CipherProof @CipherProof

$SKHY Is Expanding Beyond HBM SK Hynix is developing Processing-in-Memory (PIM), which processes data inside or near memory to reduce data movement, latency, and power consumption The company claims PIM can deliver 300× the memory capacity of SRAM in the same chip area and is exploring hybrid bonding to stack compute and memory dies without sacrificing density Meanwhile, High Bandwidth Flash (HBF) combines 3D NAND density with HBM-style interconnects to deliver higher capacity at lower cost, targeting long-context and large-batch inference SK Hynix is also developing SALT-KV, software that distributes KV cache across HBM, DRAM, and SSDs to reduce dependence on expensive memory

At AI Infra Summit 2026, @SKhynix presented a “New Spectrum” of solutions for evolving AI workloads. From HBF and PIM to SALT-KV, discover how SK hynix is advancing AI infrastructure for the inference and Agentic AI era. 🔗news.skhynix.com/en/ai-infra-su… #SKhynix #AIInfraSummit

keyboard_arrow_left Previous keyboard_arrow_right Next

Trump calling @JensenHuang live during @theallinpod interview. “The grt thing about Jensen is that he can create the best AI chip in the world but can’t figure out how to put me on speaker phone.” They agree that AI is safe, grt for jobs, economic growth & natl security. 😂🇺🇸🚀

Advanced memory is foundational to the next generation of AI.  Congratulations to @SKhynix on this milestone in Indiana and their continued investment in U.S. semiconductor manufacturing.

SK hynix @SKhynix

4 weeks ago

A new chapter for AI memory begins in Indiana. @SKhynix has broken ground on its first U.S. #HBM production base, strengthening the U.S. AI ecosystem and Korea-U.S. cooperation. 🔗 Read more: news.skhynix.com/en/groundbreak… #SKhynix #AIMemory

8 18 238 78K 23

A new chapter for AI memory begins in Indiana. @SKhynix has broken ground on its first U.S. #HBM production base, strengthening the U.S. AI ecosystem and Korea-U.S. cooperation. 🔗 Read more: news.skhynix.com/en/groundbreak… #SKhynix #AIMemory

$SKHY Very Interesting. SK Hynix has started a research & development in CPO. Rumor is: $MU is also following suit on Optical Interconnects. Seems like AI Memory Roadmap is far from done. “SK hynix is moving from a memory supplier to a bigger role in the AI infrastructure. According to the company, it has teamed up with leading researchers worldwide to lay out a roadmap for co-packaged optics (CPO) in the journal Nature Electronics.” “As noted by SK hynix, in the long term, CPO could extend optical interconnects directly to the memory interface, overcoming the physical limits of conventional packaging. Thus, its proposed optics-centric architecture uses a photonic interposer to directly link memory and processors, improving system-wide data movement efficiency.” “Notably, beyond its HBM leadership, SK hynix’s paper tackles data bottlenecks at the rack and pod levels, outlining an optics-centric architecture that links memory and processors through optical interconnects.” “As hyperscale AI clusters scale to thousands of GPUs and HBM stacks, conventional copper links are hitting their limits, with higher speeds and longer distances driving up power consumption, latency and signaling complexity, SK hynix explains, adding that this is creating a new bottleneck: the bandwidth wall, which increasingly constrains data movement across AI systems.” “Against this backdrop, CPO addresses the challenge by integrating optical transceivers (TRx) directly into the processor package, shortening high-speed electrical paths while using optical links for longer-distance data transfer.”

Published in Nature Electronics, a new paper from @SKhynix and University of Virginia researchers explores how CPO can overcome the “bandwidth wall” through optical connectivity. 🔗 Read the full story below. news.skhynix.com/en/cpo-in-natu… #SKhynix #CPO #AIInfrastructure

keyboard_arrow_left Previous keyboard_arrow_right Next

Cathie Wood says to avoid memory stocks like $MU $SNDK and $SKHY. If you were not bullish Memory stocks, you should be now. Cathie Wood is even better than inverse Cramer. Her $ARKK flagship innovation ETF is flat YTD, while AI stocks have ripped. Mrs. Innovation missed the biggest tech innovation paradigm. She was a one hit wonder during the 2021 bubble and hasn't had a win since, $ARKK is still down 50% from the highs.

