BlackRock just published a paper saying AI is the most underappreciated demand driver for crypto (holy shit) This quote was killer: "AI represents machine-native intelligence, digital assets represent machine-native money" These two themes have been treated as separate trades With the famous tweet: "If you're in crypto, pivot to AI" But I've long said that a bet on crypto is a bet on AI and that's starting to play out Blackrock notes 2 specific areas where these technologies converge: 1. Agentic Commerce requires machine-native payment rails As AI agents start actually DOING things (buying data, booking travel, renting compute), they need a way to pay. And that's where it gets interesting An AI agent can't easily walk into a bank and open an account. Card rails need human onboarding, charge merchant fees that make a sub-penny API call pointless, and take days to fully settle and clear disputes A stablecoin wallet runs 24/7, settles in seconds, and doesn't care if the owner is a person or a piece of software 2. Compute is emerging as a new and potentially large market for digital assets As compute becomes one of the largest in-demand products in the world, BlackRock thinks standardized, tokenized claims on compute become a real digital asset market, with agents shopping for GPUs and paying per job. There are already early signs of Compute markets and digital assets converging with @USDai_Official, who uses stablecoins to provide financing for GPUs and uses the interest to give yield to USDAI stablecoin holders As the compute market expands, this market should have significant growth across the digital assets ecosystem Crypto has long been judged/valued based on human use cases, but I think it needs to be looked at as financial technology for AI, rather than humans (though it will work for both) Companies like Coinbase and Circle are already going all in on becoming the payment structure for AI and so to are many DeFi protocols. If you want to see what assets I hold in my portfolio to capture the agentic financial upside, you can check out my portfolio here: Don't forget to give me a follow @kylereidhead for more insights on AI, crypto and markets

Meta just found the product that turns its massive AI spending into the next great consumer platform. Muse’s daily US downloads recently surpassed 300,000, putting it above Instagram, WhatsApp, Facebook, Threads and Messenger individually and it also reached No. 1 on the US App Store. The reported Meta Pay integration could make Muse considerably more valuable by allowing users to connect cards already stored in Meta Pay and authorize purchases and Meta is also assembling a powerful commerce network around Muse. Stripe’s Link already lets Muse check out at more than one million businesses and can create single use virtual cards for approved purchases at other merchants. Shopify is bringing Shop Pay checkout to Muse across Shopify stores, connecting Meta’s audience with Shopify’s merchants, product catalogs and payment infrastructure. PayPal is also partnering with Meta to enable shopping and checkout across PayPal’s global merchant network. These partnerships allow Meta to connect every stage of online shopping. Instagram and Facebook provide product discovery, Muse handles research and recommendations, Shopify and PayPal provide merchant access, and Stripe or Meta Pay completes the transaction. This could give Meta visibility across the entire customer journey from advertisement to purchase, making its advertising platform even more valuable. Meta could also monetize Muse through subscriptions, enterprise services, API usage and eventually transaction or referral revenue,. The opportunity is becoming more important as Meta expects to spend between $130 billion and $145 billion on capital expenditures during 2026 and Muse provides early evidence that this investment could create a major consumer business instead of only improving Meta’s existing advertisements. Meta already controls social discovery through Instagram and Facebook, communication through WhatsApp and Messenger, and AI assistance through Muse and if Meta Pay becomes the final payment layer, Meta could control the entire path from product discovery to checkout.

