Your blueprint to holistic wealth. Wall Street Veteran & Market Maverick Samurai : 30+ years in hedge funds, PhD in Economics & Computer Science. Father of 5, sharing insider insights to level the playing field. Substack is the new town square.
Cite asCryptoQuant has tracked the directional calls on MSFT by @eltoromarketinsights on Substack since May 2024: not yet scored on MSFT (as of Nov 18, 2026).
Other assets covered (11)
3 MSFT calls on record, a thin history.
Relative view by default: fewer than 30 MSFT calls or days on record.
Sentiment timeline shows relative sentiment within this analyst's history. A perma-bull showing 8 bullish : 2 bearish in a bear market sets that as their baseline. If you notice any errors, claim @substack:eltoromarketinsights to submit corrections.
Hover the chart to read the thesis of that period · click a day to pin it
Follow the Capex - 6 Trends - 30 Names Driving Tech Trends 2026
The Signals Shaping What’s Next. How $700 Billion in AI Spending Creates the Opportunity of the Decade Hi and welcome back for a Quant data driven analysis . [ Full Disclaimer ] Unlock Premium - Dive into my quant portfolios, real-time alerts on regime shifts , and deep-dive models on additional hidden gems. Supercharge your edge and unlock now your hedge for 2026 Share Tech Trends 2026 I spend most of my time watching data, reading earnings calls, and tracking the money flows behind technology companies. The signal-to-noise ratio in tech coverage right now is awful. Everyone talks about AI like it is the only thing happening. AI sits near the center of nearly every trend I am tracking, but treating it as one big theme misses the real picture. The real picture is that AI is splitting into specialized lanes. Infrastructure spending, edge deployment, autonomous systems, cybersecurity arms races, and full-scale software rewrites. Each of these branches creates distinct opportunities with varying risk profiles and timelines. Some of these are 12-month plays. Some are 5-year holds. If you are not separating them, you are probably paying too much for the wrong exposure. Here I breaks down the six tech trends I think matter most for 2026, explains why the math behind each one works, and identifies 30 US names best positioned to benefit. Fifteen are large-cap leaders. Fifteen are mid-cap names with higher upside potential and better risk-to-reward ratios for those willing to size appropriately. These are not vague “buy tech” ideas. Each pick ties back to a specific signal I am watching in the data. Trend Map Subscribe now 1. The AI Infrastructure Build-Out Is Not Slowing Down Five companies, Microsoft, Amazon, Alphabet, Meta, and Oracle, have committed to spending over $700 billion on AI data center infrastructure in 2026 alone. Estimates from Cathie Wood’s team project this number climbs to $1.4 trillion annually by 2030. Those are not aspirational figures. Contracts are signed. Ground is broken. Shovels are in the dirt. This is a replay of the cloud build-out from 2015 through 2020, but the dollars are bigger and the timeline is shorter. Training next-generation AI models requires an absurd amount of compute, and the companies building them are spending whatever it takes to get there first. Inference demand is scaling on top of that. The five names I am watching here each own a different piece of the infrastructure stack: NVIDIA remains the dominant player with roughly 90% GPU market share for AI workloads. Their CUDA (Compute Unified Device Architecture) software moat is real; most AI code runs on it, and switching costs are high. TSMC manufactures nearly all advanced AI chips, making it a picks-and-shovels position on the entire hardware boom. Broadcom is carving out territory in custom application-specific integrated circuits (ASICs) for hyperscalers. Micron is the only US-based manufacturer of HBM3E high-bandwidth memory, the specialized RAM that AI GPUs demand and cannot source fast enough. Vertiv rounds out the group as the infrastructure plumbing play; somebody has to cool and power these data centers. The math I am focused on: TSMC projects AI chip revenue will grow at a 60% compounded annual growth rate from 2024 through 2029. Nvidia’s fiscal 2027 revenue growth is forecast at 65%. These are the largest semiconductor companies on earth posting growth rates that belong to Series B startups. Nvidia’s forward earnings multiple has compressed from 60x to 25x as revenue caught up to expectations. When a company grows into its valuation like that, you are getting the growth cheaper than a year ago. Names to watch: , , , , 2. AI Moves from Chatbot to Coworker 2025 was the year of the AI chatbot. 2026 is when AI becomes an agent; something that does not just answer questions but does work. IBM’s VP of Quantum and AI, Ismael Faro, put it well when he described a shift in software practice from “vibe coding” to an objective-validation protocol where users define goals, agents execute tasks, and humans approve at checkpoints. Four companies are leading the shift from chatbot to coworker: Microsoft is betting big with Copilot built into the entire Office and Azure system. Their Azure AI revenue is growing at triple-digit rates. Palantir’s Artificial Intelligence Platform (AIP) targets government and defense contracts where AI agents process classified data and coordinate operations. ServiceNow is building AI agents into IT workflow automation, stacking new AI-driven products with premium pricing on top of its existing platform. Datadog is layering AI-driven observability on top of cloud infrastructure monitoring, with 83% of customers now using two or more products. The signal I care about here is enterprise adoption velocity. Datadog’s land-and-expand model, with 52% of customers using four or more products, is the same playbook that made Salesforce a trillion-dollar company. More AI agents means more infrastructure to monitor, which means more Datadog revenue. It is a self-reinforcing loop. Where most people get this wrong is assuming agentic AI is speculative. Large enterprises are already deploying AI agents for customer service, code generation, and IT operations. The question is not whether agentic AI will generate revenue. It is which companies will capture the most of that spend; the ones embedded in enterprise workflows with high switching costs. Names to watch: , , , 3. Cybersecurity Becomes a $250 Billion Arms Race Here is a pattern I keep seeing in the data: every dollar spent on AI infrastructure creates a matching need for AI security. More AI agents operating autonomously means a larger attack surface. More sensitive data flowing through AI pipelines means higher-value targets. The attacks themselves are getting faster. Modern breaches happen in milliseconds; defenses need to operate at the same speed. CrowdStrike’s Falcon platform processes trillions of security events per week using AI-driven threat detection. Palo Alto Networks is consolidating dozens of point security tools into a single integrated platform, reducing vendor fatigue and capturing a bigger share of enterprise budgets. Wedbush’s Dan Ives flagged both as top picks for 2026. The trend underneath all of this is zero-trust architecture. The old model of “protect the perimeter” is dead. Every user, device, and data request needs continuous verification. As quantum computers get closer to cracking encryption standards like RSA and ECC, companies are beginning the migration to quantum-resistant algorithms. NIST finalized its first post-quantum cryptography standards in 2024. Every government agency, financial institution, and healthcare system will need to make this switch, and CrowdStrike and Palo Alto Networks are both putting money into post-quantum development. Names to watch: , 4. Autonomous Systems Go Mainstream Self-driving vehicles are no longer a science project. The Guardian reported in late 2025 that we are on the verge of global robotaxi deployment, with firms from both the US and China launching in cities worldwide throughout 2026. Tesla plans to begin supervised robotaxi operations in Austin this year, and Waymo is expanding across US metro areas. But autonomous does not just mean cars. AI-powered robots are becoming standard in manufacturing, logistics, and warehousing. Arm’s CES 2026 preview showed robots moving beyond lab environments into the real world. Tesla’s Optimus humanoid program is getting real money and serious engineering talent. Whether it produces revenue in 2026 is debatable, but the R&D spend tells you where management sees the long-term opportunity. For those building a portfolio, Tesla is the most direct autonomous bet, but it comes with Elon Musk risk and a valuation that prices in execution across cars, energy, robots, and autonomous driving simultaneously. NVIDIA supplies the compute through its DRIVE platform, and the whole system runs on TSMC-manufactured chips. Morgan Stanley’s Adam Jonas models a robotaxi-only scenario that values Tesla above $500 per share. But if regulatory delays push the timeline out another 18 months, the current price already bakes in a lot of optimism. The expected value here depends on the probability you assign to successful execution. Names to watch: , , 5. Cloud 3.0 and the AI Sovereignty Push Capgemini calls it Cloud 3.0, a move away from the “all-in on one hyperscaler” model toward hybrid, multi-cloud, and sovereign architectures. The reason is AI. Running AI workloads requires different infrastructure than traditional cloud computing. Companies are spreading their bets across AWS, Azure, Google Cloud, and Oracle Cloud Infrastructure (OCI) based on which platform runs their AI models most cost-effectively. The four hyperscalers are each carving out distinct advantages: Oracle has been the biggest surprise. OCI landed massive contracts with OpenAI, Meta, and xAI for AI training workloads, positioning itself as a legitimate fourth hyperscaler. AWS remains the market share leader with Bedrock driving AI adoption across enterprise customers. Azure’s OpenAI partnership gives it a distribution advantage hard to match. Google Cloud is using Gemini and its Willow quantum chip to separate from the pack. The sovereignty angle is new. Governments worldwide are demanding that AI infrastructure stay within national borders. The European Union’s AI Act is forcing compliance changes; India and Brazil are building their own frameworks. The companies that can spin up sovereign cloud regions fastest will capture government contracts their competitors cannot touch. Names to watch: , , , 6. Quantum Computing: Still Early, but the Clock Is Ticking I include quantum here not because it will generate revenue for your portfolio in 2026. It almost certainly will not in any direct way. But the companies putting money into it now are the same ones that will dominate when it does matter. The timeline is compressing. Microsoft’s Jason Zander said we are in a “years, not decades” era for quantum advantage. Google’s Willow chip showed a 30% improvement in error correction, one of the hardest problems in making quantum computing reliable. The practical near-term angle is hybrid computing: quantum processors working alongside classical AI and supercomputers. IBM and Microsoft are both pushing this approach, and the applications in drug discovery, materials science, and financial modeling are real even if the revenue is still speculative. For portfolio exposure to quantum, you are buying Alphabet and Microsoft. Both have deep quantum research programs funded by their cloud revenue. No pure-play quantum names are worth the price at current valuations. The post-quantum cryptography opportunity, migrating encryption to quantum-resistant algorithms, is more actionable near-term, and that benefits cybersecurity players like CrowdStrike and Palo Alto Networks. Names to watch: , The 15 Must-Have US Names for 2026 Below is the full list. Every name on this list ties back to one or more of the six trends above. I have weighted the list toward infrastructure and semiconductors because that is where the most durable spend is concentrated, but I have included cybersecurity, software, and autonomous plays to cover the full opportunity set. Prices reflect market close on 2/18/2026. Large Cap 15 Mid-Cap Names With the Highest Probability of Outperformance The large-cap names above are the backbone. They are the ones with proven revenue, signed contracts, and market dominance. But if you are looking for higher expected value and are willing to accept more variance, the mid-cap tier is where the math gets more interesting. Mid-caps sit in a sweet spot. They are big enough to have real revenue, proven customers, and shipping products. But they are small enough that a single contract win or a breakout quarter can move the needle on the entire business in a way that no longer happens for a $3 trillion company. When I screen for mid-caps, I am looking for three things: Direct exposure to one of the six trends above. Accelerating revenue growth. A price that has not yet fully captured the opportunity. These 15 names are more volatile than the large-cap list. Some of them are down 20 to 40% from their highs. That is the point. If the thesis is right, you are buying at a discount to where these companies are headed. If the thesis is wrong, the position sizing math matters even more. Size small, spread wide, and let the probabilities play out. Mid Cap The Mid-Cap Thesis: Where the Math Works Marvell and Ciena are the networking plays. The AI buildout needs more than GPUs; it needs the silicon and optical equipment that connects them. Marvell’s Q3 fiscal 2026 report showed 38% data center growth, and Morgan Stanley keeps an Overweight rating. Ciena’s data center business is doubling year-over-year with a $1 trillion addressable market by 2028. CoreWeave is the closest thing to a pure-play AI cloud company, with revenue going from near zero in 2022 to an expected $10 billion-plus in 2026. The risk is concentration; Microsoft accounted for 62% of 2024 revenue. Astera Labs makes PCIe (Peripheral Component Interconnect Express) and CXL (Compute Express Link) connectivity chips inside AI servers. As GPU clusters get bigger, the connectivity bottleneck becomes the binding constraint, and Astera sits right at that chokepoint. Morgan Stanley named it their top small-cap semiconductor pick for 2026. On security, Zscaler is the established zero-trust leader with roughly 100% upside to Wedbush’s $350 target. SentinelOne is riskier but the reward scales higher; their Purple AI platform does autonomous threat response, and the name sits 89% below analyst consensus. IonQ is the speculative quantum position. I would not put more than 1 to 2% of a portfolio here, but if quantum computing hits a commercial inflection point within 3 to 5 years, the current valuation looks like a bargain. Size small and let expected value math work. Lumen and Super Micro are the value plays. Both are beaten down with direct ties to AI infrastructure; Lumen’s fiber network connects data centers, and $SMCI builds GPU-optimized server racks with liquid cooling. UiPath rounds out the list as the agentic AI automation play at the lowest price point. If agentic AI adoption accelerates the way IBM and Microsoft project, UiPath’s enterprise installed base becomes very valuable. Mid-cap names to watch: , , , , , , , , , , , , , , The Edge AI Opportunity Nobody Is Pricing In Edge AI is the trend I think is most underpriced. It means running AI models on devices instead of sending data to the cloud. The compute happens locally, and the advantages stack up: Lower latency for real-time applications like autonomous driving and factory robotics. Better privacy because sensitive data never leaves the device. Reduced bandwidth costs as billions of devices generate AI workloads. Apple is the biggest consumer play here. Apple Intelligence runs on-device by default, and older iPhones cannot support it. When over a billion active iPhones need a hardware refresh to access AI features, the revenue implications are enormous. Meta fits here too; Llama is the most widely adopted open-source AI model family, and that distribution feeds data and partnerships back to Meta while competitors pay for OpenAI or Anthropic access. The semiconductor angle is specialized neural processing units (NPUs) and tiny machine learning (TinyML) chips. AMD and several Arm-based chipmakers are building AI inference hardware for low-power devices. This is the next wave of chip demand after the data center GPU boom matures. Billions of devices versus millions of servers. A Word on Risk and Position Sizing If you have read my piece on the math of probabilities, you know I think about risk in terms of expected value and position sizing, not gut feelings. The same math applies here. Concentration in tech, even in great companies, means you are correlated to a single sector’s performance. If AI spending disappoints or a major regulatory crackdown hits, the whole basket falls together. The probability distribution of outcomes for these 30 names is skewed positive. The expected value of holding them over a 2 to 3 year horizon is likely above zero for most of you, but the variance is high on the mid-cap side. A 20 to 30% drawdown in any individual name during a risk-off period is well within normal distribution. If you cannot stomach that, you are sized too large. My approach: spread the exposure across the six trend categories so no single thesis sinking wipes out your returns. The infrastructure names ($NVDA, $TSM, $AVGO, $MU, $VRT) carry the most near-term conviction because their revenue growth is backed by signed contracts and committed capex. The software and cybersecurity names ($MSFT, $PLTR, $CRWD, $PANW, $NOW, $DDOG) are more sensitive to enterprise spending cycles but have stronger recurring revenue. Tesla and Oracle are the highest-variance positions. Both could outperform or underperform by wide margins depending on execution. What the Data Is Telling Us When I pull all the data together, the picture is this: 2026 is not the year AI hype peaks. It is the year AI spending becomes structurally embedded in corporate budgets the same way cloud spending became entrenched a decade ago. Gartner estimates worldwide IT spending will hit $5.5 trillion in 2026, with AI-related spend accounting for $1.5 trillion. Capgemini calls it the “Year of Truth” for AI, when organizations shift from experimentation to measuring real ROI. The companies that perform best will not be the ones with the best marketing or the biggest hype cycles. They will be the ones with real revenue tied to real workloads, where you can trace a dollar of AI spend from a customer’s budget into the company’s income statement. Bottom Line The six trends I have laid out here are running on different timelines but feeding into the same structural shift: AI infrastructure, agentic AI, cybersecurity, autonomous systems, Cloud 3.0, and quantum computing. Technology is moving from a cost center to a revenue driver for nearly every industry on the planet. The companies building the infrastructure, securing the data, and deploying the intelligence are the ones worth owning. These 30 names are not the only ones that will work in 2026, but they represent the strongest intersection of trend exposure, revenue quality, and competitive positioning I am finding in the data right now. The 15 large-caps give you the stable core. The 15 mid-caps give you the upside potential. Together, they cover the full probability distribution of outcomes across all six trends. Size your positions according to your risk tolerance, spread across the categories, and let the math play out over time. The edge, just like in my approach to the markets, comes from probability, patience, and discipline. Share Keep watching the signals, not the noise. Likes and shares helps the publication. 🙏 Disclaimer: This article is for informational purposes only and does not constitute investment recommendations or advice. Always do your own research and consider your financial situation before making decisions. Remember what goes up must come down (eventually) Stay safe and act wisely and this is in no mean financial advice. [ Full Disclaimer ] Thank you for supporting this newsletter. It will keep improving. Harry
View original →#37 Weekly Update - FOMC ahead - A Look at 11 Top Names
S&P 500, Tesla, Bitcoin, Nvidia, Palantir, Amd, Google, Amazon, Apple, Meta, Microsoft, Salesforce S&P 500, Tesla, Bitcoin, Nvidia, Palantir, Amd, Google, Amazon, Apple, Meta, Microsoft, Salesforce Hi and welcome back for a Quant data driven analysis. [Full Disclaimer]
View original →AI - Bubble - Revolution - Both? [Guest Post]
OpenAI is seeking a $500B valuation Hi and welcome back for a guest analysis . [ Full Disclaimer ] Thesis Overview Artificial Intelligence has defined the decade’s hottest trade. Since ChatGPT’s viral breakout in 2023 and Nvidia’s historic guidance, trillions have been minted—carrying Microsoft, Alphabet, Meta, Palantir, and infrastructure upstarts like Nebius and CoreWeave into investor spotlights. But history is whispering warnings. In 1999, Cisco was briefly worth half a trillion dollars before collapsing 80% in the dot-com bust. Today, OpenAI is reportedly seeking a $500B valuation with barely any revenue, while Palantir trades at 210x forward earnings. The bubble parallels aren’t hard to see. Meanwhile, the AI revolution is devouring everything in its path . AI startups absorbed more than 70% of all U.S. venture capital in Q1 2025, starving sectors like biotech and fintech of funding. Data center construction is now surpassing office buildings, with local governments already blocking $60B+ worth of projects over water and energy usage. Goldman Sachs projects AI-driven data centers could drive U.S. electricity demand up 165% by 2030 , while water use may top 1.7 trillion gallons annually by 2027 —enough to fill 2.6 million Olympic swimming pools. The question for investors is simple but existential: Will AI deliver enough productivity to justify this enormous resource drain, or are we marching toward another speculative bubble with painful consequences? But first, a little bit about me, The Pragmatic Investor I call myself The Pragmatic Investor because my approach is built on one thing: what works in practice. Markets are noisy, strategies are endless, and theories are only as good as their results. Over years of research and investing, I’ve distilled my philosophy into what I call the Pragmatic Investing Pyramid — a three-pronged framework of Macro, Fundamentals, and Technicals . Macro gives the big picture: cycles, policy, and geopolitics that drive long-term wealth. Fundamentals keep us grounded in the reality of business models and balance sheets. And Technicals provide the timing edge, with robust tools (and even a proprietary algorithm I designed) that uncover high-probability setups. My Substack is about making this approach accessible. Each week, I deliver a clear macro update, technical analysis of key indexes, and deep-dive fundamentals on one stock — all aimed at helping you build a resilient portfolio and boost returns with actionable trade ideas. Check Out My Substack The Bull Case: AI as the Next Internet For bulls, AI isn’t hype — it’s history repeating. Just as the internet seemed overhyped in the late ‘90s yet went on to transform every industry, AI may be in its 1997 moment, not its 2000 collapse . Massive TAM: McKinsey estimates AI could add $15 trillion to global GDP by 2030. Enterprise AI spending could top $1.3 trillion annually by then. This is a market-size unconstrained opportunity, comparable to cloud in the 2010s. Real Monetization: We’re already seeing payoffs. Meta’s AI-driven Advantage+ ad tools helped lift Q2 revenue 22%. Microsoft’s Copilot is embedded across Office, with enterprise adoption scaling fast. Nvidia is on track to post nearly $46B in quarterly revenue. Neoclouds like Nebius and CoreWeave are signing gigawatt-scale contracts with hyperscalers and AI labs. Capex Supercycle: Big Tech will spend more than $200 billion in 2025 on AI infrastructure. That level of sustained investment all but ensures demand for chips, servers, and energy providers. Adoption Curve: According to MIT, 78% of organizations already embed AI in their workflows — a far faster adoption rate than the internet or smartphones. Early use-cases (automated content, code generation, customer service) are already delivering measurable returns. Bulls argue the shakeout risk is overstated. Just as Amazon, Google, and Apple emerged from the dot-com rubble, today’s leaders — Microsoft, Nvidia, Meta — will be the durable winners of the AI era. The Bear Case: Warning Lights Everywhere Bears don’t deny AI’s potential, but they argue investors are paying tomorrow’s prices today — while the economics are still shaky. MIT’s “GenAI Divide”: 95% of corporate AI projects are not profitable, and only 5% show measurable impact on P&L. Integration is messy, expensive, and often fails to scale. Bubble Valuations: Palantir trades at 210x forward earnings. Nvidia’s multiple still implies decades of flawless execution. OpenAI’s $500B whisper valuation is absurd for a company that has yet to prove durable profitability. These echo the 1999 froth, when Pets.com and Webvan commanded billions before imploding. Execution Risk: The “killer apps” of AI — autonomous agents and robotics — aren’t ready. Agentic AI can’t yet reliably replace office workers. Robotics breakthroughs needed for factories, logistics, and construction remain years away. Until then, productivity gains are incremental, not transformative. Societal and Environmental Backlash: Communities are blocking new data centers over water and power use. Tucson recently rejected a massive Amazon facility and imposed strict water-use rules. Data centers could soon consume as much electricity as entire mid-sized countries. If the narrative shifts from “AI jobs engine” to “AI kills jobs and guzzles water,” regulators will tighten the screws. Jobs Risk: If AI does deliver massive productivity, it may do so by hollowing out mid-level white-collar work — analysts, programmers, assistants, even sales reps. With consumer spending accounting for 70% of U.S. GDP, that could tip the economy into recession rather than prosperity. The bear case isn’t that AI fails — it’s that valuations assume flawless adoption in a world where friction, politics, and human behavior say otherwise. My Two Cents Both bulls and bears are right — but on different timeframes. In the short run, the market is pricing in perfection. MIT’s data shows most projects are duds. Sam Altman himself admits AI is in a bubble. And resource constraints — from electricity to venture capital — are real. A correction, even a brutal one, would not be surprising. But zoom out. Global AI spend is already $750B in 2025 and is set to grow fivefold by 2034 . This isn’t just hype. Tech giants, startups, and governments are deploying billions into real infrastructure and applications. AI is moving beyond flashy demos into production: hospitals are rolling out diagnostic models, factories are using predictive AI to cut downtime, and marketing teams are replacing manual workflows with LLMs. The internet bubble destroyed capital but left the pipes in place. Out of the rubble emerged trillion-dollar giants. AI feels similar. The 95% of failed pilots don’t negate the 5% that will reshape industries. I count myself among the bulls — but with caution. This revolution is real, scalable, and far from a fad. That doesn’t mean every stock survives. Investor Playbook How to navigate the hype cycle without blowing up your portfolio: Stick With Mega-Caps: Microsoft, Meta, and Google have real AI monetization already flowing through earnings. They can outspend rivals, weather downturns, and are unlikely to get disrupted. Semiconductor Exposure: Nvidia remains the backbone of AI training, but size carefully — it’s priced for perfection. AMD offers a cheaper inference-focused play. Infrastructure Moonshots: Nebius and CoreWeave are speculative bets. They could become the “next AWS” — or crash if capacity outpaces demand. Keep position sizes small. Avoid Bubble Multiples: Palantir at 210x earnings is uninvestable at these levels. Likewise, avoid smaller AI SaaS names trading at 40–50x sales. These will get crushed if sentiment turns. Macro Hedges: AI’s energy demand makes natural gas, uranium, and nuclear-linked utilities attractive. Gold and Bitcoin offer protection if the bubble pops and spills into broader markets. Final Thoughts The AI trade is at a crossroads. On one side, the bull case is undeniable: adoption is happening faster than the internet, capex is entrenched, and productivity gains are real. On the other hand, the warning signs are everywhere: failed projects, bubble valuations, and resource constraints that could spark backlash. History teaches that bubbles burst — but revolutions endure. Amazon survived 2000, Pets.com didn’t. Investors who sized exposure wisely and stuck with cash-generating leaders were rewarded. The same playbook applies now. Own the giants with proven monetization. Keep speculative infra bets small. Avoid chasing froth. Hedge against macro shocks. AI will reshape the world. The question isn’t whether it’s real — it’s who survives the shakeout. If you’ve found this analysis useful, I invite you to join me on The Pragmatic Investor Substack , where I track the latest AI breakthroughs, market shifts, and trade setups in real time. Every week, I cut through the noise with clear macro updates, technical setups, and deep dives into the companies shaping the AI revolution. Whether you’re looking to stay ahead of the bubble debate, find the next Nvidia, or simply manage risk with conviction, my newsletter delivers the insights and actionable ideas you need. Disclaimer: This article is for informational purposes only and does not constitute investment advice. Always do your own research and consider your financial situation before making investment decisions. Remember what goes up must come down (eventually) Stay safe and invest wisely and this is in no mean financial advice. [ Full Disclaimer ] Thank you for supporting this newsletter. It will keep improving. Harry
View original →The 750 Billion Cloud Titans - How AI Transformed "Megatrons"
Amazon, Microsoft, and Google's Infrastructure Battle Hi and welcome back for a Quant data driven analysis . [ Full Disclaimer ] Share The cloud infrastructure market hit a remarkable milestone in Q1 2025: $91 billion in quarterly revenue, up 22% year-over-year. Yet beneath these numbers lies a more profound transformation. What started as a race to rent server space has morphed into something far more complex—a battle for AI supremacy that's reshaping how these tech giants invest, innovate, and compete. I've tracked this market for years, watching AWS maintain its comfortable lead while Microsoft and Google played catch-up. But the AI boom changed everything. In just 18 months since ChatGPT's launch, we've witnessed the most aggressive infrastructure buildout in tech history, with combined capital expenditures from the big three cloud providers exceeding $250 billion in 2025 alone. Here's what makes this moment particularly fascinating: each company is betting on radically different strategies to win the AI-powered future of computing. Subscribe now The Market Reality Check Let me start with the numbers that matter. AWS still commands 30% market share, generating roughly $27 billion in quarterly revenue. Microsoft Azure sits at 21% with $19 billion, while Google Cloud holds 12% at $11 billion. The remaining 37% gets carved up among smaller players, though that percentage continues to shrink as the big three consolidate their grip. But market share tells only part of the story. Growth rates reveal where the momentum lies. Azure posted 34% year-over-year growth in Q1 2025, with AI services contributing 16 percentage points of that expansion. Google Cloud grew 32%, while AWS—despite its massive scale—managed a respectable 17%. The driver behind these divergent growth rates? AI workloads that consume 5-10x more compute resources than traditional applications. When a single large language model training run costs millions in compute time, cloud providers discovered they're no longer selling commodity infrastructure—they're selling the picks and shovels for the AI gold rush. The Infrastructure Arms Race This realization triggered an infrastructure spending spree that would have seemed absurd just two years ago. Amazon allocated $100 billion for data centers in 2025. Microsoft committed $80 billion. Google pledged $75 billion. To put this in perspective, these three companies will spend more on infrastructure this year than the entire GDP of many small countries. But raw spending only scratches the surface. The real story lies in how they're spending it. The Nuclear Option Perhaps nothing illustrates the scale of ambition better than the rush to secure nuclear power. Data centers already consume about 2% of global electricity, and AI threatens to double or triple that figure. Renewable sources like solar and wind, while important, can't provide the consistent baseload power that AI training demands. Microsoft made the boldest move, signing a 20-year deal to restart Three Mile Island's Unit 1, securing 835 megawatts of carbon-free power. The symbolism is striking—a nuclear plant that became synonymous with disaster in 1979 will power the AI revolution. Amazon took a different approach, investing $500 million in X-energy to develop small modular reactors (SMRs) while purchasing a nuclear-powered data center campus from Talen Energy. Their bet: next-generation reactor designs will provide more flexible, scalable power solutions. Google, characteristically, hedged its bets. Beyond partnerships with Kairos Power for SMRs, they're investing in geothermal and solar projects. But their recent deal with Elementl Power to develop three new nuclear sites signals they understand the physics problem: you can't train trillion-parameter models on intermittent power. Custom Silicon: The Real Differentiator While nuclear partnerships grabbed headlines, the more strategic battle unfolds in silicon. Each cloud giant recognized that Nvidia's GPU dominance creates both a bottleneck and a margin problem. Their solution? Build their own chips optimized for their specific workloads. Amazon's Trainium2 represents their most ambitious effort yet. Built on a 5nm process with HBM3 memory, these chips deliver 30-40% better price-performance than GPUs for training workloads. More importantly, they're designed specifically for the kinds of models AWS customers deploy, from recommendation engines to large language models. Google's TPU v6 takes a different approach, emphasizing raw performance over cost. With twice the performance per dollar of TPU v4, Google positions these chips as the premium option for cutting-edge research. Their recent disclosure that they trained a 128-billion parameter model on 50,944 TPU v5e chips in under 12 minutes showcases this raw computational power. Microsoft arrived late to the custom silicon party with Maia, announced in November 2023. But they're moving fast, integrating these chips specifically for Azure OpenAI workloads. It's a targeted strategy—rather than competing across all AI workloads, they're optimizing for their partnership with OpenAI. What's telling is what these chips represent strategically. They're not just about cost savings or performance gains. They're about control—over supply chains, over optimization paths, and ultimately over the kinds of AI applications that get built on each platform. The Agentic Revolution Infrastructure and chips create the foundation, but the real competition now centers on agentic AI—systems that can autonomously perform complex tasks with minimal human oversight. This isn't just another feature; it's a fundamental shift in how software works. Microsoft leads in enterprise adoption, with over 230,000 organizations using Copilot Studio to build AI agents. Their strategy leverages existing enterprise relationships, embedding agents into familiar tools like Office 365 and Teams. When a law firm can create an agent that drafts documents in their house style, or when Wells Fargo builds an agent that helps employees navigate 1,700 internal procedures, the value proposition becomes concrete. Amazon's approach with Bedrock Agents emphasizes flexibility and infrastructure. They're not prescribing how agents should work but providing the tools for developers to build whatever they imagine. Their recent launch of multi-agent collaboration—where specialized agents work together under a supervisor agent—shows they're thinking about complex, real-world workflows. Google's Vertex AI Agent Builder might be the most technically ambitious. Their Agent Development Kit (ADK) promises multi-agent systems in under 100 lines of code, with support for their Agent2Agent protocol enabling cross-platform collaboration. They're betting that open standards will attract developers who don't want vendor lock-in. But here's what's remarkable: we're still in the earliest stages. These agents remain relatively simple, handling discrete tasks like customer service queries or document processing. The real revolution comes when agents start collaborating across organizations, when they develop persistent memory and learning capabilities, when they move from reactive to truly proactive. Security as the New Battlefield Nothing crystallized the importance of cloud security quite like Google's $32 billion acquisition of Wiz—the largest cybersecurity deal in history. At 50x revenue, the price seemed astronomical. Until you understand what Wiz represents. Cloud security differs fundamentally from traditional cybersecurity. It's not about protecting a perimeter; it's about securing a constantly shifting landscape of containers, serverless functions, and interconnected services. Wiz cracked the code with an agentless approach that provides visibility across multi-cloud environments without the overhead of traditional security tools. The acquisition serves multiple strategic purposes for Google. First, it instantly establishes them as a serious player in enterprise cloud security, competing directly with Microsoft's $15 billion security business. Second, it addresses the multi-cloud reality—enterprises use an average of 2.7 cloud providers, and they need security that works across all of them. Microsoft responded by extending Defender's AI security posture management to cover Google Cloud and AWS. They're also rolling out specialized security agents—from phishing triage to vulnerability remediation—that leverage their Copilot infrastructure. Amazon, notably, has taken a partnership approach, working with companies like CrowdStrike and Bitdefender rather than building everything in-house. It's consistent with their philosophy of providing the platform while letting others build specialized solutions. What's emerging is security as a key differentiator in enterprise cloud decisions. When a ransomware attack can cost millions and AI systems introduce new attack vectors, security moves from IT concern to board-level priority. The Market Dynamics Ahead Looking forward, three trends will define the next phase of this competition: Consolidation Accelerates : The $32 billion Wiz acquisition won't be the last mega-deal. Smaller cloud and AI companies face a stark choice: get acquired or get left behind. The scale requirements for competing in AI—from infrastructure to talent—create natural consolidation pressure. Multi-Cloud Becomes the Norm : Despite each provider's best efforts to lock in customers, enterprises increasingly demand flexibility. Tools and services that work across clouds will command premium valuations. This explains why Wiz, despite being acquired by Google, will maintain support for AWS and Azure. AI Workloads Reshape Economics : Traditional cloud metrics like storage costs or compute hours matter less when AI workloads dominate. What matters is time-to-insight, model performance, and increasingly, energy efficiency. Providers that optimize for these new metrics will capture disproportionate value. Investment Implications For investors, this transformation creates both opportunities and risks. The obvious play—betting on the cloud giants themselves—faces headwinds from massive capital requirements and regulatory scrutiny. The more interesting opportunities lie in the ecosystem: Energy Infrastructure : Companies facilitating the nuclear renaissance for data centers, from SMR developers to grid modernization plays, benefit from multi-decade tailwinds. Specialized AI Chips : While the giants build their own silicon, specialized chip designers targeting specific AI workloads can still capture significant value. Multi-Cloud Tools : Security, observability, and management tools that work across clouds become increasingly valuable as enterprises avoid vendor lock-in. Vertical AI Applications : The infrastructure battle creates the foundation, but the real value creation will come from AI applications that transform specific industries. The Deeper Game Step back from the quarterly earnings and strategic moves, and a deeper pattern emerges. This isn't really about cloud infrastructure anymore. It's about who controls the means of intelligence production in the 21st century. Amazon, Microsoft, and Google aren't just building data centers; they're constructing the neural substrate for artificial intelligence at planetary scale. The winner won't be determined by who has the most servers or the fastest chips. It will be determined by who creates the platform where the next generation of AI breakthroughs happen. In that context, $250 billion in annual infrastructure spending doesn't seem excessive. It seems like table stakes for the biggest technological transformation since the internet itself. The cloud titans understand this. The question is whether the rest of us do. Keep watching the signals, not the noise. Share Disclaimer: This article is for informational purposes only and does not constitute investment advice. Always do your own research and consider your financial situation before making investment decisions. Remember what goes up must come down (eventually) Stay safe and invest wisely and this is in no mean financial advice. [ Full Disclaimer ] Thank you for supporting this newsletter. It will keep improving. Harry
View original →- 15% AI & Magnificent 7 - Next Decade Winner(s) - Going Lower in Q2?
Microsoft leads AI readiness via OpenAI, Azure and Office distribution; buy lower over time.
View original →