Memory: Up to 10% Share Reduction, CPO Adds New Optionality


Memory makers are no longer just printing profits — they are starting to retire serious amounts of stock. SK hynix could ultimately shrink its share base by nearly 10%, while Sandisk’s remaining authorization equals roughly 6.7%.


Sandisk Fired First. SK hynix Just Raised the Stakes

$SanDisk Corp.(SNDK)$   kicked things off by adding $14B to its repurchase authorization, leaving $15.5B available. At its Aug. 19 closing price, that is equivalent to roughly 6.6% of the share base. The company had already repurchased $4.5B of stock in the prior quarter.

$SK hynix(SKHY)$   then raised the stakes. Its KRW40T ($28.7B) program will buy back and cancel 24.07 million shares, or around 3.3% of issued shares. That distinction matters: these shares actually disappear.

And the announced program may not be the end. Goldman estimates another KRW70T of repurchases and cancellations through the current shareholder-return period. On a static price basis, the combined program could approach a 10% reduction in shares outstanding. Goldman consequently raised its 2027 and 2028 EPS estimates by roughly 10%.

Now Samsung is entering the conversation. Reuters reported today that the company could unveil a KRW100T+ ($71.8B+) shareholder-return package later this month, including a special dividend. Korean media estimates have pushed the potential total as high as KRW150T–160T, or roughly $108B–$115B.


Then CPO Adds a New Angle

SK hynix today highlighted a CPO technology roadmap featured in Nature Electronics, extending its thinking beyond HBM toward a broader AI bottleneck: moving data efficiently between memory, processors and massive AI clusters. The architecture ultimately pushes optical interconnects closer to memory and compute, including photonic-interposer concepts linking processors and memory.

The timing is important because CPO itself is no longer just a future roadmap. Spectrum-X Ethernet Photonics, a CPO-based switching platform, is already in full production.

$Lumentum (LITE.US)$ has perhaps the cleanest direct exposure. Its ultra-high-power lasers and external laser source modules sit at a critical CPO bottleneck: providing reliable light outside a hot switch ASIC package. Lumentum is already offering dedicated UHP laser and ELS products for CPO.

$Coherent (COHR.US)$ offers broader content. It appears across lasers, laser modules, microlenses and FAU assembly. Coherent itself has demonstrated a 6.4T silicon-photonics CPO platform using its own ELS and high-power InP lasers.

$Corning (GLW.US)$ is the lower-beta picks-and-shovels option, spanning fiber alignment grooves, optical fiber, FAU assembly and shuffle boxes in the BofA map.

The bigger picture is simple: AI spending keeps moving from one bottleneck to the next.

HBM was one of the first winners. Memory is now generating enough cash to aggressively shrink share counts. With NVIDIA CPO entering production, optics increasingly looks like the next place where scarcity, pricing power and earnings revisions can emerge.



@TigerStars  @CaptainTiger  @TigerWire  @Daily_Discussion  @Tiger_chat  @Tiger_comments  @MillionaireTiger  

# 💰Stocks to watch today?(21 August)

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Report

Comment1

  • Top
  • Latest
  • littlesweetie
    ·08-20 23:32
    A near 10% shrink is huge, but memory is still a cycle game. I care more about inventory days and capex discipline than the buyback math.
    Reply
    Report