Shay Boloor @StockSavvyShay

a month ago

Cathie Wood says she avoids memory stocks like $MU and $SKHY because she still sees HBM as the most cyclical and commoditized part of the semiconductor stack. Her bigger point is that inference architectures from $CBRS and $NVDA via Groq can reduce or eliminate the need for HBM

356 144 1K 704K 534

Spoke to a sell side analyst today from Asia who covers memory stocks so we talked about Samsung, Kioxia, SK Hynix and SK Square. SK Square is the holding company that owns 20% of SK Hynix and trades at a -45% discount to the NAV. After this conversation, I'm wondering if I'm still not bullish enough on memory stocks... more specifically SK Hynix and SK Square. Using current sell side estimates and not including cash or any potential stock buybacks, SK Hynix is trading at: 4.7x 2026 EPS 3.6x 2027 EPS 3.3x 2028 EPS Here's the wild part... this analyst's EPS estimates are 4% higher for 2026, 40% higher for 2027 and 70% higher for 2028. Which means... based on this analyst's estimates, SK Hynix is trading at: 4.6x 2026 EPS 2.6x 2027 EPS 1.9x 2028 EPS Remember, this doesn't include current cash or the cash that will be generated over the next few years which will be used to buyback stock which means the actual EPS numbers would be higher than current estimates which also means the multiples mentioned above would be materially lower. NFA. DYOR. *We are long memory stocks at @FirstWaveFund including SK Square which we believe is the best way to get exposure to SK Hynix.

Jonah Lupton @JonahLupton

3 months ago

Every couple of months I post our top positions at @FirstWaveFund and since we sent our May investor letter a few days ago I don’t mind sharing those positions... here are the top 12 in alphabetical order: AAOI, ALAB, APP, CRDO, HIMS, HROW, IREN, MELI, MU, NBIS, RDDT, TMDX All

80 262 726 2.2M 795

Memory stocks are going higher and this table is the clearest proof yet that the entire industry has changed at a structural level (Save this). This table shows how Micron, Samsung, SK Hynix, an unnamed US NAND maker and Kioxia each disclosed massive long term agreement programs on their second quarter 2026 earnings calls and the details reveal a memory industry that no longer behaves like the boom bust commodity business it's been for decades. Micron signed 16 Strategic Customer Agreements covering roughly 20% of its DRAM volume and about a third of its NAND volume, with 14 of those deals worth a combined $100 billion in minimum contracted revenue running from 2026 through 2030. These are take or pay contracts, which means customers are contractually obligated to buy specific volumes and pay for them regardless of whether they actually take delivery, a structure Micron's own executives describe as non cancellable. Samsung is running an almost identical playbook at even larger scale. Samsung finalized agreements with its top five global data center customers and is in late stage talks with five more and once complete, these deals are projected to cover 60% to 70% of Samsung's total memory production capacity. Samsung has already collected about a quarter of its total expected deposits from customers and some contracts include price floors that guarantee minimum pricing even if the broader market weakens. SK Hynix has locked in roughly 10 agreements with its key customers, including a three year DDR5 supply deal with Microsoft and a discussion with Google for up to a five year general-purpose DRAM contract that could extend to seven years if tied to next-generation HBM supply. SK Hynix is also structuring advance payments of 10% to 30% of total contract value on these deals, which is a dramatic jump from the sub-5% advance payments that were standard just a year or two ago. The scale of this shift becomes clearer when you zoom out. South Korea's presidential adviser confirmed that SK Hynix and Samsung have collectively committed to $950 billion in long term memory supply partnerships with US tech companies, split between $750 billion from SK Hynix, largely tied to Nvidia and $200 billion from Samsung's partnership with Broadcom, both running through 2030. Even the US NAND maker referenced in the table has locked in 8 long term agreements guaranteeing a minimum of $93.9 billion in total revenue at floor pricing, and Kioxia is targeting 50% long term agreement coverage by 2028. Why does this actually make memory stocks a better investment than in past cycles? In the old memory model, customers pulled back the moment demand softened, leaving suppliers stuck with excess inventory, collapsing prices, and vanishing margins during every downturn. Take or pay contracts flip that dynamic entirely, since customers are now financially penalized for walking away, which removes the demand side volatility that used to wreck memory makers' earnings every few years. The pricing structure embedded in these deals protects the downside while preserving upside exposure. Micron's contracts include price floors and ceilings, with the floor set high enough that management says it guarantees gross margins well above our peak quarterly margins in any past cycle, while the ceiling caps pricing only at recent elevated market levels rather than historical lows. That means even in a hypothetical downturn, Micron's largest customers are contractually locked into paying prices near the top of the historical range, a mechanism that structurally raises the floor under the entire industry's profitability. This is exactly why we’ve stayed so bullish on the memory trade at Milk Road, and subscribers are already up massively on some of the positions we’ve made around this supercycle. If you want to see exactly what we’re buying and trading across Micron and the rest of the memory supply chain, come join us for just $1 using the link. link.milkroad.com/d8cj6v

