Milk Road AI
@MilkRoadAI • 32,074 subscribers
Get smarter about AI investing. Capitalize on the biggest technological change in history across the infrastructure & app layers of AI. By @MilkRoad
Shorts
Videos
This is WILD! One week before SpaceX's historic IPO, Google signed a deal to pay SpaceX $920 million per month from October 2026 through June 2029 for access to 110,000 Nvidia GPUs, CPUs, and related infrastructure (Save this). That is $11 billion per year and up to $30 billion over the life of the contract. This comes less than a month after Anthropic committed $1.25 billion per month for full access to the Colossus 1 data center in Memphis, 200,000+ GPUs, 300+ megawatts of power capacity, through 2029. Two of the most consequential AI labs in the world combined committed value over $70 billion. The question that haunted SpaceX's IPO roadshow was why did Elon keep spending billions constructing Colossus, Macro Hard and Macro Harder, three facilities totaling nearly 2 gigawatts of AI compute when xAI's revenue wasn't yet on the same trajectory as OpenAI or Anthropic? Wall Street was pricing in a risk that Elon was building capacity ahead of revenue which would mean sustained cash burn without a clear payback timeline. That concern was legitimate on its face, because xAI had been aggressive on model development but had not yet demonstrated the enterprise revenue numbers to justify the infrastructure cost. The answer is that the compute itself was always the product. Amazon has AWS, Microsoft has Azure, Google has Google Cloud, Elon just confirmed that he has been quietly building the fourth major hyperscale AI cloud and his first two paying customers are Google and Anthropic, the very companies most aggressively competing in the AI race. xAI's Colossus facility in Memphis was built at a speed that no traditional data center developer could match, it went from groundbreaking to operational in roughly 122 days. That is what happens when you have direct Nvidia relationships, a construction operation built around SpaceX-style execution, and a founder who treats infrastructure buildout the same way he treats rocket launches: compress every timeline and eliminate every bottleneck. The result is that SpaceX now has three operational facilities, Colossus, Macro Hard, and Macro Harder with Macro Hard and Macro Harder in Blackwell architecture running 1.2 gigawatts combined. Colossus 1, built on H100s and optimized for inference, is the facility that went to Anthropic first. The Blackwell-era facilities are where the next-generation training workloads happen and Google's deal suggests they are renting into that capacity as it comes online through the second half of 2026. Elon's compute leasing business would generate approximately $45 billion in incremental annual revenue on top of the mid-$20 billion range analysts had been modeling for SpaceX more than enough to fully subsidize the infrastructure investment and take the financial pressure off xAI delivering immediate AI product revenue. That changes the entire valuation conversation of SpaceX completely! Milk road remains bullish on Space and come join Milk Road Pro and get our full SpaceX IPO breakdown, how we're thinking about the $1.75 trillion valuation and our entire AI thesis. Link below!
Milk Road AI741,539 Aufrufe • vor 2 Tagen
Bill Ackman was asked how he would underwrite SpaceX at $750 billion and his answer was the most honest thing anyone has said about the biggest IPO in history (Save this). "You underwrite SpaceX the way you underwrite a venture capital investment." His business school professor taught him a framework that has guided his entire career, it's people, opportunity, context, deal. On all three of the first criteria, People, Opportunity, and Context Ackman's verdict was the same, SpaceX is one of one, and nothing else in the market comes close. He even acknowledged feeling bad for Blue Origin before noting that their being so far behind is not harmful to SpaceX but rather a structural tailwind that leaves SpaceX with a near monopoly on low cost orbital access for years to come. And at $1.75 trillion, the number SpaceX is actually targeting on June 12, the question is no longer whether this is the best business on earth, but what the present value math looks like when you extend it five years forward and stress test every assumption about Starlink, launch economics, and AI compute revenue. He said that even Amazon is going to have to become a bigger SpaceX customer, because Blue Origin is so far behind that Amazon has no real alternative for low-cost orbital access. He also said something that almost no one is giving enough weight heading into Thursday's listing: "Time has become increasingly valuable in the AI era. You lose a month, you lose a couple months today, and it means a lot." The Colossus and Macro Hard facilities are compounding infrastructure assets where every month of operational delay means less contracted revenue, less negotiating leverage with customers like Google and Anthropic, and a progressively weaker moat against the hyperscalers who are now racing to build competing compute capacity. Come join Milk Road Pro for our full SpaceX IPO breakdown, how we're stress-testing the Deal leg of Ackman's framework at $1.75 trillion, what our five-year revenue model actually looks like, and our full AI thesis. Link below.
