Rambus
Hybrid, capital-light. (1) Fabless product sales of DDR5 server/client DIMM chipsets — RCD plus companion chips (MRCD/MDB, data buffers, PMIC, SPD Hub, temperature sensor) — manufactured by foundries. (2) High-margin patent-royalty stream on memory-interface standards (Samsung, SK hynix, Micron). (3) Silicon-IP + security-IP licensing (HBM/PCIe/CXL controllers, root-of-trust). FY25 mix ≈ 49% product / 40% royalty / 11% contract & other (fact, FY25 10-K release).
- 2026-08-04This market capitalisation previously read ~$13.4B (as of Jun 26 2026). Restated to $10.00B on this refresh, roughly 25% lower.
- 2026-08-04This share price previously read ~$123.50 (as of Jun 26 2026). Restated to $92.21 on this refresh, roughly 25% lower.
Sources — 13 figures with citations
- Q2 FY26 revenue by line and totalsfiled2026-06-30Total revenue $207.385M vs $172.209M Q2'25 (+20.4%); product $99.154M vs $81.325M (+21.9%); royalties $84.261M vs $68.607M (+22.8%); contract and other $23.970M vs $22.277M. H1'26 total $387.574M vs $338.873M; H1 product $187.156M; H1 royalties $153.903M. Q1 FY26 royalties therefore $69.642M and Q1 product $88.002Msec.gov — 8-K Item 2.02 Exhibit 99.1 filed 2026-07-27, Condensed Consolidated Statements of Income. Q1 FY26 lines derived by subtracting Q2 from the six-month column: 153.903 - 84.261 = 69.642 royalties; 187.156 - 99.154 = 88.002 product. Hence royalties +21.0% QoQ and product +12.7% QoQ.
- Q2 FY26 gross profit, gross margin and product-level marginderived2026-06-30Total cost of revenue $41.973M (product cost $39.575M, contract/other cost $0.778M, acquired-intangible amortization $1.620M); gross profit $165.412M = 79.76% of revenue. Q2'25: cost $34.770M, gross profit $137.439M = 79.81%. Product gross margin = ($99.154M - $39.575M) / $99.154M = 60.1%sec.gov — Dollar figures are filed in the income statement; the percentages and the product-level margin are computed from them. Royalties carry no cost line, i.e. ~100% gross margin, which is why blended GM is a mix output.
- Q2 FY26 operating expenses, operating income and marginsfiled2026-06-30Operating expenses $92.681M vs $74.446M (+24.5%): R&D $51.099M (vs $46.331M), SG&A $38.262M (vs $28.115M), restructuring $3.320M (vs nil). GAAP operating income $72.731M = 35.1% (printed as 35%) vs $62.993M = 36.6% (printed 37%). Non-GAAP operating income $93.661M = 45.2% (printed 45%) vs $79.031M = 45.9% (printed 46%)sec.gov — Income statement plus the Supplemental Reconciliation of GAAP to Non-GAAP Results. Non-GAAP add-backs: stock comp $15.886M, acquired-intangible amortization $1.620M, restructuring $3.320M, facility closure $0.104M.
- Q2 FY26 net income, EPS and tax ratefiled2026-06-30Pre-tax income $79.489M; provision for income taxes $11.884M (14.95% effective); net income $67.605M vs $57.935M. GAAP diluted EPS $0.61 vs $0.53; non-GAAP net income $84.352M and non-GAAP diluted EPS $0.77 vs $0.62. Basic shares 108.428M, diluted 109.977M. Interest and other income net $6.758Msec.gov — Income statement and non-GAAP reconciliation. The company states a 16% assumed long-term non-GAAP tax rate for 2026 versus 20% for 2025.