Another SaaS sell off is just inevitable at this point because of the rise of AI agents. Goldman Sachs expects the software profit pool to shift heavily toward AI agents, with the agent market growing to more than $50 billion by 2030 while traditional SaaS falls from roughly $30 billion to around $20 billion. This does not mean software is disappearing but the overall market could actually become larger, but more of the money is expected to move away from traditional SaaS products and toward the agents operating them. Today, businesses pay monthly fees for employees to use Salesforce, ServiceNow, Workday and dozens of other applications but agents can move across those applications and complete the work themselves. Instead of opening several programs, the user simply explains the desired outcome. The agent becomes the main interface, while the software underneath turns into a tool working quietly in the background. This puts pressure on the traditional per seat model because companies may not need as many employees actively using each application. It also weakens customer lock in because users become more loyal to the agent than to the software it operates. Yes, I know many large software companies now offer usage based pricing alongside their legacy seat based models. That should help them adapt as customers move toward paying for completed work rather than software access but the market rarely makes that distinction during a major sell off. If investors begin dumping SaaS stocks because they fear AI agents will weaken seat growth and pricing power, even the strongest software giants will probably get grouped into the same trade and sold alongside the weaker companies. We already saw broad software weakness when new agent products raised similar concerns earlier this year. Muse will be the next catalyst because it is bringing personal agents to a much larger audience. Once regular users become comfortable allowing one agent to work across multiple applications, investors will begin questioning how many separate subscriptions and paid seats businesses truly need. OpenAI and Anthropic will likely respond soon their own version of Muse very soon. And once Muse, OpenAI and Anthropic begin competing to control the user’s entire workflow, SaaS companies will no longer compete only against other software products but also compete against the AI agents deciding which software gets used in the first place. Not every SaaS company will lose because those with proprietary data, deep integrations and strong usage based models should be more defensible but if another broad SaaS sell off begins, the market may not care about those differences at first. The weaker companies will fall and the giants will likely get dragged down with them. If you enjoyed reading this, make sure to follow me @MelvinInvests for more and I’ve already started positioning my portfolio around what I think is coming next as AI agents begin reshaping the software stack. If you want to see exactly what I’m holding, what I’m avoiding, and how I’m positioned for this shift, you can check out my full Milk Road portfolio using the link below.

We are heading into a DEFLATIONARY, PRODUCTIVITY BOOM (yes, you read that right) The cost of intelligence is falling 99%+ and the only thing propping up inflation is oil, which will head down to the $30s Sounds crazy? It's not (Investors, read this!) AI inference costs have dropped more than 99% while demand for tokens rose 25x. That's the cost of digital labour falling 99%+ and when robotics scales, the cost of PHYSICAL labour starts down the same curve Blockchain is doing the same thing to the cost of financial transactions Genome sequencing is doing the same thing for healthcare, as it went from $2.7 BILLION per genome in 2003 to under $100 today This is all deflation from technology collapsing costs while demand EXPLODES, this is not deflation from a demand collapse (this is a good thing) So why is inflation higher if all of this is happening? OIL Crude spiked 60% to $100+ on the war, and energy bleeds into the price of everything. But that's a war premium from a supply shock, not a demand story and it won't last for long and OPEC members are already producing over quota, so when the war ends an oversupply hits a market that doesn't need it. ARK sees $30-35 oil in the coming years and I fully agree Strip out the war premium and nearly every structural force in the economy is pushing prices DOWN while output goes UP. We're heading into a productivity boom combined with deflationary pressures This couldn't be a better time for the stock market and I believe this plays our for years ahead, not just a few months That's the world I'm positioned for inside my Milk Road PRO portfolio. If you want to see the exact assets I hold right now and see moves in real-time you can check it out here: Don't forget to give me a follow @kylereidhead for more insights on AI and markets

Another SaaS sell off is coming soon and I think the rise of AI agents will be the reason why. Agents like Muse and Grok are bringing AI agents to the consumer market in masses. What was recently experimental technology is becoming something regular people can use to complete real work across their applications. Grok Bot, for example, can operate software through its own cloud computer and continue working after the user steps away. This is a much bigger threat to software companies than a chatbot that only answers questions. Instead of opening Salesforce, ServiceNow and Workday separately, an employee can tell an agent what needs to be done. The agent can then move between those applications and complete the entire workflow. This means the agent becomes the main interface, while the software underneath becomes a tool working quietly in the background. That puts pressure on the traditional SaaS business model. If agents perform more of the work, companies may not need as many paid software seats across CRM, project management and customer support. Agents could also make switching between software providers easier. Once users interact with Muse or Grok instead of the actual application, they become less attached to that software. The agent could choose whichever service offers the best price or performance. That weakens customer lock in. The agent controls the user experience and decides which software gets used, while SaaS companies risk losing pricing power and their direct relationship with customers. We have already seen how quickly investors react to this threat. New agent capabilities from Anthropic contributed to weakness in software stocks earlier this year as the market questioned whether AI would help SaaS companies or replace parts of them. Muse and Grok could cause another sell off because they are making agents available to regular people. I also think this will force OpenAI and Anthropic to respond very soon. OpenAI has already shown where it is heading. Earlier this year, it hired Peter Steinberger, the creator of OpenClaw, to help build its next generation of personal agents. I am sure they are going to release something soon and then Claude will respond with their own version. If Meta, Grok, OpenAI and Anthropic all begin releasing more capable agents, SaaS companies will no longer compete only against other software products. They will also compete against the agents controlling how customers use those products. The most exposed companies will be those selling basic productivity tools, simple workflows and large numbers of employee seats. Businesses with proprietary data, deep integrations and strict compliance requirements should be safer but they too will be affected by this sell off. If you enjoyed reading this, make sure to follow @MelvinInvests for more.

