Marvell agreed to buy Lumentum in an all-stock deal with an implied $115B enterprise value (~$102.5B equity). LITE holders get 4.5 MRVL shares each; Marvell holders would own ~67%; close targeted H1 2028. CEO Matt Murphy called optical connectivity a "once in a generation" architectural shift.
The tape voted instantly: MRVL $184.11, −6.3% and LITE $747.39, −11.1% intraday, while independent rival COHR rose +1.5% to $213.40. At current MRVL prices the ratio implies ~$829 per LITE share — a ~11% spread to LITE's quote — wide for a fixed-ratio deal and a direct measure of skepticism about paper value and a 17-month wait to close.
This did not land alone. The same morning Oracle filed for $10B in common stock plus $10B in senior notes — pre-flagged at its Oct 28 investor day alongside a ≥$225B FY2030 revenue guide and a promise of positive free cash flow only in H2 2028 — and Anthropic's amended S-1 showed $17B in Q3 revenue and $895.4B in purchase/lease commitments. Three paper promises in one day: equity paper (MRVL), debt+equity paper (ORCL), and off-balance-sheet compute paper (Anthropic).
The AI build has outgrown operating cash flow, so the market's marginal funding currency switched from cash to claims. Oracle is the cleanest exhibit: $125.34B in debt, S&P cut to BBB− in July (one notch above junk), 5-yr CDS at 296 (up 270% in a year, per S&P Global), its 4.95% 2031 bond yielding 7.27%, stock −33% YTD and −50% over 12 months — yet it must still fund data centers. Vendor financing, customer prepayments and hardware contributions are now explicitly part of its plan.
Anthropic makes the same point from the demand side: 9-month revenue $33.24B (+~1,200% y/y) against a $104B net loss (mostly convert/derivative revaluation) and $895.4B of committed future compute spend. Q3 growth itself decelerated — +47% q/q versus +144% in Q2 and +132% in Q1 — even as commitments compound. When $17B of quarterly revenue sits beneath $895B of commitments, someone must warehouse duration risk — today that someone is public shareholders and bondholders.
Marvell chose the cheapest warehouse: its own multiple. With a ~65× trailing P/E and beta above 2.3, issuing shares at ~$184–196 to buy 83% y/y revenue growth (LITE) and 41.7% gross margins is rational if the paper holds. The market's −6% reply says it doubts the paper holds through H1 2028.
Today's cross-asset tape is a repricing of duration, not a tech wreck. SPY $761.02 (+0.7%), AMZN $252.99 (+5.3%), GOOGL $283.79 (+1.4%) and NVDA $207.31 (+1.3%) bounced as oil collapsed (USO $159.43, −5.6%; WTI ~$99.85, −4.1%) on U.S.–Iran diplomacy hopes ahead of Nov 3 midterms, with the 10-year at 4.914% (−1bp). Gold held firm ($362.58, +1.9%) and BTC sagged at $77,442 (Fear 37). In that bounce, the two paper issuers fell 6–11% while the unencumbered optics pure-play rose. That divergence is the signal.
Second- and third-order effects most coverage misses:
My highest-conviction positioning: long COHR independence; hold LITE only if you already own it; no MRVL chase; no ORCL bottom-fish.
| Vehicle | Action | Entry | Target | Invalidation | Horizon / Conviction |
|---|---|---|---|---|---|
| COHR (flagship) | BUY | $200–215 scale | $265–280, stretch $310 | Daily close <$185 | 1–3 mo / SPECULATIVE-HIGH |
| LITE | HOLD if owned; no add | — (prior $720–780 filled) | Ratio-implied ~$830; deal stretch $950–1,000 only if MRVL reclaims $210 | Daily close <$640 | 1–3 mo / HOLD |
| MRVL | AVOID / WATCH | Revisit only >$210 reclaim with volume | — | — | WATCH |
| ORCL | AVOID | Revisit only on FCF-positive proof or <$100 washout | — | — | — |
Sizing opinion: COHR half-size starter; add second half only on a $205 hold with volume or a daily close above $225. This respects the open book: prior WATCH-ORCL $135–138 was never triggered (ORCL $123.15 — correctly avoided); MACRO-LITE-1 $720–780 filled and now capped by the ratio — thesis changed from earnings compounder to merger hold, stated here.
| Scenario | Weight | Trigger | Price path |
|---|---|---|---|
| Bull — optical scarcity repriced | 30% | COHR prints CPO/scale-out revenue + Apple/Ternus debut or AMD Nov 3 confirms AI capex holds | COHR $265–280 in weeks; LITE spread collapses toward ratio |
| Base — paper overhang lingers | 50% | MRVL grinds $175–195 into year-end; ORCL deal digests; 10Y 4.85–5.10% | COHR $215–245 chop; LITE $740–830 tied to MRVL |
| Bear — funding strike | 20% | 10Y daily >5.10% or S&P daily <$7,600 (SPY ~$760) or META <$600 book rule fires | COHR breaks $185 → $160; LITE back toward $640 kill; halve all longs per book rule |
Load-bearing assumption: that 1.6T/CPO demand is real end-demand, not pull-forward funded by commitments that later impair. If hyperscaler capex is cut post-midterms or after NVDA Nov 17, optics falls with it and cash (XLP/XLE, T-bills) was the right call.