{
  "@context": "https://frontierpicks.com/schemas/dossier.v1.json",
  "ticker": "RMBS",
  "name": "Rambus Inc.",
  "url": "https://frontierpicks.com/dossiers/RMBS/",
  "json_url": "https://frontierpicks.com/dossiers/RMBS.json",
  "status": "DORMANT",
  "current_conviction": "LOW",
  "graded_conviction": "LOW",
  "archetype": {
    "code": "a2",
    "n": 2
  },
  "current_thesis": "Memory theme is ripping — Micron +68% YTD with HBM sold out through 2027 — but Rambus has decoupled, down ~42% from its 2026-06-03 high of $174.10 to $101.44 while the cycle proxy sits near records. Interface content is a DIMM-unit story, not a DRAM-price story, and the 2026-07-27 Q2 print is the binary that resolves which.",
  "invalidation_trigger": "A weekly close below $95 breaks the July shelf and the 200-day, confirming the de-rating is a unit-volume problem rather than a momentum unwind; secondary: Q2 product revenue printing under the $95M guide floor on 2026-07-27, or a 2H26 guide that fails to raise on the MRDIMM/SOCAMM2 ramp.",
  "catalyst_date": null,
  "outcome": "INVALIDATED",
  "outcome_date": "2026-07-07",
  "invalidation_fired": true,
  "themes": [
    "ai-chips-memory",
    "ai-datacenter-infrastructure"
  ],
  "tags": [],
  "sources": [],
  "notes": [
    "Fixed content per module (RCD, PMIC, SPD hub, temp sensors) levers revenue to DIMM units shipped, not DRAM ASPs — a DRAM price spike can suppress server builds while DRAM makers print records.",
    "Royalty revenue is renewal-driven and lumpy: $69.64M in Q1, $84.2M in Q2, guided $69–75M for Q3. Quarter-to-quarter royalty swings say little about product demand.",
    "Product gross margin runs at the low end of management's stated 60–65% band; a mix rotation toward product compresses blended gross margin arithmetically.",
    "Insider record predates the de-rating: COO Sean Fan sold 37,814 shares at $151.69 on 2026-05-26; CFO Desmond Lynch resigned. No disclosed offsetting open-market purchase.",
    "Balance sheet is not a variable: $824.9M cash and marketable securities at 2026-06-30, $61.2M operating cash flow in the quarter, no dilution overhang."
  ],
  "body_markdown": "## Current Thesis\n\nThe level this coverage published on 2026-08-15 as the thesis-break gave way. Rambus closed the week of 2026-08-21 at $91.24, through the $95 shelf that had held since July, after five sessions that took it from $100.85 on 2026-08-14. That is a 46.5% discount to the 52-week high of $170.66 and a three-month price change of -36.2%.\n\nWhat makes this awkward is that the operating record moved the other way. Q2, reported 2026-07-27, was the first quarter above $200M: total revenue $207.4M (+20% YoY, +15% QoQ) against $198.3M consensus, product revenue $99.2M (+22% YoY), royalties $84.2M, non-GAAP diluted EPS $0.77 against a $0.65–0.73 guide. Q3 guidance of $210–216M puts product revenue at $110–116M, another double-digit sequential step. The unit-volume mechanism this coverage worried about — a DRAM price spike suppressing server module builds while DRAM makers print records — has not appeared in a reported number.\n\nThe separation between those two facts is the whole read now. The market's objection since 2026-07-28, when a 20% revenue beat was sold 8.9% to $87.78, is earnings quality: royalties guided down to $69–75M from an $84.2M actual, product guided up, and product gross margin landing at the low end of management's 60–65% band with supply-chain cost named. Nothing dated resolves that before the Q3 print, which lands roughly 2026-10-26 on the Q1/Q2 cadence (2026-04-27, 2026-07-27). In the interim the company itself is the marginal buyer, and it has not been enough.\n\n## Bull Case\n\n- Q2 2026 (reported 2026-07-27): revenue $207.4M vs $198.3M consensus; product $99.2M (+22% YoY) at the top of the $95–101M guide; royalties $84.2M, recovered from $69.64M in Q1; non-GAAP EPS $0.77 vs a $0.65–0.73 guide.\n- Q3 guide of $210–216M with product at $110–116M means the 2H26 MRDIMM/SOCAMM2 content ramp is inside the guidance range, not a management aspiration for 2027.\n- On the 2026-07-27 call the company disclosed a design win with a tier-one U.S. hyperscaler for next-generation HBM in future AI silicon, characterised as an IP licensing opportunity rather than a product-chip win — i.e. it lands in royalties, the high-incremental-margin line, not in the mix that is currently being marked down.\n- Silicon IP pipeline is dated and shipping: PCIe 7.0 Switch IP with time-division multiplexing at 128 GT/s was introduced 2026-05-07.\n- Cash and marketable securities of $824.9M at 2026-06-30, up $38.8M sequentially on $61.2M of operating cash flow. No dilution overhang.\n\n- Valuation compressed with the tape: market cap $9.90B and forward P/E 26.95 at the 2026-08-21 close, against $10.94B and 29.79x forward a week earlier on 2026-08-14.\n\n## Bear Case\n\n- The de-rating is now eleven weeks old and has not found a shelf. Price has broken the July $95 area on a weekly close and sits below the 2026-08-14 print of $100.85 by 9.5%.\n- Q3 royalty guidance of $69–75M is a sequential decline from $84.2M. Royalties carry near-full incremental margin; product does not. Even if both lines hit the guide, blended gross margin compresses arithmetically.\n- Product gross margin printed at the low end of the stated 60–65% band in Q2 with supply-chain cost named as a contributor. Input cost pressure inside a memory upcycle is not a one-quarter problem.\n- The $100M accelerated repurchase has been live since 2026-08-05 and the shares still fell 9.5% in the week to 2026-08-21. A known non-discretionary bid absorbed without holding the level says supply is coming from somewhere larger.