears on a volume-price divergence accumulation list, meaning there are signs of stronger money flow while the price has not moved. That conflicts with the 14 days of net outflows, so we give it little weight. History is also neutral: across 3078 past instances of entering this tier (2021-04-12 to 2026-10-02), the median return over the next 20 trading days was 0.0%, with a 55% chance of a gain (as of 2026-11-02). So the weakest tier by itself is neither a buy-the-dip signal nor a short signal. Tech is also split internally: AI and chip stocks fell while some AI hardware names rose on strong demand, so when money returns to tech it may not go to optical modules first.

On price position, Lumentum closed 14.6% below its 200-day average and 25.8% below its 50-day average, about 40% off its 52-week high, and it had already dropped 25.3% from October 9 to 16 (Trefis). Above the current price sit layer upon layer of investors holding at a loss, who may sell into any rebound.

Long term (quarters to years): worth following. The fundamentals are solid. Revenue has doubled, operating margin has climbed above 40%, next quarter is still accelerating, and an annualized forward P/E of 28 times is not outrageously expensive. The breakup fee terms also give Lumentum's side some protection. Three costs have to be accepted. If the deal closes, you will end up holding Marvell stock, with Lumentum holders owning about 33% of the merged company, so Lumentum's own upside is diluted into a bigger business. For the next year and a half, the stock will mainly follow Marvell and the swap discount, so good earnings may not produce a good share price. And if the deal falls apart, today's valuation has to return to standalone pricing while facing whatever operating uncertainty built up along the way. So holding should be driven by checkpoints. Each quarter, check the revenue increase and gross margin, and also whether the swap discount has narrowed. Raise your confidence only when both go right, and lower your expectations if either clearly deteriorates.

Short term (days to weeks): do not chase. The market is defensive, optical modules have seen 14 straight days of net outflows, and the stock sits below both moving averages. On top of that is a brand-new swap structure that needs time to digest. The discount widening from 4.9% to 11.3% says the handover from existing growth shareholders to merger arbitrage traders is not finished. Trying to catch a rebound in this setup does not offer good odds. Wait for at least one of two things: money flows stop falling, or Marvell's stock stabilizes. Specific conditions and matching price levels follow.