Milk Road AI425,291 Aufrufe • vor 1 Tag
Micron will be a $3,000 stock within a few years and Jensen Huang just spent a week in Korea telling the world exactly why (Save this). Jensen announced four new products at the Korea event and every single one of them has memory at the center of its architecture. Vera Rubin, the next generation AI supercomputer, needs massive quantities of HBM. The new Vera CPU needs large amounts of LPDDR5. RTX Spark, the first major PC reinvention in 40 years according to Jensen, needs a lot of LPDDR5. And Nvidia's new robotics and autonomous driving platforms are being built in deep partnership with the Korean memory and electronics ecosystem. Every single growth vector for Nvidia in 2026 and 2027 runs directly through memory and Micron is the only US based company that manufactures all of it. Here is what the numbers look like right now. Fiscal Q2 2026 revenue came in at $23.86 billion, up 196% year over year, with 75% gross margins and $6.9 billion in free cash flow, a quarterly record. Management guided Q3 revenue to $33.5 billion at roughly 81% gross margins, with EPS of $19.15. These are not the numbers of a cyclical memory company but rather the numbers of a company that has been structurally repriced by the largest demand supercycle in the history of the semiconductor industry. The reason the bull case reaches $3,000 comes down to three things that have never been true at the same time in Micron's history. First, the entire 2026 HBM supply is already sold out under multi-year contracts. CEO Sanjay Mehrotra told analysts that Micron can currently only fulfill 50% to two thirds of key customers' HBM demand at any price. Second, Micron has begun volume shipment of HBM4 12-Hi specifically for Nvidia's Vera Rubin platform, the exact product Jensen was talking about in Korea and has signed its first five year strategic customer agreement, converting what was historically a quarterly negotiation business into something closer to a long-term recurring revenue model. Third, Wolfe Research's bull case model points to $160 billion in calendar year 2027 revenue and $80 in EPS. At even a 20x earnings multiple, modest for a company with this growth profile, that is a $1,600 stock. UBS has already tripled its price target to $1,625. The path to $3,000 requires HBM4 to ramp smoothly, supply constraints to persist into 2027 as Mehrotra says they will, and hyperscaler AI capex to continue growing at its current trajectory, all three of which Jensen Huang just confirmed in Seoul. The HBM total addressable market alone is projected to reach $100 billion by 2028, a forecast Micron itself already pulled forward two years ahead of schedule because demand arrived faster than anyone modeled. Micron trades at roughly 9x forward earnings today. That is cheaper than a grocery chain, for a company growing revenue at 196% year over year, with its entire production sold out, supplying the infrastructure for the most important technology buildout in history. Come join Milk Road Pro for our full breakdown of the Micron bull case how we think about the HBM4 transition timeline, what multi-year customer contracts mean for Micron's valuation multiple expansion, and our entire AI thesis. Link below!
Milk Road AI164,984 Aufrufe • vor 22 Stunden
Chamath just delivered the clearest diagnosis of what is happening to enterprise software and the OpenAI Deployment Company is the most damning piece of evidence he could have picked. "The low end of the market is basically finished. There is no safe space." 90% of public SaaS stocks are down 30-80% from their 52 week highs, the median software stock is now negative over the last 3-6 months. Goldman Sachs reported that software forward P/E multiples fell from 35x to 20x, the lowest absolute level since 2014 and the smallest premium to the S&P 500 since 2010. The low end died first and fastest, because AI replaced it most directly. The small business tools, the lightweight project managers, the single function SaaS products that charged $49 a month per seat, those are being replaced by AI agents that do the same work as a workflow, not a product. You do not buy an AI powered tool, you describe what you need and it builds it and the seat based model that created the SaaS industry simply does not apply to that transaction. But Chamath's more interesting argument is about the high end and the tell he points to is perfect. OpenAI just raised $4 billion from 19 investors including TPG, Brookfield, Bain, and McKinsey to launch a consulting company and guaranteed those investors a 17.5% annual return to do it. On $4 billion in committed capital, that is roughly $700 million per year in guaranteed payouts, owed by a company that is projected to lose $14 billion in 2026. The goal of this venture is to compete directly with Deloitte, PwC, Ernst & Young, Andersen, and Cognizant. Think about what that structure reveals. OpenAI lost half of its enterprise LLM API market share from 50% to 25% between late 2023 and mid-2025, with Anthropic now leading at 32%. Its response was not to build a better model but rather to raise $4 billion, offer guaranteed PE-tier returns and hire embedded engineers to physically sit inside client organizations and make AI actually work in production. The reason, as Chamath identified, is that the high end of the market is not easy. "It's not like boop boop boop, put in a prompt and beep bap boop, it all works," he said and the data confirms exactly that. 88% of organizations running AI agents reported a security incident in the past year, 42% of C-suite executives say AI adoption is creating internal organizational conflict. The average enterprise AI consulting implementation costs $228,000 in year one versus $77,000 for platform-based approaches and most still stall before reaching production. Anthropic immediately matched OpenAI with a competing $1.5 billion consulting venture backed by Blackstone, Goldman Sachs, and Hellman & Friedman bringing the combined spend by the two leading AI labs on human powered enterprise deployment to $5.5 billion in a single month Chamath's read is that the high end, the large enterprise platforms like Salesforce with proprietary data flywheels, Palantir with its FDE model already proven at scale, Oracle with vertical specific data moats will survive and consolidate. The mid-market point solutions, the single function tools, the lightweight enterprise apps without defensible data assets, those are on the conveyor belt. The AI industry is not just disrupting the companies that use software but rather disrupting the companies that sell it.
