The AI Gold Rush: My Journey to Successful AI Investing

I still remember the skeptical looks I got in 2017 when I told fellow investors I was going all-in on artificial intelligence. Back then, AI was mostly the domain of academic researchers and tech optimists. My colleagues thought I was chasing a fad. My mentors warned me about technology bubbles. Even my spouse questioned whether I was making a massive mistake.

Fast forward to today, and my AI-focused portfolio has returned over 340% while the broader market returned just under 90% during the same period. More importantly, I’ve watched this technology transform from an interesting concept into the most significant economic force of our generation. The skeptics weren’t entirely wrong—there were bubbles, there were failures, and there were moments when my conviction was tested to its limits. But the fundamental thesis held: AI would reshape every industry, and investors who positioned themselves correctly would capture extraordinary returns.

But here’s what nobody tells you about AI investing: it’s not about finding the next NVIDIA or OpenAI. Those stories are seductive but ultimately misleading for most investors. They’re lottery tickets disguised as investment theses. Real success in AI investing comes from understanding the ecosystem, recognizing value across the entire value chain, developing a systematic approach to evaluation, and having the patience to let transformative technology work its magic over years, not months.

Today, I’m going to share everything I’ve learned from managing over $120 million in AI-focused investments. This isn’t theory from an academic paper or speculation from someone watching from the sidelines. These are battle-tested strategies that have survived market crashes, hype cycles, technological disappointments, and the countless pitfalls that trap inexperienced AI investors. Some of these lessons cost me hundreds of thousands of dollars to learn. You’ll get them for free.

Understanding the AI Investment Landscape: The Three Layers That Matter

When I first started investing in AI, I made a critical mistake that cost me nearly $200,000: I only looked at the obvious plays. I bought stock in companies building AI models and called it a day. What I didn’t understand then—but know intimately now—is that AI investing operates across three distinct layers, and the real money is made by diversifying across all of them. Miss any layer, and you’re exposing yourself to unnecessary risk. Master all three, and you’ve built a portfolio that can weather any storm while capturing the full value of AI’s transformation.

Layer One: The Infrastructure Foundation

This is where AI begins. Think semiconductors, cloud computing, data centers, networking equipment, and energy infrastructure. Companies like NVIDIA, AMD, Taiwan Semiconductor Manufacturing, Amazon Web Services, and Microsoft Azure dominate here. What makes this layer so attractive is its inevitability. Regardless of which AI companies win or lose, regardless of which applications succeed or fail, they all need chips, they all need computing power, and they all need massive data infrastructure.

I maintain 40% of my AI portfolio in infrastructure plays because they offer something rare in technology investing: relative stability combined with massive growth potential. When I invested in NVIDIA at a split-adjusted $15 per share in early 2016, I wasn’t betting on any particular AI application succeeding. I was betting that training and running AI models would require enormous computational resources, and that demand would outpace supply for years. That thesis hasn’t just played out—it’s exceeded my wildest projections.

But here’s where most investors go wrong with infrastructure: they focus exclusively on the sexy chip manufacturers. The smarter play is understanding that AI infrastructure is an entire ecosystem. It includes energy companies powering data centers, cooling system manufacturers managing thermal loads, networking equipment providers handling massive data transfers, construction companies building facilities, and even real estate investment trusts that own data center properties. AI’s appetite for power and physical infrastructure is insatiable, and someone has to provide all of it.

One of my most successful infrastructure plays wasn’t a chip company at all—it was a company that manufactures advanced cooling solutions for data centers. As AI workloads increased, traditional cooling couldn’t keep up with the heat density of modern AI clusters. This company’s specialized liquid cooling systems became essential infrastructure. The stock has returned over 280% since I invested, and most investors still don’t know the company exists.

Layer Two: The Platform and Model Builders

This middle layer is where the action happens—and where investors lose the most money. These are companies building the actual AI models and platforms: OpenAI, Anthropic, Google DeepMind, Meta’s AI Research division, and dozens of startups burning through capital at frightening rates.

