SpaceX building its own turbine blades tells you where Alpha is in the AI infra trade (Save this) Musk says 15 GW of AI compute produced in 2027 may not turn on that year, and the fastest fix is gas turbines. Only 3-4 companies in the world cast the blades those turbines need, and they are sold out through 2030, so SpaceX is casting its own and says it gets turbines running 18 months sooner. What we are seeing here is AI companies integrating energy infra cause time to power has become a competitive advantage. When billions of dollars of GPUs sit waiting for electricity, you face massive idle costs as an operator. As ~65% of datacenter spend is on chips, operators don’t care to pay higher prices for faster available energy. Keep coming back to $BE as my expression of this trade. Bloom puts power on site in months, and all hyperscalers and major neoclouds are customers by now. Companies like GE or Eaton could also be interesting as they are feeding the grid related parts sold out for years, which is the actual bottleneck we need to solve for. I've been covering the energy story inside Milk Road PRO since end of last year with calls like $BE. You can see my portfolio for $1 here:

AI agents and stablecoins are converging into 1 trade, and an $8B deal last week proved it. Every company needs 2 things: A way to get paid and labor. Labor used to mean hiring people. Now you assemble a team of intelligence instead and you buy it by the token. Stripe already handles the getting paid infra layer, routing $1.9T of annual volume for small companies, and it is moving that platform onto stablecoins. Last week it paid ~6x what OpenRouter was worth a few months ago to handle the intelligence side too. The deeper thinking view is that this is a bet on Entrepreneurship. Small companies which Stripe serves will adopt and benefit from AI much faster! Covering the names that build for this convergence of agents and digital assets inside Milk Road PRO, and there are only a couple of hours left before the price goes up:

Every single quarter, the AI trade holds its breath for $NVDA to "rescue the AI party." And every single quarter, the setup feels the same. For today's earnings: - Q2 revenue should land $94-95B - Q3 guidance should be $107-109B - And the real surprise would be non-GPU revenue (CPUs, networking, storage) coming in above $17B. @WhiteCollarExit believes that the stock will have a flat/down reaction when earnings release.

$NVDA incremental upside is coming from a $200B market it has never competed in. (Save this) Jensen's $1T Blackwell and Rubin opportunity is now the base case rather than the bull case. Vera, Nvidia's new CPU, sits outside of that number and management has visibility to ~$20B of CPU revenue this year. Adoption is real too: OpenAI, Anthropic, CoreWeave and Oracle all plan to use Vera, and SpaceXAI announced deployments for Grok. LPX takes this even further, pushing Nvidia into ultra fast inference and the data layer that keeps AI agents running. So, the deeper thinking is: Nvidia can lose some GPU share and still grow its share of the total AI spend through CPUs, networking and storage. That makes the business less dependent on defending one product and strengthens the long term earnings story. So beyond the Q3 guide, I want proof that products beyond the GPU (Vera, LPX, storage, ...) are moving toward incremental revenue. Will be covering the broader implications of NVDA earnings inside Milk Road PRO. Only a couple of hours left before the price goes up:

$NVDA earnings preview: EPS grew 200% in 2 years and the stock got 50% cheaper. (Save this) The stock is up 67% over that stretch, and the forward multiple fell from 47x to 22x. So, the whole question is what the market is still scared of? The first fear is custom silicon taking accelerator share, but NVLink Fusion lets chips like Amazon's Trainium plug into Nvidia's infrastructure, so a smaller GPU share does not mean a smaller Nvidia bill. The second is competition crushing that 75% gross margin. The chip architecture roadmap incl. Rubin reduces cost per token produced, which lets Nvidia hold premium pricing while customers still get cheaper AI. The third is circular financing. Nvidia has been funding the customers buying its own chips. The $500B facility it lined up was a great move to disconnect from hyperscaler (negative) FCF risk. The real proof is AI revenue paying for the next CapEx wave by itself. My view is that Wednesday comes down to the Q3 guide (assuming GM will remain at >=75%). Street sits near $104B and I want to see >=$107B. I also want signs that Vera, networking and storage are moving from optionality toward meaningful revenue contribution. Those are the numbers I'll be checking Wednesday night, and I'll post what they mean for my positions inside Milk Road PRO. Join before Aug 26, then the price goes up:

$NVDA earnings preview: EPS grew 200% in 2 years and the stock got 50% cheaper. (Save this) The stock is up 67% over that stretch, and the forward multiple fell from 47x to 22x. So, the whole question is what the market is still scared of? The first fear is custom silicon taking accelerator share, but NVLink Fusion lets chips like Amazon's Trainium plug into Nvidia's infrastructure, so a smaller GPU share does not mean a smaller Nvidia bill. The second is competition crushing that 75% gross margin. The chip architecture roadmap incl. Rubin reduces cost per token produced, which lets Nvidia hold premium pricing while customers still get cheaper AI. The third is circular financing. Nvidia has been funding the customers buying its own chips. The $500B facility it lined up was a great move to disconnect from hyperscaler (negative) FCF risk. The real proof is AI revenue paying for the next CapEx wave by itself. My view is that Wednesday comes down to the Q3 guide (assuming GM will remain at >=75%). Street sits near $104B and I want to see >=$107B. I also want signs that Vera, networking and storage are moving from optionality toward meaningful revenue contribution. Those are the numbers I'll be checking Wednesday night, and I'll post what they mean for my positions inside Milk Road PRO. Join before Aug 26, then the price goes up:

The investor takeaway from $MRNA AI cancer vaccine is what happens when computation becomes part of the medicine itself. (Save this) Reminder of how the Moderna tech works: An AI model picks up to 34 targets specific to the persons cancer DNA, and an individual therapy is manufactured around them (See video below). Results (according to Moonshots): - Cut's the risk of recurrence or death by 49% - $5000 per patient (!) So, AI itself is becoming part of the pharma value chain which is much deeper than R&D (=drug discovery) only. What this enables is the move from 1 drug for millions to 1 platform designing millions of individual treatments. That implies that the most valuable asset in pharma becomes the system that forces this loop. $LLY is building that system! Breaking down the entire Lilly thesis on Milk Road PRO. Join us before prices increase dramatically Aug 26:

The investor takeaway from $MRNA AI cancer vaccine is what happens when computation becomes part of the medicine itself. (Save this) Reminder of how the Moderna tech works: An AI model picks up to 34 targets specific to the persons cancer DNA, and an individual therapy is manufactured around them (See video below). Results (according to Moonshots): - Cut's the risk of recurrence or death by 49% - $5000 per patient (!) So, AI itself is becoming part of the pharma value chain which is much deeper than R&D (=drug discovery) only. What this enables is the move from 1 drug for millions to 1 platform designing millions of individual treatments. That implies that the most valuable asset in pharma becomes the system that forces this loop. $LLY is building that system! Breaking down the entire Lilly thesis on Milk Road PRO. Join us before prices increase dramatically Aug 26:

Everyone owning $MU is trying to answer one question: Do record memory prices mean the cycle is about to roll over? (Save this) The case for no starts with NVDA evaluating 33% less HBM for its next architecture. They would not redesign around a part that is abundant! Next, Memory spend currently at ~33% of total AI data center spend is expect to head to ~50% next year. Also, the Moonshots clip below explains why supply-side is paranoid about oversupply from past boom-bust cycles. Which is why they are likely to not overbuild and also sign LTAs with floor prices. However, there is one scenario (among others) where the memory trade could eventually break. Today, AI repeatedly pulls model weights from memory to generate answers. If those weights become stable enough, you could bake them directly into the chip and remove much of the memory needed. That works best for mature, repetitive workloads where the same model can run for a long time. The tradeoff is flexibility. Once the model is baked into silicon, you cannot simply load the next version onto the same chip (=stranded assets). My view is that 2026-28 is about using less HBM for each job, not eliminating memory. Beyond that, innovation like the described case above could become materials. At Milk Road PRO we have been banging the drum on memory since launching our portfolios in March. We bought the dip recently but also sold some of our AI infra positions. All with live notifications! Come join us before prices increase after Aug 26:

Everyone owning $MU is trying to answer one question: Do record memory prices mean the cycle is about to roll over? (Save this) The case for no starts with NVDA evaluating 33% less HBM for its next architecture. They would not redesign around a part that is abundant! Next, Memory spend currently at ~33% of total AI data center spend is expect to head to ~50% next year. Also, the Moonshots clip below explains why supply-side is paranoid about oversupply from past boom-bust cycles. Which is why they are likely to not overbuild and also sign LTAs with floor prices. However, there is one scenario (among others) where the memory trade could eventually break. Today, AI repeatedly pulls model weights from memory to generate answers. If those weights become stable enough, you could bake them directly into the chip and remove much of the memory needed. That works best for mature, repetitive workloads where the same model can run for a long time. The tradeoff is flexibility. Once the model is baked into silicon, you cannot simply load the next version onto the same chip (=stranded assets). My view is that 2026-28 is about using less HBM for each job, not eliminating memory. Beyond that, innovation like the described case above could become materials. At Milk Road PRO we have been banging the drum on memory since launching our portfolios in March. We bought the dip recently but also sold some of our AI infra positions. All with live notifications! Come join us before prices increase after Aug 26:

$COIN is up 30% this week, while $BTC is up 22% Even $ETH is only up 27% in the same time frame Most people don't realize that Coinbase continues to be the BEST way to invest in crypto It outperformed $ETH by 3x last bull market and $BTC by 2x Coinbase has the benefit of price speculation from Bitcoin and Ethereum, but it also limits the downside with diversified revenue streams and diversified exposure across the ecosystem Some of these revenue streams are not as impacted by crypto prices: like stablecoins, stock trading, prediction markets and more The last few years Coinbase has worked extremely hard to build out these new revenue streams that put Coinbase into a strong position to perform extremely well during bull markets and create a floor during bear markets Coinbase is also a diversified bet on the crypto ecosystem at large: - they have plenty of exposure to Bitcoin through their balance sheet and custodial operations - they have exposure to ETH through their balance sheet, custodial operations and Base, which is built on Ethereum - they have exposure to Hyperliquid, managing their USDC treasury - they have exposure to DeFi, integrating multiple applications into their application and built on top of Base - they have exposure to Solana, as they have been building and integrating across that ecosystem too - they have exposure to Stablecoins, with a large ownership in Circle and involvement in the OpenUSD stablecoin And then they have a huge venture arm across the entire industry too. If you want to bet on crypto, Coinbase is the easiest way to get access to everything that matters Of course, there are things that make it different from simply holding Bitcoin or ETH, like the fact that you can't self-custody it or that there is execution risk from the team But if you believe in Brian Armstrong, who I think is THE BEST entrepreneur in crypto, then you can sleep well at night holding his company and believing in their teams ability to execute and navigate changing regulations Coinbase has long been one of my portfolios largest positions. I first bought it at $35, near the lows of the last bear market and I continue to be as bullish as ever on the company. $COIN is one of the many ways I have exposure to crypto and the onchain finance revolution You can track my entire portfolio inside of Milk Road PRO, where we have 5 top-tier analysts sharing their real-time portfolios and market research. We're about to increase the price of the membership on Aug 26, but if you sign up today you can lock in current prices for life. Learn more here: For more insights on markets, make sure to give me a follow @kylereidhead

$MU is still valued like we are approaching cycle peak (6-7x). The latest NAND data says that is wrong. (Save this) NAND wafer spot prices went sideways since March and spiked again in August. BofA attributes it to second-tier OEMs restocking into a market with almost no inventory left. Meaning, a "small" order wave was enough to move prices, because there was no spare supply to absorb it. Customers understand that this is the new normal, which is why $SNDK and $MU are now able to lock-in multi-year agreements. Micron has signed 16 take-or-pay agreements, 14 of which carry ~$100B of minimum contracted revenue through 2030! And the part nobody is pricing is the floor. Management says those contracted floor prices deliver gross margins well above Micron's peak in any prior cycle. So, its contracted worst case now beats its historical best case. We have been buying this memory dip aggressively, and Milk Road PRO members get a live notification every time we do. The price rises Aug. 26, so lock in today's rate before then:

$MU is still valued like we are approaching cycle peak (6-7x). The latest NAND data says that is wrong. (Save this) NAND wafer spot prices went sideways since March and spiked again in August. BofA attributes it to second-tier OEMs restocking into a market with almost no inventory left. Meaning, a "small" order wave was enough to move prices, because there was no spare supply to absorb it. Customers understand that this is the new normal, which is why $SNDK and $MU are now able to lock-in multi-year agreements. Micron has signed 16 take-or-pay agreements, 14 of which carry ~$100B of minimum contracted revenue through 2030! And the part nobody is pricing is the floor. Management says those contracted floor prices deliver gross margins well above Micron's peak in any prior cycle. So, its contracted worst case now beats its historical best case. We have been buying this memory dip aggressively, and Milk Road PRO members get a live notification every time we do. The price rises Aug. 26, so lock in today's rate before then:

$BE's pricing power is so underrated. Energy is only 5-10% of total data center cost. So when Bloom raises prices, it barely moves the needle for operators. "They have a lot of flexibility in terms of increasing prices and therefore increasing margins." Bloom's Q2 earnings showed massive growth in operating margins because of this.