Milk Road AI1,651,965 Aufrufe • vor 22 Tagen
Larry Ellison, the man who built Oracle into a $500 billion enterprise software empire and he said something that every investor needs to hear (Save this). "By 2029, I can guarantee you, AI is not going to be the problem." The problem is going to be compute specifically, who has enough of it and who does not. Ellison described the current AI race in terms that strip away all the abstract commentary about models and capabilities and reduce it to the one thing that actually determines who wins: "Me and Elon begging Jensen for GPUs. Please take our money. We need you to take more of our money, please." Citigroup raised its forecast for AI infrastructure spending to $2.8 trillion through 2029, with hyperscalers already spending at a $490 billion annual rate by end of 2026 and the firm estimates global AI compute demand will require 55 gigawatts of new power capacity by 2030 at a cost of approximately $50 billion per gigawatt. Sam Altman publicly thanked Jensen Huang this past March for significantly increasing NVIDIA's capacity at AWS, the CEO of the most important AI lab in the world writing a thank you note to the chip supplier because compute is still the binding constraint on everything OpenAI wants to build. Ellison's point about getting there first is the part of this clip that deserves a second read. He named three specific races, self-driving, reading cancer biopsy slides, and synthesizing video and said that being first in each one is a big deal. The logic is that in winner take most AI markets, the first mover trains the best model, the best model attracts the most usage, the most usage generates the most data, and the most data trains the next best model, a compounding loop that the second-place finisher never catches up to. "The guys in this race are very smart and they understand they need to be best at something," Ellison said. What makes this clip so important right now is the timing. The AI GPU chip market is projected to grow at a 32.4% CAGR through 2029, reaching $145 billion in incremental spend, and NVIDIA's data center revenue is already running at a pace that would have seemed impossible three years ago. Every major hyperscaler, Microsoft, Amazon, Google, Oracle, Meta is no longer funding AI capex from operating cash flows alone, they are borrowing to keep up, because falling behind in compute now means ceding the winner-take-most race Ellison just described. At Milk Road, we have been positioned in NVIDIA, AVGO, AAOI, MU, and Bloom Energy and more. Come join Milk Road Pro and get the full picture on how we are playing every layer of the GPU demand supercycle that Larry Ellison just guaranteed will not slow down before the end of the decade, link below/bio.
Milk Road AI479,776 Aufrufe • vor 9 Tagen
This is WILD! MIT just solved one of the hardest unsolved problems in robotics (Save this). For decades, the fundamental problem with soft robots and wearable exoskeletons has not been compute or AI, it has been actuation. The moment you try to give a soft robot meaningful strength, you run into the same wall every engineer has hit since the field began, fluid-driven systems require external pumps, hydraulic reservoirs, and heavy infrastructure that makes the entire thing impractical to wear or embed into fabric. MIT's new Electrofluidic Fiber Muscles solve that problem by eliminating external infrastructure entirely. The key insight is electrohydrodynamic pumping using electric fields to generate pressure directly from electricity, with no moving parts, no motors, and no external fluid reservoir. The fibers are less than 2 millimeters thick, can be woven into fabric like ordinary textile, and operate in complete silence because nothing physically moves inside them, it is just ions propelling fluid through a closed circuit. The performance numbers published in Science Robotics are not conceptual, they are empirical results from actual hardware. These fibers achieve a power density of 50 watts per kilogram, matching skeletal muscle, with a contraction strain of 20% and a response time of 0.3 seconds. A single bundled configuration lifted 4 kilograms, 200 times its own weight while a separate configuration drove a robotic arm through a 40-degree bend compliant enough to safely complete a human handshake. Another configuration launched objects in under 100 milliseconds, which is faster than a human flinch reflex. The design mirrors biological muscle architecture in a way that prior artificial muscle approaches never achieved. The fibers are organized into antagonistic pairs, one contracts while the other extends, exactly like biceps and triceps and because the system runs in a closed loop, the relaxing fiber serves as the fluid reservoir for the contracting one, which is what allows the whole system to operate untethered with no external tank. The applications are not hypothetical but rather are the exact use cases the industry has been waiting years for the hardware to catch up to. Exoskeletons for physical labor, prosthetic limbs that move with the natural compliance of biological tissue, assistive garments for patients with motor disorders, and soft robots capable of safe physical contact with humans are all immediately unlocked by a muscle technology that is silent, lightweight, and weavable into clothing. The deeper significance is what this technology does when it meets the AI robotics wave that is already underway. Every major humanoid robot program, Figure, 1X, Boston Dynamics, Tesla Optimus is currently bottlenecked by the same hardware limitations these fibers address, actuators that are too rigid, too loud, too heavy, or too dependent on infrastructure to operate naturally alongside humans. Electrofluidic fiber muscles do not just solve a materials science problem but rather they remove one of the last physical barriers between robots that live in labs and robots that live in the world.