- Net cash, cash flow and capex at 2026-06-30filed2026-06-30Cash, cash equivalents and marketable securities $824.9M, up $38.8M vs 2026-03-31; composed of $87.720M cash and equivalents plus $737.228M current marketable securities. Q2 operating cash flow $61.2M; capex $12.2M; $14.3M paid for taxes on net share settlement of equity awards. No debt line on the balance sheet — only EDA-tool software licence liabilities ($14.815M current, $13.367M long-term) and operating-lease liabilitiessec.gov — Stated in the liquidity paragraph and the Condensed Consolidated Balance Sheets. Implied Q2 free cash flow = $61.2M - $12.2M = ~$49.0M, a ~23.6% FCF margin (derived).
- Q2 FY26 beat versus company guidancefiled2026-07-27Revenue $207.385M against a guided range of $192M-$198M; non-GAAP diluted EPS $0.77 against a guided range of $0.65-$0.73seekingalpha.com — The $192-198M revenue range is restated inside the Q2 8-K exhibit itself ('exceeded the Company's guidance range of $192 million to $198 million'); this source captures the original Q2 guidance including the $0.65-$0.73 non-GAAP EPS range issued with Q1 FY26 results, which the Q2 release references but does not restate numerically.
- Q3 FY2026 guidancefiled2026-07-27Total revenue $210M-$216M; product revenue $110M-$116M; royalties $69M-$75M; contract and other $25M-$31M; GAAP operating costs and expenses $137M-$133M; non-GAAP $119M-$115M; GAAP diluted EPS $0.59-$0.67; non-GAAP diluted EPS $0.75-$0.82. Assumes 18.7% GAAP / 16% non-GAAP tax rate, 110M diluted shares, $7M interest and other income; excludes $15.7M stock comp, $1.5M intangible amortization, $0.8M othersec.gov — '2026 Third Quarter Outlook' table. Royalty midpoint $72M vs $84.261M in Q2 = -14.6% QoQ; product midpoint $113M vs $99.154M = +14.0% QoQ.
- Implied Q3 blended gross margin from guidance mixderived2026-07-27~78.4%, versus 79.8% delivered in Q2 FY26 — roughly a 135bps mix-driven declinesec.gov — Arithmetic at Q3 guide midpoints: product $113M x 60.1% GM = $67.9M gross profit; royalties $72M at ~100% = $72.0M; contract/other $28M less ~$0.8M cost = $27.2M. Total ~$167.1M on $213M revenue = 78.4%. Assumes product and contract cost ratios hold at Q2 levels — an assumption, not a company disclosure.
- Inventory build and balance-sheet compositionfiled2026-06-30Inventories $74.836M at 2026-06-30 vs $44.098M at 2025-12-31 (+69.7%); accounts receivable $144.094M vs $137.476M; unbilled receivables $22.768M; total assets $1,612.328M; total stockholders' equity $1,466.659M; goodwill $286.812M; deferred revenue $21.179M (down from $29.980M)sec.gov — Condensed Consolidated Balance Sheets. The inventory build is ~1.9x the quarter's $39.575M product cost of revenue (derived).
- FY2025 total revenue (base for TTM math)filed2025-12-31$707.630Mdata.sec.gov — SEC XBRL company-concept API, FY2025 annual fact from the FY2025 Form 10-K (2025-01-01 to 2025-12-31).
- TTM revenue through 2026-06-30derived2026-06-30$756.3Msec.gov — FY2025 $707.630M - H1'25 $338.873M + H1'26 $387.574M = $756.331M. H1 figures from this exhibit; FY2025 from the 10-K XBRL fact above.
- Share price and market capitalisationmarket2026-08-03$92.21 closing price on 2026-08-03 (+$1.18 / +1.30% that day); market cap $10.00B; 108.46M shares outstandingstockanalysis.com — Regular-session close, not an intraday high. Cross-check: $92.21 x 108.46M = $10.00B, matching the quoted market cap. Enterprise value = $10.00B - $0.825B net cash = ~$9.18B, or ~12.1x TTM revenue (derived). Retrieved 2026-08-04.
- Consensus and post-print market reaction contextmarket2026-07-29Non-GAAP EPS $0.77 vs consensus of roughly $0.71-$0.73; shares fell after the print on concerns about the guided sequential royalty decline and the mix shift toward the lower-margin chip businesssahmcapital.com — Third-party commentary and sell-side consensus, not company-reported. Used only for the consensus level and the direction of the reaction. Note this class of coverage misstates which quarter the $69.64M royalty figure belongs to — the filing is authoritative: $69.642M was Q1 FY26 and $84.261M was Q2 FY26.
