Aug. 9 at 7:59 PM
$NLST Netlist’s U.S. Patent No. 7,619,912 (the ’912 patent), titled “Memory Module Decoder,” is a foundational patent covering rank multiplication technology for server memory modules (DIMMs).
What the ’912 patent covers
At a high level, it solves a long-standing limitation in computer memory systems:
• Memory controllers and industry standards (JEDEC) historically limited how many “ranks” of DRAM chips a system could address.
• A rank is a set of memory chips that the system accesses together.
• The patent describes a memory module that includes a decoder / logic circuit. This circuit receives the system’s control signals (chip-selects, bank address signals, etc.) and intelligently routes or multiplies them so the module can physically contain more ranks of lower-density (cheaper) DRAM chips while appearing to the host system as a smaller number of higher-density ranks.
This is called rank multiplication. It lets manufacturers put more memory capacity on a standards-compliant module without forcing the CPU or memory controller to change.
Why it has been so hard to defeat and repeatedly upheld
Claim 16 (a key independent claim) has an unusually long and resilient litigation history:
• It survived multiple inter partes reexaminations started by Google, Smart Modular, and Inphi that stretched over roughly a decade.
• Those proceedings involved numerous office actions, briefs, PTAB appeals, and an appeal to the Federal Circuit, which affirmed the validity decision.
• The patent office and courts adopted a claim construction requiring that a “rank” contain two or more memory devices (consistent with industry standards at the time). That construction made it much harder for challengers to map prior art onto the claim.
• Later PTAB decisions (in IPRs filed by Samsung and Micron) found Claim 16 unpatentable. Those decisions are on appeal to the Federal Circuit (oral arguments were scheduled in the September 2026 timeframe in recent reports). Netlist and others have argued that the later PTAB panels applied a different claim construction and that the serial challenges themselves are an example of process abuse.
Juries in the Eastern District of Texas have also found the patent infringed (and in some cases willfully infringed) by major memory makers, resulting in large damages awards that included significant portions attributed to the ’912.
Why it is considered a keystone for AI data-center memory and Nvidia systems
Modern AI training and inference in data centers require very high memory capacity and bandwidth next to the GPUs. High-capacity server DIMMs (especially Load-Reduced DIMMs / LRDIMMs and high-rank modules) are critical because:
• GPUs like Nvidia’s are extremely memory-hungry. If the system cannot supply enough capacity or bandwidth efficiently, the expensive GPU sits under-utilized.
• Rank multiplication is one of the practical ways to deliver higher capacity on a module while staying within the electrical and standards constraints of the server platform.
• Many of the high-capacity memory modules used in hyperscale data centers (Google and others) and in systems that pair with Nvidia GPUs have been accused of practicing this technology.
In short, the more efficiently and densely you can pack usable DRAM capacity next to the GPU, the better the overall system efficiency and economics of AI infrastructure. Netlist has positioned the ’912 (along with related patents) as covering important implementations of that capability.
Why it is viewed as unique and original
Before Netlist’s approach (priority dates going back to 2004–2005 provisional applications), the industry largely accepted the rank limitations imposed by the memory controller and standards. Adding more ranks usually required either:
• Higher-density (more expensive) chips, or
• Changes on the host side that broke compatibility.
Netlist’s contribution was putting the intelligence (the decoder/logic) on the module itself so that more physical ranks of lower-cost chips could be controlled while the module still looked standard to the system. That combination of rank multiplication + per-device addressability + standards compliance was novel at the time and has proven durable through years of challenges.
Current status note: Netlist and Samsung reached a multi-year patent license and supply agreement in early August 2026 that settles their disputes. Litigation and appeals involving other parties (and the validity of Claim 16) continue. The patent remains a central asset in Netlist’s portfolio for high-performance server and AI-related memory technology.
This is a technical and legal summary based on publicly available patent, PTAB, court, and company materials. Patent strength and coverage ultimately depend on claim construction, specific product implementations, and final court rulings
The ptab errored in stating point 26 was unoatebtable as argued by Samsung and micron ! Now Samsung admits it is patentable abd will help nlst defeat on appeal again inky mu in Demosthenes . 912 will bevuohejdcas oatentable .( 90.% sure ) .
RPM I continue to urge you to buy as many shares as reasonable and affordable under 5 dollars. Add when you can and know where the price is going by Dec 2028 and buy .
Rpm August 9 2026 9 pm bst
I continue to pray tge lord leads me in tge tight direct for all our sajes ! Amen