Milk Road AI1,198,732 Aufrufe • vor 23 Tagen
The man who turned $225 million into $13.7 billion told you something most people still haven’t absorbed (Save this). And this video is from before any of it played out the way he said it would. GPT-2 was released in 2019, a model so brittle that OpenAI withheld it for months out of concern, and that could barely hold a coherent paragraph before veering into nonsense. Four years later, GPT-4 arrived and scored in the top 10% of bar exam takers, passed the medical licensing exam and performed at expert level across standardized tests that GPT-3.5 barely cleared at median. That is a preschooler to smart-high schooler jump in four years and Aschenbrenner’s entire thesis is that another jump of exactly that magnitude, on the same timeline, gets you to AGI. The part of the clip that is both funny and devastating is the $20/month observation. Aschenbrenner calls out Ross Douthat, one of the most prominent opinion writers in the country for complaining that ChatGPT couldn’t pronounce his name, then using that as evidence that the singularity wasn’t imminent. What Douthat had used was the free tier. GPT-3.5, the little green icon, not GPT-4. The deeper point is not funny at all, the most influential voices in culture were forming their entire worldview on AI’s capabilities off a degraded, rate-limited product that was already two generations behind what researchers were actually working with inside the labs. The benchmarks now fully vindicate the trajectory Aschenbrenner was describing. Stanford’s 2026 AI Index Report found that on SWE-bench Verified, a real software engineering benchmark performance jumped from 60% to near 100% in a single year. Old benchmarks kept getting retired because frontier models maxed them out faster than new ones could be designed. The 2025 retrospective on Situational Awareness found his core predictions compute scaling, capability jumps, and the timing of AI displacing knowledge workers had largely tracked ahead of schedule. Then he put his money behind it in the most direct way possible, and the fund grew from $225 million to $13.7 billion in AUM. His core bets were on the physical infrastructure layer of AI, power, compute, and the companies building the backbone of the trillion-dollar cluster he described precisely in that video. Nebius with 684% year over year revenue growth, sold out GPU capacity, and a $17.4 billion Microsoft contract is the clearest pure-play expression of that infrastructure trade in public markets today. He just filed to disclose a 5.6% stake: 12.4 million shares. Nebius is a core Milk Road position, and our subscribers are up massively. Come join Milk Road Pro to get our full thesis, link below!
Milk Road AI323,621 Aufrufe • vor 11 Tagen
This is WILD! Anthropic just became the most valuable AI company on earth and what Chamath said months ago explains exactly why this moment matters (Save this). Anthropic closed a $65 billion Series H round at a $965 billion post-money valuation surpassing OpenAI's $852 billion valuation from March and making Anthropic the highest-valued private company in history. Just three months ago, in February, Anthropic had raised $30 billion at a $380 billion valuation meaning the company nearly tripled in value in a single quarter. Claude's run rate revenue crossed $47 billion today, up from $30 billion in April, up from $9 billion at the end of 2025, a pace of revenue growth that has no comparable precedent in business history. Now go back and watch what Chamath said, because he called the entire arc of this. "I've never seen a business like this. And I'd say the same thing about Anthropic. Nobody in the history of the world has ever seen two businesses like this at this scale. These are trillion dollar companies. They both are. And they both deserve to be." He said that before the $965 billion number. Chamath Palihapitiya also said something that most people skipped over, that OpenAI and Anthropic need to get public as fast as humanly possible because of what happens after. Chamath laid out a specific sequencing thesis, SpaceX goes public first and does great, the next company does good to great, then appetite runs out, because the market simply cannot absorb trillions of dollars of new demand in rapid succession. Today, Anthropic's $65 billion round may be precisely the move that locks in its position before that window narrows fortifying the balance sheet before the public markets get crowded. But Chamath's deeper warning cuts through the celebration, once SpaceX, OpenAI, and Anthropic are all public, the AI technology baked into all three will cannibalize the moats of every other tech company, compressing tech sector P/E ratios toward non-tech levels and making the software businesses of the last decade obsolete. "It will eliminate and it will cannibalize and it will erode most of the moats that support this differential trading," he said directly. "I'll buy the first five or six years of this story, but I'm not buying year 15 of this anymore because these three guys are going to build something." The companies getting valued at near-$1 trillion today are not just winning but rather are the instruments by which everything else eventually gets repriced.
Milk Road AI293,046 Aufrufe • vor 10 Tagen
The man who turned 225 million dollars into 5.5 billion dollars explained on camera exactly why he made his biggest bet. This is Leopold Aschenbrenner, the same person whose Bloom Energy position is now worth close to 2 billion dollars after Oracle's 2.8 gigawatt fuel cell deal laying out the power math that drove every investment decision his fund has made. In 2022, the GPT-4 training cluster consumed roughly 10 megawatts of power and cost about 500 million dollars. AI compute has been scaling at roughly half an order of magnitude per year meaning the largest training cluster doubles in power requirement every 12 to 18 months without stopping. By 2024, the largest cluster was approximately 100 megawatts, the equivalent of 100,000 high-end GPUs and costs in the billions. By 2026, right now, the leading training cluster requires a full gigawatt of continuous power and that is the output of a large nuclear reactor. By 2028, the projection reaches 10 gigawatts, more electricity than most US states generate in total. By 2030, the trillion-dollar cluster, 100 gigawatts, over 20 percent of everything the United States currently produces in electricity, consumed by a single AI training installation. And that is just the training cluster. Inference, the continuous compute required to actually run AI products for hundreds of millions of users requires multiples of that on top. Meanwhile, total US electricity production has barely grown five percent over the last decade and the grid was not built for this. And the transformer shortage, the switchgear backorders, and the canceled data center projects that are making headlines right now are the first visible symptoms of a power system hitting a wall that Aschenbrenner saw coming years before the rest of the market. This is exactly why he built a 875 million dollar position in Bloom Energy, a company that generates electricity directly at the data center site using fuel cells, completely bypassing the grid bottleneck that is already stopping half of all planned US data centers from opening on schedule. The thesis was never complicated. The bottleneck in AI is not the models, not the chips, and not the software. The bottleneck is whether civilization can generate enough electricity to run the machines fast enough to matter.