The thesis on this name
State of the Memory Supercycle
The DDR5 register-clock-driver (RCD) / interface-chip and memory-IP licensing play — content on every DDR5 RDIMM via the RCD/companion-chip trio, plus a high-margin IP-royalty stream tied to memory interface standards. As the cycle pushes high-capacity server RDIMMs (the primary AI procurement target per TrendForce) and DDR5 speeds climb, Rambus's per-module content and royalty base grow with both unit volume and ASP. An asset-light, royalty-heavy way to ride DDR5 server-memory volume without commodity-price exposure on the downside — the de-rate in a glut is cushioned by recurring IP. Mispriced as a sleepy IP name versus its DDR5-server attach.
Earnings, margins, COGS & capex
Rambus has re-rated into a structural AI-memory-content story: FY25 total revenue rose 27% to $707.6M on a 41% surge in DDR5 product revenue (RCD share gains to mid-40%), while the royalty book is a stable, ~100%-margin annuity (~$279M FY25). The model is exceptionally capital-light — ~80% GAAP gross margin, ~37% GAAP operating margin, ~$360M operating cash flow, and a $786M net-cash balance sheet with no debt. The near-term tension is that royalty revenue is flat-to-down (Samsung step-down, contract timing) even as product compounds, so blended growth and margin mix depend on the product ramp out-running the royalty drag.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~20¢ is cost of goods and ~45¢ operating expense, leaving ~35¢ of operating profit (~33¢ net).
Revenue trend
Margins
flat→down slightly as lower-margin product mix grows
down (opex +16% YoY in Q1, R&D ramp for HBM4/CXL)
down
stable-high
up — strong cash conversion
COGS structure
COGS is concentrated in the product segment: foundry wafer + assembly/test for the DDR5 chipset (RCD, MRCD/MDB, data buffers, PMIC, SPD, TS), plus amortization of acquired/internal IP. Royalty and most IP-license revenue carry near-zero incremental COGS, so blended gross margin is set by the product-revenue share — as product climbs toward half of revenue, blended GM drifts from the high-90s (IP) toward the company's stated 60–65% product model.
Capex
Very low — ~$35.3M FY25 (~5% of revenue, ~$8–10M/quarter). Funds design tools/EDA, test equipment, and lab/IT, not fabs (fabless model). The real 'investment' is R&D opex, not capex.
Latest earnings
Beat on both lines. Revenue $207.385M exceeded the company's own $192-$198M guidance range (stated in the release). Non-GAAP diluted EPS $0.77 exceeded the guided $0.65-$0.73 range and consensus of roughly $0.71-$0.73 (third-party). Management called it 'a new all-time high in revenue and non-GAAP earnings'. Notably, the stock still de-rated after the print.
Q3 FY2026 guidance is above Q2 on revenue but mixed underneath: total revenue $210-$216M (midpoint +2.7% QoQ, a sharp deceleration from +15% QoQ in Q2); product revenue $110-$116M (midpoint +14% QoQ, continued ramp); royalties $69-$75M (midpoint -14.6% QoQ from $84.261M — a sequential step DOWN in the ~100%-gross-margin line); contract and other $25-$31M. GAAP EPS $0.59-$0.67; non-GAAP EPS $0.75-$0.82. Assumes a 16% non-GAAP tax rate, 110M diluted shares, $7M of interest and other income. Rambus gives quarterly guidance only — no full-year outlook.