Both OpenAI and Anthropic say the AI race needs to slow down yet both companies released newer models on the exact same day. Anthropic released Claude Opus 5.5, while OpenAI released GPT-6 Sol and GPT-6 Luna on September 22. Anthropic says Opus 5.5 delivers performance close to its more powerful Claude Fable 5.1 model while costing about 40% less to run than Opus 5. The model is designed for coding, computer use and long professional tasks that require the AI to work independently. OpenAI expanded the GPT-6 family with two cheaper models. GPT-6 Sol is built for difficult coding, research and computer-use tasks, while GPT-6 Luna is designed for faster and cheaper work at scale. OpenAI priced both models 50% below their GPT-5.6 equivalents, making GPT-6 more practical for businesses running millions of AI tasks.9to5mac The same day releases show how intense the competition has become. Both companies may genuinely want more time for safety testing, but neither company wants to give the other one a major advantage. If Anthropic slows down while OpenAI continues releasing models, OpenAI can win more developers and enterprise customers. If OpenAI slows down while Anthropic keeps moving, Anthropic can do the same. This creates a race in which both companies want everyone to slow down but neither company wants to slow down first.

Micron looking REAL GOOD here $MU dropped 50% after the AI infra trade topped back in June on: - open-source model fears - China supply fears - deleveraging of KOSPI and Situational Awareness - reignited Iran war and Oil price Through all of that the fundamentals of the demand for memory and for Micron specifically improved significantly Multiple banks and analysts have now agreed the memory cycle will go longer than previous cycles, with supply constrained, prices rising and margins holding past 2029 All of this is happening while we get closer and closer to December, when Micron will be able to turn on stock buybacks, something they haven't been able to do for years Microns FCF is heading into the $100s of Billions next year and the bulk of it will be used to buyback their stock The market is very likely to front run this and its why $MU is beginning to break out from its multi-month consolidation pattern I think we will see ATHs in Micron very soon as the market begins to turn the AI infra trade back on and risk enters the markets again throughout Q4 $MU is part of my portfolio and If you want to see all of the assets I hold, which span across AI infra, AI application layer and even some crypto equities, you can check it out here: Good luck out there!

Nebius just turned its spare GPU capacity into a live auction. Starting October 8, the price of its preemptible virtual machines will change based on real time demand, available capacity, GPU type and region. Previously, Nebius set a fixed discount for these machines but under the new system, prices can fall when Nebius has extra GPUs available and rise when demand becomes stronger. Customers can either follow the changing market price or set the maximum price they are willing to pay. If the spot price rises above a customer’s limit, Nebius can stop that virtual machine and give the capacity to someone willing to pay more. Nebius is making this change because dynamic pricing allows Nebius to lower prices when demand is weak, attract extra workloads and keep more of its expensive infrastructure running. When demand becomes stronger like right now, Nebius can raise the spot price instead of continuing to sell scarce capacity at a fixed price. This could improve both utilization and revenue because Nebius can fill otherwise idle GPUs while charging more whenever customers begin competing for limited capacity. Bullish on Nebius and if GPU demand stays this strong, dynamic pricing gives $NBIS another lever to squeeze more revenue and higher margins out of every GPU it already owns.

Elon Musk warned that electricity will become the next major AI shortage. He expects today’s silicon shortage to shift toward transformers and eventually electricity as AI, transportation and heating electrify. Here are five under the radar stocks positioned to benefit (Save this).