Milk Road AI1,217,301 Aufrufe • vor 1 Monat
Cathie Wood just flagged the sleeper trade inside the AI boom that most people are completely missing. Everyone has been chasing GPUs. Nvidia, the data center buildout, the chip arms race. That trade has been obvious for two years. But OpenAI's CFO Sarah Fryer said something quite different: people are going to be really shocked by how agentic AI activates CPUs. Right now, for every CPU in an AI workload, there are 4 to 5 GPUs. That's the current ratio. Wood thinks that ratio is going to 1 to 1. Think about what that means. AI inference at scale, agents running autonomously, pipelines executing tasks across systems. The compute mix shifts dramatically away from pure GPU dominance. CPUs become a first-class citizen in the AI stack. Cathie called it going "back to the future." Intel has taken off. Flex (formerly Flextronics) is booming. Stocks that were giants in the dot-com bubble are resurging because the underlying demand for their products is real again. The GPU trade made sense at the training stage. You need massive parallel compute to train frontier models. But agentic AI runs differently. Agents are constantly orchestrating, reasoning, calling APIs, executing workflows. That workload looks a lot more like traditional computing. And traditional computing runs on CPUs. If Cathie Wood is right about the ratio collapsing to 1:1, the CPU demand signal embedded in the AI buildout is orders of magnitude larger than the market is currently pricing.
Milk Road AI234,536 Aufrufe • vor 14 Tagen
Microsoft canceling its internal Claude Code licenses this week likely has almost nothing to do with AI costs becoming untenable. Microsoft's own statement says Claude models remain accessible through Copilot CLI and Anthropic's Claude still runs inside Microsoft 365. The real story is that Microsoft's own product (GitHub Copilot) was being embarrassed by a competitor internally this is a platform strategy play, not a cost crisis call. And the post assumes token prices are rising and unsustainable but the actual data goes the other way. As Sam Altman said, a hard reasoning problem that cost X on the OpenAI API 18 months ago now costs 1,000x less. 1,000x in 18 months is just that doesn't happen very often. His stated mission is to relentlessly drive the cost of intelligence down as close to zero as possible and make it a low-cost asset available to the world. A16z separately documented a 10x cost decline per year for equivalent model performance and Nvidia's Blackwell platform delivered 4x to 10x inference cost reductions in production deployments in early 2026. Yes, Uber burned through its 2026 AI budget in four months. But look at why, 95% of engineers were using AI tools monthly, ~70% of committed code was AI-generated, and 11% of real-time backend updates were made autonomously by agents with zero human intervention. That's not an AI cost problem. That's a CFO who budgeted for a pilot and got a revolution. Also, this post ignores the Jevons Paradox, when a resource gets cheaper, total consumption rises, not falls. Tokens inferenced on platforms like OpenRouter grew 25-fold since December 2024, OpenAI’s revenue grew at roughly a 250% annual rate to around a $20B run rate, while Anthropic’s surged from roughly $9B to over $30B in just a few months. Those aren't the numbers of companies whose economics are imploding. Even Gartner projects inference on a 1-trillion parameter model will cost 90%+ less by 2030. The actual Gartner warning is that agentic AI costs more per task because agents run longer chains, that's a design and scoping problem, not a structural industry collapse.
Milk Road AI250,588 Aufrufe • vor 17 Tagen
The CEO of the world's largest asset manager just said something that should reframe how every investor thinks about the AI trade. Larry Fink, managing $11.5 trillion at BlackRock, stood at the Milken Institute Global Conference and said four words that matter, "We just don't have enough compute." "The United States is short power. We're short compute. We're short chips. And there's going to be shortages in all three and memory, four things. I actually believe a new asset class will be buying futures of compute." Think about what that means. Fink is predicting that compute becomes a tradable commodity like oil, like grain, like natural gas where investors buy forward contracts on future capacity because the shortage is so structural and so predictable that a derivatives market will emerge to price it. That is not a minor observation from a finance executive but rather the chairman of the most powerful capital allocator on the planet telling you that compute scarcity is a multi-year, investable megatrend. The data backs him up completely. Data centers will consume 70% of all memory chips produced globally in 2026. Advanced HBM production from Samsung, SK Hynix, and Micron is sold out through 2026 and into 2027 and a single AI server consumes 10-20x more memory than a conventional workload server. DRAM supply growth is running at just 16% annually while AI infrastructure demand is growing at 80%+. The chip crunch, the power crunch, and the compute crunch are not temporary dislocations, they are structural, and they will get worse before they get better. Fink also said something the bears keep getting wrong: "There is not an AI bubble. There is the opposite. We have supply shortages. Demand is growing much faster than anyone has ever anticipated." This is why the Milk Road Pro portfolio is built the way it is, long the companies producing and supplying the constrained resources: chips, memory, compute infrastructure, and power. Check out Milk Road Pro, link below to access our full thesis and plays.