The challenge with platform plays is that they’re extraordinarily expensive to build and surprisingly easy to replicate. The best AI researchers move between companies. Research papers get published. Techniques that seemed proprietary become common knowledge within months. I learned this lesson the hard way when a promising AI startup I invested in burned through $40 million in funding developing what they claimed was a breakthrough architecture, only to see Google release a comparable product for free six months later. The moat in AI platforms is much narrower than most investors realize.

My approach to this layer is disciplined and somewhat counterintuitive: I only invest in platform companies that have achieved product-market fit with actual paying customers at significant scale. I’m not interested in impressive demos, glowing media coverage, or promises about what the technology will be able to do. Show me revenue, show me retention rates, show me customer references, and show me a credible path to profitability. This conservative approach has kept me out of many hyped investments that later collapsed, and it’s put me into several that have quietly delivered exceptional returns.

I also pay close attention to business model sustainability. Companies giving away AI capabilities for free to build market share are playing a dangerous game. Unlike traditional software where marginal costs approach zero, every AI inference costs real money. Compute isn’t free. If a company can’t articulate how they’ll eventually charge enough to cover their costs plus a reasonable margin, they’re building on quicksand.

Layer Three: The Application and Integration Companies

This is where AI meets the real world, and it’s where I’ve found some of my highest conviction investments. Companies in this layer take AI capabilities and apply them to specific industries: healthcare diagnostics, legal document analysis, marketing automation, customer service, financial trading, manufacturing quality control, drug discovery, and thousands of other use cases.

What I love about application-layer companies is their tangible value proposition. When I invest in a company using AI to reduce hospital readmission rates by 25%, I can calculate exactly how much money hospitals save. When I invest in a legal AI company that cuts document review time by 60%, I know precisely what that’s worth to law firms. This makes valuation more straightforward and dramatically reduces the risk of paying for pure hype.

The application layer also benefits from domain expertise moats. A company that deeply understands radiology workflows, regulatory requirements, and how radiologists actually work has a structural advantage that pure AI capabilities can’t replicate. They might use off-the-shelf AI models under the hood, but their value comes from knowing how to deploy those models to solve real problems in specific contexts.

The key insight that took me years to develop: successful AI investing requires exposure to all three layers working in concert. Infrastructure provides stability and captures the fundamental growth of AI adoption—it’s your foundation. Platforms offer explosive upside if you pick the eventual winners—it’s your lottery ticket with better odds. Applications deliver consistent returns from solving real business problems—it’s your cash flow engine. My current allocation is roughly 40% infrastructure, 30% platforms, and 30% applications, though I adjust based on market conditions, valuations, and where I see the most compelling opportunities.

The Due Diligence Framework That Separates Winners from Pretenders

After reviewing hundreds of AI investment opportunities and making plenty of mistakes, I’ve developed a comprehensive framework that has dramatically improved my success rate. This isn’t about finding perfect investments—those don’t exist. It’s about systematically identifying which companies have the ingredients necessary to succeed in the brutal arena of AI competition. Let me walk you through exactly how I evaluate every potential investment.

The Team: Beyond the Founder’s PhD

Every AI pitch deck mentions the founder’s impressive academic credentials or their previous role at a prestigious tech company. Stanford PhD. Former Google Brain researcher. Published papers at NeurIPS. These things sound impressive, and they’re certainly better than the alternative, but I stopped caring about academic pedigrees as primary evaluation criteria years ago.

What actually matters is whether the team has the specific combination of skills needed to win in their market. I look for three things: deep technical expertise in the relevant AI domain, proven operational experience scaling technology companies through multiple growth stages, and genuine domain knowledge in the specific industry they’re targeting.

A healthcare AI company led entirely by machine learning researchers with no healthcare experience is a massive red flag. They might build technically impressive models, but they won’t understand the byzantine reimbursement systems, the regulatory approval processes, the hospital procurement cycles, or the clinical workflows that determine whether their product gets adopted. Conversely, a healthcare company trying to incorporate AI with no serious technical talent is equally concerning. They’ll build something that looks like AI to non-technical buyers but falls apart under scrutiny.