- Total revenue
- $207.385M, +20.4% YoY — first quarter above $200M
- Product revenue
- $99.154M record, +21.9% YoY / +12.7% QoQ; ~60.1% product gross margin
- Royalties revenue
- $84.261M, +22.8% YoY and +21.0% QoQ from $69.642M in Q1 FY26 — but guided to $69-75M in Q3
- Licensing billings
- $84.1M (operational metric: amounts invoiced to licensing customers, adjusted for advance payments on variable agreements)
- Non-GAAP EPS
- $0.77 all-time high, +24.2% YoY; GAAP $0.61
- Non-GAAP operating margin
- 45.2%, down ~70bps YoY from 45.9%
- Net cash
- $824.9M cash + marketable securities, +$38.8M QoQ, zero debt
- Restructuring
- $3.320M restructuring charge plus $0.104M facility-closure cost in Q2 FY26 — no comparable charge in Q2 FY25
- Inventories
- $74.836M at 2026-06-30 vs $44.098M at 2025-12-31 (+69.7%) — a build ahead of the DDR5 product ramp, and a working-capital drag
- Product roadmap
- Complete chipsets for DDR5 9600 server and client memory modules; PCIe 7 Switch IP
- Effective tax rate
- 14.95% GAAP in Q2 FY26 ($11.884M on $79.489M pre-tax); 16% assumed non-GAAP rate for FY2026 vs 20% for FY2025
Growth drivers
- DDR5 RCD share gains (mid-40% exiting 2025, up from early-40% in 2024) plus the standard DDR4→DDR5 server transition expanding the per-DIMM interface-chip TAM
- Companion-chip attach — PMIC, SPD Hub, temperature sensor and data buffers multiply silicon content per RDIMM beyond the RCD
- AI-memory content step-ups — MRDIMM (MRCD/MDB at 12,800 MT/s, server launches 2026), DDR5 9600 client chipset for AI PCs, and LPDDR5X SOCAMM2 server modules
- Interface-IP licensing into AI accelerators — HBM4/HBM4E memory-controller IP (claims industry-fastest/-first), plus PCIe 6/7 and CXL controllers
- Security-IP attach (root-of-trust, crypto cores) into data-center and automotive SoCs
- Stable patent-royalty annuity (Samsung, SK hynix to mid-2034, Micron) funding the R&D engine
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-02-18. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
Rambus is executing on both engines at once — record product revenue (+21.9%) and royalties up 21% sequentially — and the print beat its own revenue and EPS guidance while the stock de-rated ~25% from late June, so an all-time-high earnings quarter is now being bought at roughly 30x with $825M of net cash and no debt.
- Both revenue engines fired: product $99.154M (record, +12.7% QoQ) and royalties $84.261M (+21.0% QoQ from $69.642M). Royalties had been DECLINING (-6% YoY in Q1 FY26); that reversed to +22.8% YoY this quarter.
- First quarter above $200M and a beat against the company's own $192-198M guide — Rambus guides quarter-by-quarter with no annual number, so beating its own near-dated range is a higher-quality signal than beating a sell-side estimate.
- Margin structure is intact: gross margin 79.8% flat YoY and non-GAAP operating margin 45.2% versus 45.9% — a ~70bps give-back while revenue grew 20%. This is not a company losing pricing power, it is one absorbing a heavier product mix.
- Balance sheet is a genuine option: $824.9M of cash and marketable securities, up $38.8M in the quarter, against zero debt. That is ~$7.60/share, ~8.2% of the price, and $7M/quarter of interest income is already flowing through guidance.
- The product roadmap is aimed at the right bottleneck: complete chipsets for DDR5 9600 server and client modules plus PCIe 7 Switch IP. Content per server rises with every speed-grade transition, and Rambus sits on the interface rather than competing on DRAM economics.
- Q3 guidance is above Q2 at the midpoint ($213M vs $207.4M revenue; $0.785 vs $0.77 non-GAAP EPS) with product guided +14% QoQ — the chip ramp is accelerating even as royalties normalise.
- The de-rating did the work: ~$123.50 in late June to $92.21 on 2026-08-03 is roughly -25%, against an all-time-high non-GAAP EPS quarter. Backing out net cash, enterprise value of ~$9.18B is ~12.1x TTM revenue versus a business compounding revenue in the high teens with 45% non-GAAP operating margins.