If you want to become god-level with AI engineering, then learn these 21 concepts:

Andrew Ng just released 2-hour course on 100% Graph engineering: 1 prompt → 100 agents → loops → graphs from scratch: 10% → 9:14 - build your first agent from scratch 30% → 33:11 - Loop engineering 55% → 1:02:46 - Graph engineering 75% → 1:30:15 - agents that rewrite themselves 100% → 1:49:05 - full graph system that work without you most graph tutorials stop at the diagram - this one has you running one in the first 20 minutes watch this brilliant course, build the graph - then read the full architecture below ↓

codila @0xCodila

2 months ago

x.com/i/article/2079…

46 333 2K 5.3M 6K

Instead of watching 1 hour of Netflix tonight, watch this ex-Google Chief Scientist Jeff Dean’s lecture. It’s the clearest explanation I’ve seen of the full AI engineering stack - from building LLMs from scratch all the way to one human coordinating 100 agents. The best part is that it’s useful whether you’ve never touched a model or you’ve been shipping agent systems every day for the past year. Bookmark it & watch the whole lecture this weekend, because it might end up being the most valuable thing you learn all week.

You know what's the craziest part right now? Capex keeps getting raised, on the road to $1 trillion next year. Amazon's CEO says demand still outpaces supply, with no balance in sight through 2028. And after all that, $MU & $SKHY still trade at 4-5x forward earnings.

This is the REAL reason the AI infra trade topped (Save This) It wasn't China, Korea or Michael Burry shorting On June 16 the market started pricing a Fed HIKE. Days later semis peaked and the whole capex trade rolled over When rates go higher, the market doesn't want long-term capital expenditures and negative cash flows. And it definitely doesn't like leverage and debt Long-term capex makes perfect sense when rates are low or falling. Distant cash flows are worth more today, and the debt you raise to build is cheap. But when the market starts pricing HIKES, that math runs in reverse. Future returns discount harder, and every dollar of financing costs more, right when these companies need to raise more than ever The market spent a year assuming easy money was coming back. The last six weeks repriced that completely: 24% odds of a hike TODAY, ~82% odds by September, two hikes now priced for 2026 And once you see it, the selloff stops looking so random Oracle, the most debt-dependent builder ($40B raise, negative $24B FCF), had its worst month since 1990. Galaxy had to price its datacenter notes at 9.875%. NVIDIA's credit default swaps widened on the OpenAI backstop news. Big Tech bond demand hit a record-low 1.7x cover. Google and Meta took buybacks to ZERO to fund capex Meanwhile the one megacap with no capex bill, Apple, sits at all-time highs with its best month vs the Nasdaq in 20 years The market isn't repricing AI demand. It's repricing the cost of capital for the builders Again, the order books didn't change: HBM is sold out through 2027, hyperscaler capex is $600B+ and likely gets RAISED again this week. What changed is what it costs to finance it all or at least, the market THINKS that is going to change. I don't think the FED will raise rates today nor do I think they will do it in 2026 at all, so I think the market is getting this wrong and that's why I think the dips are meant for buying. Follow me @kylereidhead for more insights on AI and markets and track my real-time portfolio inside Milk Road PRO to see what companies I'm buying (link in bio)

SK Hynix just said second half of year will be better for ASP/earnings than first half of year on higher value product mix changes: "As HBM for shipments ramp up in earnest and the one nanometer conventional DRAM shipment increases, we expect big growth in the second half to be higher than the level in the first half." "Considering customer demand and the changing product mix, the growing HBM for sales and the higher contribution from value add products will have a positive impact on our blended ASP as well." "This will result in higher shipment volumes and continued improvement in the product mix, which will then push up our ASP as well as earnings in the second half."