You don't understand markets if you think a 2nd wave of INFLATION like the 1970s is coming Here's what everyone sharing this chart misses: The 70s second wave was driven by demographics (Baby Boomers) Today's demographics are the exact OPPOSITE (AI Agents) In the 1970s the baby boomers, the largest generational cohort ever, all entered the workforce, bought homes and started families AT THE SAME TIME Demand for everything exploded at once. Homes, cars, dishwashers, food. And supply couldn't respond, because in the 70s building new manufacturing capacity took YEARS Too much demand chasing too little supply for a decade straight. That's why inflation kept coming back in waves and rates had to go to the moon to kill it Today is the complete opposite... The boomers aren't entering the workforce, they're LEAVING it. Labor force participation is declining as the largest cohort in history (finally) retires And the demographic entering the workforce at size? AI agents. Millions of digital workers that don't buy homes, don't buy cars, don't buy dishwashers and don't eat. They add supply (labour, output, productivity) while adding basically zero demand for goods The 70s formula was demand shock + slow supply. Today's formula is demand fade + instant supply. You couldn't design a more opposite setup for inflation This is why I've been saying the same thing since 2023 while people called for the second wave every single year: inflation is NOT sticky, and the second wave isn't coming Sure, the price of Oil is holding inflation higher than it should be, but it's not creating another double digit inflation wave. Not even close. Even if we get another rate hike, it won't matter, as the growth from AI and agents far outweighs the restrictions of 25bps This is a big part of why I think we are entering the EVERYTHING bull market and why you should be allocated to the market and not sidelined, regardless of what the doomers will tell you. By the way, if you're unsure how to invest in todays market, you can see my entire real-time portfolio here: Don't forget to give me a follow @kylereidhead for more insights on AI and markets

Wall Street’s earnings forecasts are exploding higher and the market is nowhere near finished (Save this). Analysts usually begin with optimistic forecasts and gradually lower them as reality catches up but this time, the exact opposite is happening. Goldman Sachs chart shows that global earnings estimates for 2026 have risen from roughly $52 per share to more than $61 per share. The 2027 estimate has increased even faster, rising from approximately $56 per share to more than $71 per share.. These increases are unusual because earnings estimates for most previous years moved sideways or lower as the year approached and the increases for 2026 and 2027 suggest that companies are earning considerably more money than analysts expected. So why this all happening? it's because of AI is one of the biggest reasons for this earnings growth. Goldman Sachs estimates that global AI investment will exceed $1 trillion in 2026, with approximately $581 billion of that spending occurring in the United States. This money is flowing into semiconductors, memory, networking equipment, cloud infrastructure, data centers, cooling systems and power equipment. The spending cycle also appears to have more room to run because Goldman Sachs expects global AI investment to increase from 0.9% of global GDP in 2026 to 1.3% in 2027 and 1.4% in 2028. The bank also estimates that hyperscaler capital spending could reach approximately $1.1 trillion in 2027, compared with Wall Street’s forecast of roughly $920 billion. This continued investment helps explain why the 2027 earnings line is rising so quickly. The growth is also beginning to spread beyond the largest American technology companies. Companies that provide chips, memory, electricity, construction, networking and cooling equipment are earning more money as the AI infrastructure buildout expands. This is important because a bull market becomes stronger when earnings growth spreads across more sectors, countries and companies. This is why I’m still comfortable staying heavily exposed to AI and infrastructure names. Earnings expectations are not rolling over, they’re still moving higher. If you want to see the positions I’m holding and the trades I’m making around this trend, check out my Milk Road Pro portfolio below.

Our analyst @MelvinInvests has been screaming that $META was a buy since June. Meta is now up over 11% today, and our Pro members were able to track every move he made along the way, from the first buy to every DCA. If you want to see exactly what Melvin is buying in real time, you can track his full portfolio below.

Grok 4.7 just pushed SpaceXAI into the top four AI labs in the world. The new model scored 46 on the independent Artificial Analysis Intelligence Index, which is two points higher than Grok 4.6. That still places Grok behind the most powerful models from OpenAI and Anthropic, which scored as high as 53 but the gap is becoming much smaller. The biggest improvement appeared in tasks that resemble real work. Grok 4.7 scored 1,657 Elo on AA-Briefcase, which measures how well AI can complete long and complicated professional assignments. That score increased by 111 points from Grok 4.6 and placed the model alongside the newest Claude models at the frontier of agentic knowledge work. Grok also made a major jump in coding.Grok 4.7 paired with Grok Build scored 56 on the Artificial Analysis Coding Agent Index, compared with 47 for Grok 4.6. That result moved Grok Build into fourth place among coding agents and pushed it ahead of GPT-5.6 Sol. The model also scored 71% on DeepSWE, which placed it close to GPT-5.6 Sol and slightly ahead of Claude Fable 5.1 on that particular software-engineering benchmark. Grok 4.7 is showing that SpaceXAI can now compete in those areas. The company is also offering the model for $2 per million input tokens and $6 per million output tokens, which is far below the listed prices of several competing frontier models. Grok 4.7 also supports a 500,000 token context window, which allows it to process large collections of documents and extended conversations in a single session. However, the model uses a considerable amount of computing power to achieve these results. Artificial Analysis found that Grok 4.7 used roughly 81,000 output tokens per Intelligence Index task, compared with 36,000 for Grok 4.6 and 27,000 for GPT-6 Astra. This means that Grok is becoming much more capable, but it is sometimes reaching better answers by thinking longer and using more tokens.That tradeoff will matter for companies running thousands or millions of tasks because a low price per token does not always produce a low total cost per task. Grok 4.7 is not the best model on every benchmark, and it has not taken the overall crown from OpenAI or Anthropic. However, SpaceXAI no longer looks like a company merely trying to catch up. Grok is now competing near the frontier in intelligence, coding and professional knowledge work while charging considerably less than several of its largest rivals. @elonmusk has finally turned Grok into a legitimate frontier model and the gap between SpaceXAI and the leaders is closing much faster than most people expected.