The Q3 guide is the tell: royalties are guided DOWN 14.6% sequentially from $84.261M to a $72M midpoint while product is guided UP 14%, so the mix is shifting from the ~100%-gross-margin licence line into the ~60%-gross-margin chip line — mathematically eroding blended margin, adding inventory and cyclicality, and turning a capital-light royalty compounder into a memory-cycle component supplier.
- Quantified mix erosion: at the Q3 guide midpoints (product $113M at ~60% GM, royalties $72M at ~100%, contract/other $28M) blended gross margin computes to roughly 78.4%, versus 79.8% delivered in Q2 — a ~135bps step down guided in a single quarter, driven purely by mix.
- Royalty revenue is lumpy, not a run-rate. It went -6% YoY in Q1 FY26, +22.8% YoY / +21.0% QoQ in Q2, and is guided -14.6% QoQ for Q3. A line that swings 20 points quarter to quarter cannot carry a premium multiple on a trailing print.
- Operating leverage is running backwards: opex grew +24.5% YoY against +20.4% revenue growth, so GAAP operating margin fell 36.6% to 35.1% and non-GAAP 45.9% to 45.2%. R&D alone was $51.099M, up 10.3% YoY, and SG&A $38.262M, up 36.1%.
- A new $3.320M restructuring charge plus $0.104M of facility-closure costs appeared this quarter with no prior-year comparable. Restructuring inside a record quarter is a signal that some part of the portfolio is not working.
- Working capital is now absorbing the growth: inventories $74.836M versus $44.098M at 2025-12-31 (+69.7%), roughly 1.9 quarters of product cost. FCF margin fell to ~23.6% in the quarter from ~46% in FY2025 — the asset-light thesis is diluted by a physical chip business that carries inventory and takes DRAM-cycle risk.
- The GAAP effective tax rate of 14.95% and the 16% non-GAAP assumption (down from 20% for FY2025) flatter YoY EPS growth. Part of the +24.2% non-GAAP EPS increase is a lower tax rate rather than operating improvement.
- The market disagrees with the print: shares closed at $92.21 on 2026-08-03, roughly 25% below the ~$123.50 level of late June, on a quarter that beat guidance on both lines. Even after that, ~30x annualised non-GAAP EPS is not a distressed multiple.
- No annual guidance. Rambus guides one quarter at a time, so there is no company-provided visibility beyond September — a structural information disadvantage when the debate is precisely about the 2027 royalty and DDR5 trajectory.
What it is worth
Blend: forward P/E + EV/FCF cross-checked vs sell-side targets. At ~$123.50 (Jun 26 2026) RMBS trades ~P/E 62x / P/S 20x — a growth/scarcity premium for memory-adjacent IP. Anchor on ~$2.50–2.90 FY26 non-GAAP EPS (Q2 guide $0.65–0.73) and a 35–50x multiple reflecting capital-light ~46% FCF margins.
~$95–110 (multiple compresses toward 30–35x on royalty decline/mix-margin erosion or a DRAM-cycle slowdown)
~$145 (consensus avg target
~50x non-GAAP EPS, product offsets royalty drag)
~$170+ (sustained 40x+ on accelerating product + HBM4/CXL IP revenue; revisits the Jun-2026 all-time high)
Re-rated on the AI-memory-content thesis; the multiple already discounts product compounding, so upside needs royalty stabilization + HBM4/CXL revenue to show up. Not financial advice.