Milk Road AI418,026 Aufrufe • vor 1 Monat
Nebius will be a TRILLION dollar company and here is exactly why (Save this). Brad Gerstner's Altimeter just said on camera that they are invested in ClickHouse, and explained exactly why in one sentence: "If you're in the data infrastructure layer, then token consumption is driving a lot more consumption of your basic services." The flip side of that point is equally important. Gerstner added that the closer you are to a point solution, a single use app built on top of AI, "that feels like you're on the front of the conveyor belt heading toward the guillotine." Models get better, apps get commoditized and the companies that own the foundational infrastructure that every AI application must run through keep compounding. ClickHouse is exactly that foundational layer. It is a real time analytical database engine originally built inside Yandex, optimized for the exact query patterns that AI agents, LLM observability pipelines, and machine learning infrastructure generate, massive write volumes, complex aggregations, and sub-second response at scale. It processes hundreds of billions of rows per second, serves over 2,000 enterprise customers including Cloudflare, Uber and ByteDance, and grew 300% in a single year. In January 2026, a $400 million Series D valued ClickHouse at $15 billion more than double its $6 billion valuation just eight months prior. Here is where Nebius comes in. Nebius holds a 28% stake in ClickHouse, an asset that traces back to its Yandex origins. At ClickHouse's current $15 billion valuation, that stake is worth approximately $4.2 billion, sitting largely unrecognized on Nebius's balance sheet while most market coverage focuses entirely on the AI cloud business. A ClickHouse IPO, which the company is actively positioning toward, would force the market to mark that position to full public market value for the first time and could alone reprice Nebius meaningfully. But that hidden asset is just one layer of the bull case. The core AI cloud business just printed 684% year over year revenue growth, $399 million in Q1 2026 against $50 million a year prior. AI specific revenue grew 841% and now represents 98% of total revenue. The moat underneath those numbers is 3.5 gigawatts of secured power capacity, a $27 billion five year contract with Meta, a $2 billion strategic investment from Nvidia, and a Microsoft partnership ramping to full run rate in 2027, all stacked on top of a ClickHouse stake that the market is still not fully pricing in. Milk Road Pro remains massively bullish on Nebius, we called it early, we are up huge on the position, and we continue to track every development across AI infrastructure before it becomes obvious to the rest of the market. Come join us to see our full Nebius thesis and every other position in the portfolio, link below!
Milk Road AI216,036 Aufrufe • vor 22 Tagen
This is one of the craziest AI launches of 2026 and it came out of basically nowhere (Save this). A company called Subquadratic just shipped SubQ, and the benchmarks are almost hard to believe. To understand why this is such a big deal, you have to understand the fundamental problem that has defined AI for the last decade. Every large language model in existence is built on transformer architecture, and transformers use a mechanism called standard attention that checks every single word in a sequence against every other word. Double the context length and compute doesn't double, it quadruples, triple it and compute goes up nine times. This quadratic scaling is why frontier models have been stuck at roughly 1 million tokens, why running them at those lengths gets expensive fast, and why the AI labs have essentially been printing money charging you more the longer you need the model to think. The industry has known this problem existed since 2017 but they scaled it anyway. SubQ is built from the ground up to solve it. Instead of processing every possible token relationship, SubQ's sparse attention architecture identifies which relationships actually matter and ignores the rest meaning compute is used where it counts and wasted nowhere else. The result is that compute scales linearly with context length instead of exponentially, and the implications of that one architectural shift are enormous. At 12 million tokens, SubQ reduces attention compute by nearly 1,000x compared to standard frontier models and at 1 million tokens, it runs 52x faster than FlashAttention. And it does all of this while posting frontier level accuracy, scoring 95% on the RULER 128K long-context benchmark versus Claude Opus 4.6's 94.8%, and an 81.8 on SWE-Bench Verified coding tasks, besting Opus 4.6 (80.8) and DeepSeek 4.0 Pro. The cost comparison is where it gets genuinely insane. SubQ runs at under $1.50 per million tokens less than 5% of what Claude Opus charges. On the RULER benchmark, running the test with SubQ cost $8, running the same test with Claude Opus cost $2,600 and that's a 300x cost reduction at equivalent or better accuracy.. Subquadratic launched with $29 million in funding, SubQ is available today for early access via API, and SubQ Code, a coding agent built on the architecture ships alongside it. The transformer has been the unchallenged foundation of every major AI system since 2017. SubQ is the first serious evidence that something structurally better might have just arrived.
Milk Road AI277,002 Aufrufe • vor 1 Monat
Chamath Palihapitiya just laid out the most important valuation question nobody on Wall Street wants to answer. For 20 years, the Mag 7 won because they had the greatest business model ever invented, asset- ight software. You write the code once, you sell it to a billion people, the marginal cost of the next customer is basically zero. There is essentially no factories, no raw materials, no union workers, no physical infrastructure, just pure leverage, scale the revenue, barely scale the costs. That's how you get 30x, 50x, 60x earnings multiples and the market was paying for compounding economics that had no natural ceiling. But AI just blew that model up. The hyperscalers, Amazon, Microsoft, Google, Meta are now projected to spend between $600 and $725 billion on capex in 2026 alone, up from $250 billion just two years ago. That number is climbing, not plateauing and it's not just the chips and the data centers, it's the energy contracts underneath all of it. When Microsoft re signed Three Mile Island, they locked in a 20 year forward purchase agreement at more than $100 per megawatt hour nearly double the prevailing spot rate of $60 for wind and solar in the same region. That's a long term liability commitment baked into operating cash flows for two decades. Here's where Chamath's math gets uncomfortable. These five or six companies are now collectively spending so much that their capex has exceeded their free cash flow meaning they can no longer self fund growth from operations alone. In 2025 alone, hyperscalers raised $108 billion in new debt and projections put the total debt issuance over the next few years at $1.5 trillion. These are companies that, for two decades, were net cash accumulators and now they're going to the debt markets like everyone else with term loans, revolvers, and structured credit facilities. That's Chamath's core point and it's a devastating one for anyone still modeling these companies the old way. When a company is asset light, investors pay a premium for that lightness and the multiple reflects the belief that returns on capital will stay high indefinitely, because there's no heavy physical plant dragging them down. But when Google starts looking like a utility locked into 20-year energy contracts, carrying hundreds of billions in debt, spending half its revenue on physical infrastructure, the rational multiple compresses. You don't price a utility at 30x earnings, you price it at 12x. His conclusion is that stop trying to value the hyperscalers themselves and follow the money instead. A trillion dollars a year is flowing out of these companies into power companies, data center operators, chip manufacturers, cooling systems, fiber networks, rare earth metals. The companies on the receiving end of that spending are already underpriced because the market is still staring at the senders while ignoring who's cashing the checks. The asset-light era minted the most valuable companies in human history and the asset heavy era that's replacing it might be the best argument yet for owning everything around them instead.
Milk Road AI268,305 Aufrufe • vor 1 Monat
Micron is one of the most UNDERVALUED stocks in the entire AI trade right now and everyone should be buying at these prices. (Save this). Jensen laid out the situation in one sentence, the supply chain is lined up, the HBM is lined up with the Grace Blackwell GPUs, the only problem is that demand is much greater than the overall capacity of the world. And Michael Dell said it before Jensen even finished that memory is the single biggest supply constraint in the entire AI buildout right now. Every HBM chip that Micron, SK Hynix, and Samsung produce consumes three times the silicon wafer area of standard DRAM. Nvidia's Rubin GPU requires 288GB of HBM per chip, a 260% increase over the H100 in just two generations. Every major hyperscaler has locked up contracts through 2026, and Micron has said publicly it can only fulfill about two-thirds of medium-term demand for some customers. And it's HBM production is sold out entirely for 2026 and HBM4 is also already sold out. The numbers tell the story, DDR4 spot prices surged roughly 15x in eight months. DRAM contract prices rose 90-95% in a single quarter, TrendForce called it "essentially unprecedented" in the history of the memory market. Micron has rallied roughly 68% year to date in 2026, and yet it still trades at a P/E of 37.6x against an industry average of 75.3x. The shortage does not resolve until new fabs come online, Micron's new factories are not producing until 2027 and 2028 at the earliest, and the memory shortage is forecast to run until at least 2027. Milk Road Pro has been covering the HBM memory trade as a core AI infrastructure thesis before it became a consensus Wall Street call and our Pro members are already up massively in $MU. Come join us at the link in bio/below to see our full portfolio and the names we're watching before the rest of the market catches on.
Milk Road AI146,326 Aufrufe • vor 20 Tagen
The man who turned 225 million dollars into 5.5 billion dollars just laid out on camera exactly when he believes the world changes permanently with specific dates. Leopold Aschenbrenner's argument follows a single trend line that has held for over a decade without breaking. Right now in 2025 and 2026, the models being built are already smarter than most college graduates across the board. By 2027 and 2028, AI hits expert level as capable as the best professionals in any field operating not as a chatbot but as what he calls a drop-in remote worker. You assign it a project, It goes off, writes drafts, runs tests, iterates, and comes back with finished work fully autonomously, for hours at a time. The key unlock he describes is what he calls unhobbling, today's models are already more capable than most people realize, but artificially constrained by how they are deployed. Once agents can use computers freely and run long-horizon tasks without human checkpoints, the economic value unlocks almost overnight. His best guess for true AGI is the 10 gigawatt cluster range, a single data center drawing more electricity than most US states produce in total. By 2030, the trillion-dollar training cluster consumes over 20 percent of all US electricity production for a single training run. This is the direct line between that prediction and his 875 million dollar Bloom Energy position. He did not buy a power company because he liked the chart but rather bought a power company because he ran the math on what AGI physically requires to exist, and concluded that electricity is the asset class of the decade. The position is already worth close to 2 billion dollars, and his own timeline says the demand that drove it is just getting started.
Milk Road AI375,498 Aufrufe • vor 1 Monat
David Sacks just said what every honest analyst in Silicon Valley is already thinking (Save this). Nobody has ever seen anything like this. Anthropic has grown at 10x per year for three straight years and going into 2026, the conventional wisdom was that the rate of growth had to slow at this level of scale but then the numbers came in. Q1 alone is $10B ARR to $30B, in April, $30B to $44B and that's $96 million in new ARR added every single day. Inference margins are now above 70%, up from 38% last year and the only thing holding them back was compute. That's solved now, the SpaceX deal and others Anthropic has been quietly signing unlocks the supply side. This is exactly why we are bullish on Nebius and AMD. When a single company is adding nearly $100M in ARR per day, the real trade isn't the frontier lab but rather the infrastructure underneath it. Nebius, one of the fastest-growing neoclouds on the planet posted 547% YoY revenue growth in Q4 2025, exited the year with $1.25B ARR, and is guiding for $7–9B ARR by year-end 2026. Their revenue backlog has reached $46B, with projections of $16B in revenue by 2028 and NVIDIA locked in a $2 billion stock buy agreement with them giving Nebius early access to cutting-edge chips while every other cloud scrambles for supply. AMD is the other side of the same coin. Data center revenue hit $5.78B in Q1, up 57% year-over-year with total company revenue at $10.25B, up 38%. Meta has committed to deploying up to 6 gigawatts of AMD Instinct GPUs. Data center GPU revenue is forecast to surge 114% year over year to $15B in 2026. MI400-series chips hit the market in H2 and analysts project segment operating margins climbing to 31% as the next generation ramps. The model is simple, Anthropic is printing revenue and that that revenue pays for compute. That compute flows through companies like Nebius and AMD. This is why Milk Road PRO remains bullish on them and our positions are up massively. Our analysts have broken down the full thesis, the allocations, and the price targets. Go PRO at Milk Road to see everything, link below!
Milk Road AI184,643 Aufrufe • vor 1 Monat
Chamath Palihapitiya, one of the most connected investors in tech and his warning is the clearest framing of the AI compute crisis anyone has put into words. "It is a five alarm fire for them. They need to have land, power, shell." He's talking about Anthropic and OpenAI and the threat he's describing is called the Friendster effect. Friendster was the dominant social network before MySpace and Facebook and it didn't lose because it had a bad product but rather it lost because it couldn't keep the site up. Demand outpaced infrastructure, the experience degraded, and users left for one that actually worked. Chamath's argument is that OpenAI and Anthropic are approaching exactly that moment. The numbers are already showing it, Anthropic is growing so fast that it had to cut Claude's thinking depth during peak hours, cap agentic sessions, and test removing Claude Code from its $20 plan entirely. GitHub Copilot paused new signups, paying enterprise customers are hitting usage walls they've never seen before. Dario Amodei himself admitted there is "no hedge on earth" against the risk of over-purchasing compute meaning he's deliberately staying lean on capacity, even as the demand wall approaches. The core problem is structural, OpenAI and Anthropic grew up renting capacity from hyperscalers AWS, Azure, Google Cloud. That was fine when they were small but now they're so large that dependency is a strategic liability. Every token they sell runs on someone else's infrastructure and every capacity decision belongs to someone else. And when demand spikes faster than anyone planned, there's nothing they can do in real time except throttle. Building your own infrastructure takes 18 to 24 months minimum, you need to acquire land, secure power, construct shell and none of that happens fast. That's why Google's $40 billion commitment to Anthropic this week is about more than just money but rather about securing the land, power, and infrastructure that Anthropic needs to not become Friendster. Whoever controls the compute controls the frontier and right now, the AI labs with the best products are the most dependent on infrastructure they don't own.
Milk Road AI260,300 Aufrufe • vor 1 Monat
Marc Andreessen just coined a term that perfectly describes what's actually happening to programmers right now and it's the opposite of what the doomers predicted (Save this). He calls them AI vampires. Andreessen's says that programmers using Codex, Claude Code and AI coding tools are not being replaced but they're working harder than ever, sleeping less than ever, with massive bags under their eyes and they are completely euphoric. What's remarkable is that the phenomenon extends far beyond professional engineers. Andreessen described an a16z partner who had never written a single line of code in his career, who built an entire AI powered work system for himself and when asked if he'd ever looked at the underlying code, the answer was simply "hell no." The data behind the anecdote is extraordinary. Andreessen says the leading-edge programmers at a16z portfolio companies are now 20x more productive than they were a year ago, the most dramatic increase in programmer productivity in the history of the industry. The METR May 2026 AI usage survey found technical workers self reporting a 1.4–2x change in work value from AI tools, with 75% of software engineers using AI for at least half their work. The software engineer hiring rate is actually increasing up to 22.77% of new hires in 2025 from 19.32% in late 2023 and companies are now bidding more aggressively for senior engineers specifically because AI empowered engineers have a higher ROI than ever before. The US economy added 115,000 jobs in April 2026 alone, beating the 62,000 consensus forecast precisely as AI adoption hit its highest level on record. This is exactly what basic economics predicts and what almost no one who writes about AI and jobs bothers to say. Classic marginal productivity theory says: when you increase the productivity of a worker, you don't diminish human work, you expand it. The worker becomes more productive, gets paid more, does more, and more jobs are created in the process. Andreessen's ATM analogy holds here because ATMs were supposed to eliminate bank tellers but instead, teller employment rose because lower operating costs let banks open more branches. The no-code AI market has exploded from $4.3 billion in 2023 to $21.2 billion in 2026 not because programmers are being replaced, but because the universe of people who can now build software has expanded by orders of magnitude. The blind spot, as Andreessen notes, is that productivity is now outrunning comprehension. The a16z partner building AI systems he's never looked at the code for represents something genuinely new, software being summoned faster than it can be understood. That's not necessarily dangerous but it does mean the verification, security, and governance layer of the AI development stack is more important now than it has ever been.
Milk Road AI136,646 Aufrufe • vor 26 Tagen