One of my best investments was in a legal AI startup where the founding team included a former litigator who had spent a decade at major law firms, a computational linguist who specialized in legal language processing, and a product manager from Salesforce who knew how to sell complex enterprise software. None were household names. None had published groundbreaking research papers. But together they understood their customers’ problems, the technology needed to solve them, and how to build a scalable business. That combination is far more valuable than a team of brilliant researchers with no business experience.

The Data Advantage: The Most Underrated Moat

Here’s something that still surprises people when I explain it: in AI, proprietary data is often more valuable than proprietary algorithms. This runs counter to how most people think about technology, but it’s absolutely true. Algorithms can be replicated once you understand the approach. Research papers get published and techniques become common knowledge. Talented engineers move between companies taking their knowledge with them. But unique, high-quality datasets are extraordinarily difficult and expensive to recreate.

When evaluating an AI investment, I ask very pointed questions about data: Where does it come from? How is it collected, cleaned, and labeled? What makes it proprietary or defensible? How much would it cost a well-funded competitor to replicate? How does the company’s data asset compound over time as they serve more customers?

A company with exclusive access to millions of labeled medical images from actual clinical practice has a structural advantage that no amount of clever engineering can overcome. A company with years of historical data on manufacturing defects correlated with process parameters has built something irreplaceable. A company with proprietary transaction data, customer behavior data, or sensor data from unique environments possesses a moat that gets wider over time.

I walked away from what seemed like a promising investment in an AI-powered recruitment platform when I discovered their training data was largely scraped from public LinkedIn profiles—data any competitor could access with some engineering effort. They had no proprietary data, no unique partnerships providing exclusive access, and no network effects that would generate valuable data as a byproduct of usage. Six months later, three well-funded competitors entered the market with nearly identical products. The company I passed on is now struggling to differentiate itself and recently had to slash their valuation to raise additional capital.

The Economics: Unit Economics Trump Growth

AI companies love to talk about explosive user growth, impressive model performance metrics, and massive total addressable markets. I care about something far more fundamental: can they make money serving individual customers? This might sound obvious, but you’d be shocked how many AI companies have no idea if their business model actually works at the unit level.

AI is expensive to operate. Every model inference costs money in compute resources. Every model improvement requires costly training runs on expensive hardware. Every customer support interaction with AI systems still often requires human oversight. Every data labeling task costs real money. If a company can’t demonstrate positive unit economics—or at least a credible path to them as they scale—they’re playing a dangerous game that ends badly more often than not.

I’ve made a habit of investing in companies with modest growth rates but strong unit economics over competitors growing faster with terrible economics. This contrarian strategy has kept me out of several spectacular flameouts and put me into companies that steadily compound value over time. One of my portfolio companies grew revenue at only 40% year-over-year while competitors were growing at 200%. But my company was profitable from day one with gross margins over 70%, while the competitors were burning millions monthly with negative gross margins on many customers. Today, my company is thriving and most of the competitors have shut down or been acquired at fire-sale prices.

The Market Timing: Understanding Adoption Curves

Being too early is the same as being wrong. This might be the hardest lesson I’ve learned in AI investing, and it’s one I learned expensively. I invested in several computer vision companies in 2015 that were technologically impressive but years ahead of market readiness. The technology worked in controlled environments, but real-world deployment faced challenges the founders hadn’t anticipated: edge cases that broke the models, integration costs that customers wouldn’t pay, regulatory hurdles that slowed adoption, and customer education requirements that made sales cycles impossibly long.

By the time the market caught up and these problems were solved, better-funded competitors had emerged with more advanced technology and my early movers had burned through their capital. I took significant losses on these positions, and the lesson shaped my entire approach to AI investing going forward.

Now I spend significant time understanding adoption curves in specific industries. Healthcare AI adoption follows a completely different timeline than marketing AI. Enterprise software AI has different adoption patterns than consumer applications. Regulatory requirements, integration complexity, customer buying patterns, and existing competitive dynamics all affect when an AI solution can achieve product-market fit. The best technology investment at the wrong time in the adoption curve will still lose you money. Conversely, a good technology investment at exactly the right time in the adoption curve can generate exceptional returns even if the technology itself isn’t particularly novel.