SWOT
Strengths
- Content-on-every-DIMM position — Rambus RCD/companion silicon ships on essentially every DDR5 server RDIMM, a 3-vendor oligopoly with mid-40% RCD share (fact, exiting 2025)
- Two-engine model — ~100%-margin royalty/IP annuity funds the product R&D, giving ~80% gross margin and ~46% FCF margin (derived, FY25)
- Fortress balance sheet: $786M net cash, no debt, $360M annual operating cash flow (fact, FY25/Q1 FY26)
- Standards-aligned IP moat — foundational memory-interface patents licensed by Samsung, SK hynix (to 2034), Micron — hard to design around (fact)
- First-mover on next-gen interfaces — industry-first/-fastest HBM4/HBM4E controller IP, MRDIMM 12,800, DDR5 9600 — leverage to the AI build-out (fact)
Weaknesses
- Royalty revenue is flat-to-declining (-6% YoY Q1 FY26) — the highest-margin line is not growing, dragging blended growth/margin
- Mix shift compresses margin — as lower-margin product rises toward ~half of revenue, blended gross margin drifts down from IP-era highs (derived)
- Customer concentration in royalties — a handful of DRAM makers (Samsung, SK hynix, Micron) drive the annuity; renewal terms reset the run-rate (fact)
- Product COGS exposed to foundry/substrate pricing and DRAM-cycle volume — Rambus controls neither (analysis)
- Premium valuation leaves no room for error (P/E ~62x, P/S ~20x) — operational stumbles are punished hard (fact, Q1 reaction)
Opportunities
- AI raises DRAM content per server — more DIMMs, MRDIMM/HBM/CXL all add Rambus silicon and IP dollars per box (analysis)
- HBM4/HBM4E controller-IP attach into AI accelerators (a market currently dominated by GPU/ASIC vendors needing memory subsystems) (fact/analysis)
- CXL memory expansion and PCIe 6/7 controller IP open a new interconnect TAM beyond DDR (fact)
- Client/AI-PC chipsets (DDR5 9600, CKD) extend the interface-chip franchise from server into PC (fact)
- Capital-return optionality: net cash + strong FCF support buybacks/tuck-in IP M&A (analysis)
Threats
- Montage Technology and Renesas compete directly in DDR5 RCD/companion chips — share is contestable each generation (fact)
- Memory-cycle cyclicality — DRAM downturns cut DIMM units and pressure both product volume and royalty bases (analysis)
- Royalty cliff/reset risk — large license agreements have step-downs and finite terms; a renewal on worse terms resets the annuity (fact, Samsung step-down precedent)
- Geopolitics/export controls on China and the DRAM supply chain could disrupt customers and end demand (analysis)
- If AI memory standardizes around vertically-integrated solutions (in-house controllers at hyperscalers/DRAM makers), IP attach could be squeezed (analysis)
Moats, dependencies & bottlenecks
Moats
multi-year license terms (SK hynix to mid-2034), foundational to DRAM signaling Samsung, SK hynix, Micron all licensed; hard to design around standardized interfaces (fact)
must re-win each DDR generation Only 3 credible RCD vendors (Rambus, Montage, Renesas); Rambus mid-40% share, but contestable (fact)
first-mover, but Synopsys/Cadence compete in IP Claims industry-first/-fastest HBM4 controller; design-win stickiness once embedded in an SoC (fact/analysis)
RCD + companion chips are qualified per-platform by module makers and OEMs; requalifying a competitor is costly (analysis)
Adjacent attach into data-center/automotive SoCs; smaller, supplementary moat (analysis)
Dependencies
SK hynix, Micron) as royalty payers AND module-chip pull-through Concentrated; royalty terms reset on renewal and DIMM volumes set product demand (fact)
Memory-module / server OEMs qualifying Rambus RCD + companion chips Design-in per platform; share contestable vs Montage/Renesas each generation (analysis)
TSMC/GlobalFoundries, assembly/test) Fabless — no fabs; exposed to wafer/substrate pricing and allocation (analysis)
Roadmap and patent relevance ride standards adoption timing (analysis)
DIMM units and royalty bases swing with the memory cycle and hyperscaler capex (analysis)
Customer base and end-demand exposed to U.S.-China trade actions (analysis)
Advantages
- Capital-light — ~5% capex/revenue, ~46% FCF margin, $786M net cash, zero debt — funds R&D and returns without dilution (fact/derived)
- Dual revenue engine — product + ~100%-margin royalty/IP — smooths the DRAM cycle and self-funds growth
- Embedded on nearly every DDR5 server RDIMM via the RCD/companion-chip franchise; rising silicon content per DIMM
- First-mover interface IP (HBM4/HBM4E, CXL, PCIe 6/7) positioned for AI-accelerator memory subsystems
- High switching costs from per-platform qualification and standards-embedded patents
Weaknesses
- Royalty/IP annuity (highest margin) is declining, not growing (-6% YoY Q1 FY26)
- Blended gross/operating margin compressing as lower-margin product mixes up (non-GAAP op margin 46%→42% YoY)
- Customer concentration: a few DRAM makers drive both royalties and chip pull-through
- Foundry-dependent product COGS with no control over wafer/substrate pricing or DRAM-cycle volume
- Premium valuation (~62x P/E) magnifies any execution or guidance disappointment
- AI upside (HBM4/CXL) still early-stage and not yet a material, proven revenue line
Bottlenecks
- Royalty growth bottleneck — the highest-margin line is flat-to-down (-6% YoY), capping blended top-line and margin upside until product fully offsets it
- Foundry/substrate supply and allocation for the DDR5 chipset — Rambus controls neither cost nor capacity (fabless)
- RCD share must be re-won each DDR generation against Montage/Renesas; a single generation loss compresses the product engine
- AI-IP monetization timing — HBM4/CXL revenue is design-win-then-royalty, so it lags the AI capex it's levered to
- Margin mix mechanics: scaling lower-margin product dilutes the ~80% blended gross margin investors anchor on
Top signals & trends
Top signals
Holding/gaining vs Montage & Renesas validates the product engine; share loss would break the thesis (fact)
-6% YoY now; a renewal at lower terms resets the annuity — the single most important margin variable (fact)
These are the AI-content-per-server step-ups; design wins convert to revenue in 2026+ (fact)
Real optionality but early; watch for named accelerator/customer wins, not just product announcements (analysis)
Tells whether scale/ASP offset the mix dilution to blended GM (analysis)
Q1 in-line-but-EPS-miss drove a ~6–11% drop — the multiple amplifies misses (fact)
Trends
Directly expands Rambus silicon + IP dollars per box — the core long thesis (analysis)
Each generation expands the interface-chip TAM and resets share competition (fact/analysis)
New controller-IP attach beyond traditional DDR DIMM franchise (fact)
Opens an interconnect-IP TAM adjacent to memory (fact)
Up-cycle (AI) is a tailwind now; a downturn cuts DIMM units and royalty bases (analysis)
China and DRAM-supply-chain restrictions are a tail risk to customers and end demand (analysis)
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Foundry for advanced-node interface chips (fabless model)
Foundry partner for mature/specialty-node analog/mixed-signal chips
Outsourced packaging & test for the DDR5 chipset
Design tools/IP inputs — also competitors in IP (coopetition)
DRAM maker — both a patent-royalty licensee and a pull-through for module chips. KRX-listed; named for analysis only.
DRAM/HBM leader — royalty licensee (extended to mid-2034) and HBM-ecosystem customer. KRX-listed; analysis only.
U.S. DRAM maker — royalty licensee and module-chip pull-through; key HBM customer base
HPE, Supermicro, hyperscalers) RDIMM/server buyers whose qualification drives RCD + companion-chip volume
CXL, PCIe IP licensees) License interface/controller IP into accelerators & data-center silicon
Primary DDR5 RCD/companion-chip rival; Shanghai-listed (688008.SH). Largest direct threat to the product franchise's share. Named for analysis only — not a recommendation.
TSE-listed (6723.JP); via IDT acquisition competes in DDR5 RCD/memory-interface chips. Third leg of the RCD oligopoly.
Largest silicon-IP vendor — competes in interface IP (HBM, PCIe, CXL controllers/PHY). Direct rival to the IP-licensing engine.
Major IP/EDA vendor with memory + interface controller/PHY IP (HBM, DDR, PCIe); competes for AI-accelerator design wins.
Connectivity/CXL & PCIe retimer and memory-fabric silicon for AI servers — overlaps on the CXL/data-center-interconnect opportunity.
Broader data-center connectivity & custom-silicon players that overlap on interconnect IP; less direct on RCD but compete for AI-infrastructure budget.