[Exclusive] SK hynix to Begin Mass Production of LPDDR6 in Second Half; Xiaomi to Be First Customer SK hynix plans to begin mass-producing and shipping LPDDR6, its next-generation low-power mobile DRAM, in the second half of this year. Having completed development of the product in March, the company intends to make its first shipment to China’s Xiaomi for use in mobile devices. The launch is expected to further intensify competition among the three major memory manufacturers—Samsung Electronics, SK hynix, and Micron—in SOCAMM, a next-generation, LPDDR-based low-power memory module for AI servers. According to industry sources on July 28, SK hynix will enter full-scale mass production of LPDDR6 in the second half of this year. Xiaomi is expected to be its first customer, with the memory slated for use in the company’s next-generation flagship smartphones. SK hynix’s LPDDR6 uses its sixth-generation 10nm-class (1c) process, offering improved data-transfer speeds and power efficiency over the previous generation. In March, SK hynix announced that it had completed certification of what it described as the world’s first LPDDR6 product, saying, “We plan to complete preparations for mass production within the first half and begin supplying the product in the second half.” Regarding the reported Xiaomi deal, however, the company said it could not confirm information related to customers. SK hynix began doing business with Xiaomi in 2013, supplying LPDDR3 for certain models, and is understood to have continued supplying mobile DRAM through the latest LPDDR5X generation. Samsung Electronics and SK hynix currently lead the mobile DRAM market. Meanwhile, CXMT, China’s largest memory manufacturer, has been rapidly gaining ground following its stock-market debut, bringing its market share close to that of third-ranked U.S. chipmaker Micron. Samsung supplies LPDDR to Android smartphone manufacturers worldwide, while SK hynix has a strong position among Chinese smartphone brands. The Xiaomi deal is expected to pave the way for SK hynix to expand LPDDR6 shipments to other Chinese smartphone manufacturers, including Vivo and Oppo. Chinese smartphone makers are increasingly adopting domestically produced memory from suppliers such as CXMT in their entry-level lineups, but their flagship models—which require higher performance—remain dependent on LPDDR from the three major global memory manufacturers. LPDDR is primarily used in mobile devices such as smartphones and tablet PCs. By reducing power consumption while maintaining high performance, it has recently experienced explosive demand from AI servers. Some industry observers expect it to become a major growth driver for the memory market after HBM, or high-bandwidth memory. LPDDR supply prices have recently risen to levels comparable with those of DRAM, whose prices have been soaring. The increase stems from supply bottlenecks caused by memory manufacturers concentrating their production capacity on commodity DRAM and HBM. As LPDDR applications diversify and contract prices continue to rise, memory manufacturers posted record-high revenue in the first quarter, supported by strong LPDDR sales. Market-research firm TrendForce said, “Mobile DRAM revenue increased by 64.5% in the first quarter of this year,” adding that “continued increases in contract prices will drive further growth in mobile DRAM revenue in the second quarter.” Samsung Electronics and SK hynix are also seen as having seized the initiative by beginning preparations for mass production of the latest LPDDR6 products ahead of Micron. In January, Samsung unveiled what it described as the industry’s first LPDDR6 product built using fifth-generation 10nm-class (1b) DRAM technology. The product won a CES Innovation Award. SK hynix’s mass production of LPDDR6 is also expected to signal the start of a full-fledged race to secure an early lead in the SOCAMM market. SOCAMM packages LPDDR into a module designed for AI servers. As the AI market shifts from training toward inference, it is rapidly emerging as a next-generation memory solution for AI servers. In June, Nvidia CEO Jensen Huang visited SK hynix’s booth at Computex 2026, Asia’s largest information-technology trade show, and left the message “LOVE SOCAMM” on a 192GB SOCAMM module. Nvidia also recently unveiled Vera, its first internally designed CPU for AI servers. If HBM is the GPU’s natural partner, LPDDR is the CPU’s. As the AI market becomes increasingly advanced and demand for CPUs rises sharply, demand for SOCAMM is also expected to strengthen.

SK Hynix CEO Kwak Noh-jung: “We forecast that next year will be the worst year in the industry's history from the supply perspective. We still forecast that customer demand will remain higher than our supply capacity even beyond 2030.” $SKHY $MU $AMD $NVDA

$SKHY Chairman Chey Tae-won expects AI semiconductor demand to surge 60–100% next year with overall memory demand rising nearly 60%. He warns supply will barely increase leaving global companies “clamoring for memory chips” as the shortage worsens.

keyboard_arrow_left Previous keyboard_arrow_right Next

SOFTBANK is now live. 10x leverage, 24/7, 365.