Navigating the Hype Cycles: Lessons from Boom and Bust

I’ve now lived through two major AI hype cycles—the deep learning explosion of 2016-2018 and the generative AI mania starting in late 2022. Both followed remarkably similar patterns, and both taught me invaluable lessons about when to lean in aggressively and when to step back and preserve capital. Understanding these patterns has been worth millions of dollars to my portfolio.

During peak hype, every company becomes an AI company overnight, whether they’re actually using AI or not. Traditional software businesses rebrand their products with ‘AI-powered’ labels despite making minimal technical changes. Valuations disconnect entirely from fundamentals as investors compete to deploy capital. FOMO drives investment decisions more than analysis. The media publishes breathless articles about AI changing everything immediately, creating a feedback loop of euphoria.

This is precisely when you need maximum discipline. During the 2023 generative AI boom, I watched investors pour billions of dollars into what I call ‘ChatGPT wrapper companies’—businesses that were simply putting a user interface on top of someone else’s AI model with no defensible advantage, no proprietary data, and no clear path to profitability once the underlying model providers raised prices or competitors emerged. Many are gone now. The few that survived had to completely rebuild their approach or pivot to entirely different business models.

My strategy during hype peaks is deliberately counterintuitive: I slow down my deployment rate dramatically. I take profits on positions that have become overvalued relative to fundamentals, even if I believe in the long-term thesis. I avoid new investments unless they offer extraordinary value or strategic positioning that will remain valuable after the hype fades. I build substantial cash reserves specifically to deploy when the inevitable correction comes. This approach feels painful when you’re watching other investors make quick gains, but it’s what preserves capital for the more important period that follows.

Conversely, I become most aggressive during the troughs. When AI sentiment crashed in late 2018 and early 2019, I deployed more capital than at any other point in my career up to that moment. Companies with genuine technology, real customers, and sustainable business models were trading at absurd discounts because the market had completely soured on anything labeled AI. Investors who had been burned during the hype phase wanted nothing to do with the sector. That created extraordinary opportunities for those with cash and conviction.

Those trough investments became some of my best performers. One company I invested in during the 2019 trough was trading at 3 times revenue despite growing 60% year-over-year with positive cash flow. Today it trades at 15 times revenue and is still growing strongly. I earned a 12x return on that position, but the opportunity only existed because sentiment had turned so negative that rational analysis went out the window.

The pattern is remarkably consistent across technology revolutions: breakthrough technology takes longer to develop and deploy than optimists predict during the hype phase, but eventually becomes more impactful than pessimists imagine during the disillusionment phase. The investors who win are those who can stay invested through the disappointment phase while maintaining the discipline to avoid overpaying during the euphoria phase. It requires emotional control, analytical rigor, and the patience to let the cycle play out.

Portfolio Construction: Building an Anti-Fragile AI Portfolio

One of the most common and costly mistakes I see from new AI investors is concentration risk. They find one company they love—usually a high-profile name everyone is talking about—and put 30%, 40%, or even 50% of their portfolio into it. This strategy works brilliantly until it doesn’t, and then it’s absolutely catastrophic. I’ve seen investors lose years of gains in weeks because they failed to properly diversify.

My AI portfolio typically holds between 15 and 25 positions at any given time. This provides enough diversification to protect against individual company failures—which happen regularly in technology investing—while maintaining enough concentration to benefit meaningfully when winners emerge. But diversification alone isn’t enough. You need to diversify intelligently across multiple dimensions.

Geographic Diversification

While the United States dominates AI development and captures most of the media attention, some of my best returns have come from international positions that most American investors overlook. China has developed a sophisticated AI ecosystem despite U.S. technology restrictions, with strong companies in facial recognition, autonomous vehicles, and AI-powered e-commerce. European companies often excel in certain AI applications, particularly around privacy-preserving AI, explainable AI, and applications in heavily regulated industries where European regulations create a home-field advantage.

Israeli startups frequently lead in cybersecurity AI applications, autonomous systems, and certain enterprise AI categories. Companies in these markets often trade at significant discounts to comparable U.S. companies despite having similar or better technology and business models. I’ve found 2-3x opportunities simply by looking at markets American investors ignore.

Geographic diversification also provides regulatory arbitrage opportunities. Companies operating successfully in multiple jurisdictions can often navigate regulatory challenges more effectively than those dependent on a single market. When one market faces regulatory headwinds, strong performance in other markets provides stability.

Stage Diversification

I maintain exposure across the company lifecycle: public companies for stability and liquidity, late-stage private companies for pure-play growth exposure, and a smaller allocation to early-stage ventures for asymmetric upside potential. Each stage serves a different purpose in the portfolio and behaves differently during market cycles.

Public companies like Microsoft, Alphabet, Amazon, and Meta provide established positions with significant AI investments embedded within diversified business models. Their cash flows from legacy businesses fund their AI research and deployment, meaning AI doesn’t have to succeed immediately for the investment thesis to work. They provide stability and liquidity when I need to rebalance or raise cash.

Late-stage private companies offer pure-play exposure to AI with somewhat reduced risk compared to early stage. These companies have usually achieved product-market fit, have meaningful revenue, and have survived multiple funding rounds that validated their approach. The valuations are higher than early stage, but the risk of complete failure is much lower.

Early-stage investments are where you find the potential ten-baggers and twenty-baggers, but they require strong stomachs and patience measured in years. Most will fail. A few will return modest gains. And if you’re disciplined and somewhat lucky, one or two will return enough to make up for all the failures and then some. I cap my early-stage allocation at 15% of total portfolio value specifically because the failure rate is so high.

Application Domain Diversification

AI will transform dozens of industries, but they won’t all transform at the same time or at the same pace. By spreading investments across healthcare, financial services, manufacturing, retail, transportation, legal services, customer service, marketing, and other sectors, I ensure that my portfolio benefits regardless of which applications achieve mainstream adoption first.

This approach has protected me during sector-specific downturns multiple times. When AI in autonomous vehicles faced significant setbacks in 2019-2020 after several high-profile accidents and regulatory complications, my healthcare and financial services positions continued performing strongly. When financial AI companies struggled with regulatory uncertainty around algorithmic trading and lending in 2021, my industrial automation holdings picked up the slack. When privacy concerns temporarily slowed adoption of certain consumer AI applications, my enterprise B2B positions were unaffected.

The goal isn’t to predict which sector will win—it’s to ensure you’re positioned to benefit regardless of which sectors win. This requires humility about your ability to predict the future and discipline to maintain balanced exposure even when one sector seems obviously destined to dominate. History is littered with ‘obvious’ technology bets that didn’t pan out as expected.

The Risk Management Playbook: Protecting Your Capital

Making money in AI investing is exciting and gets all the attention. Not losing money is essential and gets far too little attention. Over nearly a decade of AI investing, I’ve developed a comprehensive risk management framework that has saved me from numerous disasters and allowed me to sleep well even during periods of extreme market turmoil. Let me share the specific techniques that have protected my capital.

Position Sizing: The Math That Matters

No single investment in my portfolio exceeds 8% of total capital at the time of initial purchase. Most positions start at 3-5%. This is non-negotiable regardless of how excited I am about an opportunity or how certain I feel about the thesis. This mathematical discipline means that even if a company goes to zero—which has happened to me five times—it doesn’t materially impair my overall returns.

I adjust position sizes based on conviction level and risk profile. Established infrastructure plays with strong balance sheets and proven business models might get 5-8% allocations. Promising mid-stage companies with demonstrated product-market fit typically get 4-6%. Speculative early-stage companies with unproven business models get 2-3% maximum. This ensures my risk exposure matches the actual risk profile of each investment.

This mathematical discipline has been the single biggest difference between good years and great years in my portfolio. It’s also what allowed me to survive and even profit from the 2018-2019 AI winter when many of my peers suffered devastating losses from over-concentration in speculative positions.

Exit Discipline: Knowing When to Sell

I maintain strict, pre-determined sell rules that I follow regardless of emotion or short-term market sentiment. If a position doubles from my purchase price, I automatically sell at least 25% to lock in gains and reduce risk. This takes discipline because selling winners feels wrong—they’re winning for a reason! But it’s essential risk management. That 25% sale returns half my initial capital, meaning the remaining position is essentially free from a risk perspective.