Meta’s Muse agents are about to ignite the next CPU boom (Save this). The more tasks these agents complete, the more CPU power Meta will need to run them. Here are the five stocks positioned to win from this.

Big Tech is sitting on $3 TRILLION of AI commitments that don't show up in the capex numbers Bears call it hidden leverage... I'd say its hidden upside and it means the AI infra trade is bigger than most think This chart is showing off-balance sheet purchase obligations, leases and guarantees. Google alone has $707B of future purchase commitments. Microsoft carries $329B in leases, Oracle $261B, Meta $279B None of this sits in the reported capex numbers yet. It's all in ADDITION to the ~$1 trillion/year already flowing The Micheal Burris of the world say this is "hidden liabilities! leverage! this is how bubbles end!".. The way I see it though is a purchase obligation is a signed ORDER. Google didn't accidentally commit $707B, they contractually locked in future compute, chips and datacenter capacity for YEARS ahead. You only do that when you're terrified of not getting supply And if you think they won't be able to pay for it, then you're effectively saying the biggest and most successful companies of our lifetime are going to fail in the next 1-3 years... there's absolutely no way that will happen So what you should be asking is who RECEIVES that $3 trillion? Chips, memory, power equipment, data center builders, neoclouds. What the bears call Big Tech's hidden liability is the supply chain's contracted revenue backlog. Three trillion dollars of it, on top of the trillions in reported capex The capex numbers you see quoted all over X are the FLOOR, not the ceiling. The AI infra trade is bigger than the market is pricing I'm positioned all through the stack (compute, memory, power, neoclouds) and if you want to see my entire real-time portfolio you can check it out here: Don't forget to give me a follow @kylereidhead for more insights on AI and markets

Wall Street’s earnings forecasts are exploding higher and the market is nowhere near finished (Save this). Analysts usually begin with optimistic forecasts and gradually lower them as reality catches up but this time, the exact opposite is happening. Goldman Sachs chart shows that global earnings estimates for 2026 have risen from roughly $52 per share to more than $61 per share. The 2027 estimate has increased even faster, rising from approximately $56 per share to more than $71 per share.. These increases are unusual because earnings estimates for most previous years moved sideways or lower as the year approached and the increases for 2026 and 2027 suggest that companies are earning considerably more money than analysts expected. So why this all happening? it's because of AI is one of the biggest reasons for this earnings growth. Goldman Sachs estimates that global AI investment will exceed $1 trillion in 2026, with approximately $581 billion of that spending occurring in the United States. This money is flowing into semiconductors, memory, networking equipment, cloud infrastructure, data centers, cooling systems and power equipment. The spending cycle also appears to have more room to run because Goldman Sachs expects global AI investment to increase from 0.9% of global GDP in 2026 to 1.3% in 2027 and 1.4% in 2028. The bank also estimates that hyperscaler capital spending could reach approximately $1.1 trillion in 2027, compared with Wall Street’s forecast of roughly $920 billion. This continued investment helps explain why the 2027 earnings line is rising so quickly. The growth is also beginning to spread beyond the largest American technology companies. Companies that provide chips, memory, electricity, construction, networking and cooling equipment are earning more money as the AI infrastructure buildout expands. This is important because a bull market becomes stronger when earnings growth spreads across more sectors, countries and companies. This is why I’m still comfortable staying heavily exposed to AI and infrastructure names. Earnings expectations are not rolling over, they’re still moving higher. If you want to see the positions I’m holding and the trades I’m making around this trend, check out my Milk Road Pro portfolio below.