$BE's 40% faster installs got the headlines, while the real story is the deployment ceiling it removes. (Save this) Bloom announced Power Connect, which moves the wiring & testing off the construction site and into its factories, so modules arrive close to plug and play. Historically about 60% of the total installation cost was site construction requiring multiple specialized trades. That works fine for a 1 MW project. It does not work anymore in an AI world where Bloom is installing hundreds of MWs (see image below). You cannot scale skilled electricians that easily, so field labor becomes the constraint on deployment. Power Connect moves that constraint into Bloom's own factory, where it controls the process. Three things improve: Speed, which is the entire Bloom value proposition since it competes on time-to-power not price. Scalability, because a standardized factory process replicates across hundreds of MW in a way on-site construction never could. That matters if Morgan Stanley is anywhere close with its 5-8 GW Bloom deployment assumption thorough 2028. Margins, since factory labor is more predictable than field construction, with fewer weather delays, local contractor variables etc. Once again, very bullish news around $BE Details like this rarely make headlines, and digging them out is the whole job inside Milk Road PRO. The price rises Aug. 26, so lock in today's rate for life before then:

$BE's 40% faster installs got the headlines, while the real story is the deployment ceiling it removes. (Save this) Bloom announced Power Connect, which moves the wiring & testing off the construction site and into its factories, so modules arrive close to plug and play. Historically about 60% of the total installation cost was site construction requiring multiple specialized trades. That works fine for a 1 MW project. It does not work anymore in an AI world where Bloom is installing hundreds of MWs (see image below). You cannot scale skilled electricians that easily, so field labor becomes the constraint on deployment. Power Connect moves that constraint into Bloom's own factory, where it controls the process. Three things improve: Speed, which is the entire Bloom value proposition since it competes on time-to-power not price. Scalability, because a standardized factory process replicates across hundreds of MW in a way on-site construction never could. That matters if Morgan Stanley is anywhere close with its 5-8 GW Bloom deployment assumption thorough 2028. Margins, since factory labor is more predictable than field construction, with fewer weather delays, local contractor variables etc. Once again, very bullish news around $BE Details like this rarely make headlines, and digging them out is the whole job inside Milk Road PRO. The price rises Aug. 26, so lock in today's rate for life before then:

There is an AI theme nobody is positioned for yet, and it has nothing to do with chips or models! It is the point where AI agents meet tokenization, and the volume of transactions this unlocks. @jvisserlabs explains it in the video below: Think about borrowing against a house you own. Today a bank underwrites it, and you wait weeks for approval. In tomorrows agentic world an agent makes the call and tokenization settles the loan instantly. This will increase velocity of capital benefiting names like $HOOD massively that are building for this. Recent datapoints supporting this: - x402 processed ~75M AI payments in the last 30 days - Stablecoin card spend hit ~$1B in July, up ~200% YoY. - HOOD already has >100K agentic accounts trading - Elon: Bot internet traffic 1000x human traffic in 5 yrs Did a full deep dive on $HOOD outlining the thesis:

OpenAI's latest print looks uncomfortable! Revenue grew only 18% QoQ while losses remain enormous, which raises the obvious question of who keeps funding the infrastructure boom. The answer is that the funding pool extends far beyond the model companies. The hyperscalers fund capacity out of existing cash flows, NVIDIA and partners are mobilizing >$500B, and Bank of America launched a $250B infrastructure financing initiative (as examples). Private capital is also stepping in. The chart below shows data center deal value hitting $130B in 2025, and Goldman estimates infrastructure and real estate funds still hold ~$1T of dry powder. Importantly, Anthropic is growing over 100% QoQ, showing model economics are not uniformly weak. So, the AI buildout does not need OpenAI to be profitable today. It needs AI to create enough value to incentivize capital to keep flowing into the space. The investor takeaway is that the pressure currently sits in the economics of some frontier models, rather than in underlying AI demand. I only start worrying when weak model economics begin feeding through into weaker capacity commitments and lower CapEx volumes. Separating the real AI risks from the noise is the whole job inside Milk Road PRO. The price rises Aug. 26, so lock in today's rate for life before then:

OpenAI's latest print looks uncomfortable! Revenue grew only 18% QoQ while losses remain enormous, which raises the obvious question of who keeps funding the infrastructure boom. The answer is that the funding pool extends far beyond the model companies. The hyperscalers fund capacity out of existing cash flows, NVIDIA and partners are mobilizing >$500B, and Bank of America launched a $250B infrastructure financing initiative (as examples). Private capital is also stepping in. The chart below shows data center deal value hitting $130B in 2025, and Goldman estimates infrastructure and real estate funds still hold ~$1T of dry powder. Importantly, Anthropic is growing over 100% QoQ, showing model economics are not uniformly weak. So, the AI buildout does not need OpenAI to be profitable today. It needs AI to create enough value to incentivize capital to keep flowing into the space. The investor takeaway is that the pressure currently sits in the economics of some frontier models, rather than in underlying AI demand. I only start worrying when weak model economics begin feeding through into weaker capacity commitments and lower CapEx volumes. Separating the real AI risks from the noise is the whole job inside Milk Road PRO. The price rises Aug. 26, so lock in today's rate for life before then: