Markus Moxham
AvoidMedium convictionOpen

Rambus Inc

RMBSNASDAQ

Avoid. Cyclical peak-multiple trap.

Thesis
Rambus has successfully transitioned from an aggressive litigation-dependent entity into a core fabless memory interface designer and pure-play AI data center beneficiary. Its zero-debt balance sheet, high free cash flow generation, and leading market share in DDR5 Register Clock Drivers (RCD) and companion chips provide exceptional profitability. However, the stock leaves minimal margin of safety at current multiples, as future upside heavily depends on sustaining 20%+ EPS growth through the DDR5-to-DDR6 transition while avoiding OEM in-housing, competitor share erosion, or patent portfolio expiration drag.
Key assumption
I assume that DDR5 companion chips and RCD market share will sustain high-margin product growth: I assume Rambus will maintain its 40%+ share in RCDs and successfully scale its companion chip suite (PMICs, SPD Hubs, temperature sensors) to offset slowing legacy patent licensing without diluting corporate gross margins. I assume the shift to DDR6 in 2027/2028 will not disintermediate third-party interface suppliers: I assume the soaring architectural complexity of next-generation standards prevents major DRAM module makers (e.g., Micron, Samsung) from internalizing companion/interface controller design.
What would prove me wrong
Aggressive market share losses to Montage Technology or Renesas: rapid adoption of rival RCD and companion chipsets in tier-1 cloud server supply chains, forcing price cuts and capping product revenue expansion. Secondly, in-house silicon development by major memory partners: leading DRAM manufacturers opting to design their own companion power-management and sensing silicon for future DDR6/MRDIMM platforms to capture margin internally.
Entry$101.60
Price now$87.65marked 10 Sep 2026
Price move−13.7%not a position
Benchmark+10.7%S&P 500
Worth avoiding+24.5 ppagainst holding the index
Held9 monthsopen
Download Markdown
  • American company that designs, develops and licenses chip interface technologies and architectures used in digital electronics products.
  • Founded in 1990 by engineers, public in 1997
  • Started with making a faster interface technology for DRAM, made it 10x faster. It licensed this technology to companies such as Nintendo, Intel, IBM, etc.
  • Marked by a history of patent battles with numerous chip makers
  • Long history of acquisitions
  • In 2023 sold its memory interface PHY IP business to Cadence Design Systems for $110 million
  • Sued multiple chip makers in the 2000s, RMBS found guilty of fraud and court infractions multiple times. History of court losses in terms of patents and the USPTO. Also had legal issues with the FTC, was overturned however.
  • Was accused of patent ambushing by the European Commission, as their royalties were deemed extortionate. Most of these antitrust issues came in 2000s, this being in 2007.
  • In 2010s, RMBS settled with Micron and some other chipmakers in exchange for royalties. RMBS said this represented a change in strategy to a less litigious, more collaborative approach, distancing themselves from accusations of patent trolling (overextending patent applicability and hoarding them).
  • Significant and consistent share price increase aligning with broader semiconductor market trends (AI).
  • DDR6 will come out in 2027/2028
  • Inside a PC, Rambus’ main contribution is in DDR5 memory modules used in servers and workstations. They are diversifying into companion chips, which appear on all DDR5 modules. They now develop PMICs (power regulation chips), SPD Hubs, and Temperature Sensors.
  • DDR5 memory modules are memory sticks (stuck into a motherboard). Each DDR5 memory module has DRAM chips, companion chips and RCD. RMBS makes chips and technology used in these sticks, not the sticks themselves.
  • RMBS has two parts of its business, it sells actual chips and licenses its technology to companies like Micron and NVIDIA as its patents are far-reaching into GPUs etc. SerDes PHYs for example (sold to Cadence) is used in CPUs, GPUs, etc.
  • RMBS is a market leader in DDR5 products, used in data centers (for AI).
  • With the release of DDR6, it will bring about designs. As standards evolve module makers might design in-house solutions. However, the higher complexity and new design cycle might be good for RMBS.
  • RMBS has been having some supply chain challenges despite robust supply chain management. They are also dependent on the rollouts of platforms due to royalties.
  • 51% of Rambus’ 5508 patents are active (2845). Their success ratio is 3866/5508. Number of patent filings is stable around 200, increasing to levels around 300 in some years. Of these an increasing proportion are being granted. Patents typically last 20 years from their filing date.
  • Financial strength (10): debt levels at historical lows and relative to industry.
  • Growth (9): high 3-year revenue and EPS growth rates vs industry and good vs history. 3-Year FCF and book growth rates have been solid, but average for industry and history. This is also true of the 3-5Y revenue and EPS growth estimates.
  • Momentum (6): overbought RSI, strong momentum (%).
  • Liquidity ratio: very high current, quick and cash ratios vs industry and history. Days inventory and days payable are very high vs history and average vs industry.
  • Strong buyback ratio.
  • Profitability: very strong margins vs history and competitors. Moat score of 6. Over the past 10 years, 5 profitable.
  • Value: average PE ratio relative to industry and history, same with forward PE ratio. Even worse PS and PB ratios. Average in most of the other ratios. High FCF yield vs competitors, low vs history. Same with forward rate of return.
  • Prior to 2023 no net income, some EBITDA though. Revenue growth is strong. High net income in 2023 and then diminished in 2024, one quarter in 2023 boosted it significantly (Cadence acquisition money). Net income is higher this year than in 2024 but trending downward instead of the usual upward observed in 2024.
  • Total assets is increasing, although decreased last quarter, stockholder’s equity increasing.
  • They do not report revenue streams, but one news outlet reports that licensing makes more revenue than the development side.
  • Low COGS, fairly high operating expense, 29.3% spent on R&D, 18.7% SG&A.
  • Operating revenue split between US and South Korea almost equally.
  • EPS increasing since 2021, with estimates expecting this trend to continue at a greater pace.
  • FCF positive since 2012.
  • Does not pay dividends.
  • Reached a similar share price in 2000 (exact same spike), however revenue and EPS were significantly lower.
  • Strong correlations between share price and EPS/revenue, estimates with median P/E would put the share price at $147 at the end of 2027 if estimates are correct and correlation prevails.
  • Historical P/E ratio is not high as when profitable RMBS has traded at above >200 P/E ratio. Since becoming profitable, P/E is slightly higher than in 2024 (18-32) and early 2025.
  • DCF ratio fair value $65 with EPS growth rate of 20% (default), while FCF (13.6%) puts it at $54.74. On Finbox most DCF models, suggest minimal upside (5 - 20%).
  • 9% insider ownership.

Updates

The note above is unedited. Anything that changed goes below it, dated.