\n- Sell-side estimates have not caught down to price. Consensus target sits at $147.50 against a $91.24 close, and the low end of the target range polled on 2026-08-14 was $100 — above the market. Revision risk is one-directional from here.\n- The insider record is old but unhelpful: COO Sean Fan sold 37,814 shares at $151.69 on 2026-05-26, and CFO Desmond Lynch resigned. Neither has been offset by a disclosed open-market insider purchase.\n- RSI(14) at 48.8 on 2026-08-21 is mid-range. A 36.2% three-month decline that has not produced an oversold reading is a grind, and grinds do not mark lows by themselves.\n\n## Setup & Price Structure\n\nReference close 2026-08-21: $91.24. Session range that day $89.50–$93.35. The structural marker below is the 2026-07-28 post-print close of $87.78, which has held as the low of the move since the earnings reaction; below that there is no dated reference until the low end of the 52-week range.\n\nThe narrative is **dead**. The dates that make the call: the record 2026-07-27 print was sold 8.9% the following session; the memory complex has held near cycle highs while this name has fallen 46.5% from its 2026-06-03 peak; the $95 weekly shelf broke on 2026-08-21; and the news flow since has been retrospective return arithmetic (Benzinga pieces on 2026-08-14 and 2026-08-20 on what a decade-old holding would be worth) rather than anything forward-looking. That is a name whose bid has left, not one being distributed at the top. The distinction matters for what comes next: the operating narrative is intact and re-testable at the Q3 print, so this is a broken price structure sitting on top of a business still guiding up — a combination that resolves at a print, not on a chart.\n\nCrowding and positioning observables, stated as observables:\n\n- A mechanical buyer is in the tape until roughly end of Q3 2026 (the accelerated repurchase term). When it settles, that bid stops.\n- The published target range brackets nothing: consensus $147.50, the polled low $100 (2026-08-14), spot $91.24. Analyst numbers sit entirely above the market.\n- No earnings date inside 30 days. The next company-controlled event is roughly nine weeks out.\n- Retail-facing coverage has rotated to backward-looking \"what $1,000 would be worth\" formats — two such items inside seven days (2026-08-14, 2026-08-20).\n- Sector-level leverage supply continues: 2x single-stock ETFs on Everspin, SiTime and UMC were flagged 2026-08-13, evidence of leveraged product proliferation across small- and mid-cap semis generally, not an RMBS-specific flow.\n\n## Catalyst Calendar (next 30 days)\n\n- **No dated company catalyst falls inside the window to 2026-09-22.** That is itself the setup condition: the tape between here and late September is semiconductor-index beta plus the repurchase.\n- **~2026-09-24 (est.) — Micron fiscal Q4 results.** The cycle proxy. Its commentary on server build plans and DRAM contract pricing sets the assumption for DIMM units shipped, which is the variable Rambus product revenue is levered to.\n- **~2026-09-30 (est.) — accelerated repurchase term expected to complete by end of Q3 2026.** Fixes the final retired share count against 108.46M outstanding and ends the mechanical bid.\n- **~2026-10-26 (est.) — Q3 FY2026 print.** Date unconfirmed; Q1 landed 2026-04-27 and Q2 landed 2026-07-27. First check on whether the guided product-heavy mix ($110–116M product against $69–75M royalties) holds gross margin inside the 60–65% band.\n\n## What Would Change Our Mind\n\nThe structure that defined the constructive case is already gone: the $95 weekly shelf broke on 2026-08-21, so the prior frame no longer stands on its own and the name has to earn a new one at the print.\n\nConstructive again requires two things together — a weekly reclaim above $101, back over the 2026-08-14 close of $100.85 and the broken shelf, and a Q3 print around 2026-10-26 that pairs product gross margin recovering toward the top of the 60–65% band with royalty guidance above $75M. Either alone is insufficient; the objection is specifically the mix.\n\nContinuation is confirmed by a weekly close below $87, which takes out the 2026-07-28 low of $87.78 and says the de-rating has extended rather than based. A secondary condition on the same side: the repurchase term completing around end of Q3 2026 with the $95 area still unreclaimed removes the only identified non-discretionary buyer.\n\nThe theme-level flip that would end the discussion: Micron guiding server DRAM demand or 2027 build plans lower at its ~2026-09-24 report. The second-order interface name has no separate bid if the cycle proxy rolls.\n\n## Correlation Notes\n\n- The revenue mechanism is content per module — RCD, PMIC, SPD hub, temperature sensors — so the driver is DIMM units shipped, not DRAM ASPs. A DRAM price spike can compress server builds and lift Micron's realised pricing at the same time. That is the mechanical reason the two charts can separate, and they have since 2026-06-03.\n- Beta to the semiconductor index dominates the next nine weeks in the absence of company news. Idiosyncratic dispersion should be near zero until the Q3 print; if it is not, the move is flow rather than fundamentals.\n- The HBM design-win disclosure of 2026-07-27 correlates this name to the HBM4 generation timeline rather than to current DDR5 module volumes — a different, longer-dated clock than the product line the market is currently marking.\n- Royalty revenue correlates to the DRAM makers' licence renewal calendar, not to the current quarter's shipments. Quarter-to-quarter royalty swings ($69.64M in Q1, $84.2M in Q2, guided $69–75M in Q3) carry no information about product demand.",
  "first_seen": "2026-04-24",
  "last_analyzed": "2026-08-23T16:38:51+00:00",
  "last_synthesized": "2026-08-23",
  "last_update_source": "watchlist_research",
  "license": "Content © FrontierPicks. Cite the canonical URL."
}