Everyone was clowning Mark Zuckerberg and Alexandr Wang but now they are back on top. Meta spent more than $14 billion to bring @alexandr_wang and several Scale AI researchers into the company after the disappointing release of Llama 4. Many people questioned whether Wang had enough experience to lead one of the world’s largest AI teams and nvestors also worried that @finkd was spending too much money without having a clear plan to earn it back. But now that narrative is completely dead after Meta’s new Muse AI agent reached No. 1 on Apple’s US App Store shortly after its launch and moved ahead of ChatGPT. Meta offers Muse for free but users who need more access can pay $20 or $100 per month which means that Meta finally has a consumer AI product that could turn its massive infrastructure spending into direct revenue. Oppenheimer estimates that Muse could eventually reach 1.91 billion users across Meta’s applications and If only 6% of those people become paying subscribers, Meta would have approximately 115 million paying Muse customers. Those customers could generate approximately $27.5 billion in annual revenue at a monthly price of $20 and Oppenheimer estimates that this revenue could produce approximately $22 billion in operating income. This scenario would increase Meta’s total 2027 revenue to approximately $332 billion and its operating income to more than $122 billion. Meta’s earnings could reach $38.83 per share, which would be 15% above the Wall Street estimate used in Oppenheimer’s model. Meta has an advantage that most other AI companies do not have. Meta already owns Facebook, Instagram, WhatsApp and Messenger, which gives the company a direct path to billions of potential Muse users. Meta does not need to build a new audience because it can place Muse inside the applications people already use every day. The company could eventually connect Muse to its AI glasses, which would allow users to access the agent without opening a phone or computer. Meta could also make money from consumer subscriptions, business agents, developer access and commissions from purchases completed through Muse. Meta is an asset I hold in my portfolio, and if you want to see the rest of my positions, how I’m sizing them, and the trades I’m making across AI, check out my Milk Road Pro portfolio using the link below.

Samsung and SK hynix are about to make an absurd amount of money and South Korea’s export data just revealed why. South Korea’s semiconductor exports surged 259.4% year over year to a record $34.1 billion during the first 20 days of September. Semiconductors now account for 47.8% of the country’s total exports, which is the highest percentage ever recorded for this period. These numbers show that the AI boom is no longer based only on future spending forecasts because massive chip orders are already being manufactured, shipped and converted into revenue. Samsung and SK hynix are benefiting from rising shipment volumes and dramatically higher memory prices at the same time. The supply constraint allows Samsung and SK hynix to raise prices across a much larger portion of their memory businesses. Samsung reported that its average selling prices increased by approximately the mid-40% range for DRAM and the high 60% range for NAND during the second quarter. These price increases can produce enormous profit growth because semiconductor manufacturing carries extremely high fixed costs. Once the factories are operating near capacity, every additional dollar generated through higher selling prices can carry a very high incremental margin. Samsung demonstrated this operating leverage during the second quarter when its revenue reached KRW 171.5 trillion and its operating profit climbed to KRW 89.5 trillion. Samsung’s semiconductor division generated KRW 127.5 trillion in revenue and KRW 89.2 trillion in operating profit, which means the semiconductor business produced almost all of the company’s profit. SK hynix offers more concentrated exposure to the AI-memory boom because it remains one of the leading suppliers of HBM for AI accelerators. The company began shipping HBM4 during the second quarter and plans to expand production throughout the second half of the year. SK hynix has also completed long term supply negotiations with more than 10 customers, which gives the company greater revenue visibility and reduces the risk of expanding production without committed buyers. Samsung offers a broader opportunity because it can benefit from rising demand for HBM, conventional DRAM, NAND and enterprise SSDs. Samsung also expects its HBM4 sales to more than triple sequentially during the third quarter, while HBM4 is expected to represent more than 60% of its total HBM revenue during the second half. Samsung and SK hynix do not need semiconductor exports to continue growing by 259% for their profits to keep rising. They only need shipment volumes to remain strong while memory prices stay elevated. If those conditions continue, revenue should keep expanding while operating profit grows even faster because higher selling prices can flow almost directly to the bottom line. Make sure to follow @MilkRoadAI for more insights.