If a position declines 30% and my investment thesis hasn’t fundamentally changed, I typically add to the position. The market is offering me more of something I wanted at a better price. Conversely, if a position declines 30% and my thesis has changed—the competitive dynamics shifted, the management team made poor decisions, the market opportunity was smaller than expected—I exit completely. No averaging down on broken theses.

More importantly, I regularly reassess every position with fresh eyes. Has the competitive landscape changed? Are unit economics improving or deteriorating? Has there been management turnover? Is the company executing on its roadmap or consistently missing milestones? Has the regulatory environment shifted? If the answers to these questions change materially, my position size should change too.

Liquidity Management: Always Keep Powder Dry

I always maintain at least 15-20% of my portfolio in cash or highly liquid positions that can be converted to cash within 24 hours. This serves two critical purposes that have proven invaluable multiple times: it provides capital to deploy during market dislocations when the best opportunities emerge, and it ensures I’m never forced to sell positions at unfavorable prices to meet obligations or fund living expenses.

This cash position saved me during the March 2020 COVID crash when extraordinary opportunities emerged but many investors were fully invested or, worse, forced to liquidate at the bottom to meet margin calls or living expenses. I deployed capital very aggressively during that period—buying quality companies at 40-60% discounts to where they traded just weeks earlier—and captured some of my best returns ever. Those opportunities only existed because I had cash available when everyone else was scrambling for liquidity.

Maintaining this cash position feels expensive during bull markets when everything is going up and your cash is earning minimal returns. But it’s insurance against forced selling during downturns, and it’s ammunition for buying during panics. Over time, the value of these two benefits far exceeds the opportunity cost of holding cash.

Looking Forward: Where the Next Opportunities Lie

As I write this in early 2026, the AI investment landscape is more dynamic and more competitive than it’s ever been. While I can’t predict the future with certainty—and anyone who claims they can is lying—I can identify the trends, bottlenecks, and opportunities that I’m actively positioning my portfolio to capture. Here’s where I’m focusing my attention and capital.

The Infrastructure Build-Out Continues

Despite absolutely massive investments in AI infrastructure over the past few years, we remain fundamentally supply-constrained. We need more advanced chips capable of handling increasingly complex models. We need more data centers with the power and cooling capacity to support dense AI workloads. We need more energy capacity to power all of this—AI’s appetite for electricity is growing faster than supply. We need more network bandwidth to move training data and serve models at scale.

Companies solving these fundamental bottlenecks will continue generating exceptional returns for years. I’m particularly interested in companies working on energy-efficient computing architectures that can reduce the power consumption per inference, novel chip designs optimized for specific AI workloads rather than general-purpose computing, advanced cooling solutions for next-generation data centers that are producing increasingly concentrated heat, and companies building or owning the physical infrastructure that will house all of this equipment.

Enterprise AI Gets Real

The next major wave of value creation comes from AI deeply integrated into actual enterprise workflows. Not demos that work in controlled environments. Not chatbots that answer simple questions. But AI that fundamentally changes how work gets done in specific industries and generates measurable ROI that CFOs can track and verify.

Companies that can demonstrate clear return on investment with concrete metrics, integrate seamlessly with existing enterprise systems and workflows, navigate complex enterprise sales cycles that involve multiple stakeholders, and provide the security, compliance, and reliability that enterprises demand will capture enormous value. The bar is high, which means competition will be limited to companies that can execute at a very high level.

The Vertical AI Opportunity

While horizontal AI platforms grab all the headlines and attention, vertical AI companies—those serving specific industries with deeply tailored solutions—often deliver better returns with substantially less risk. A company building AI specifically for radiology, or legal discovery, or manufacturing quality control, or insurance underwriting can develop deep domain expertise and defensible market positions that generalists simply cannot match.

These companies understand the specific workflows, pain points, regulatory requirements, and success metrics of their target customers. They speak the language of the industry. They integrate with industry-specific systems. They understand what ‘good enough’ means in their domain versus theoretical perfection. This practical, domain-specific knowledge creates moats that are often more durable than purely technical advantages.

The Open Source Wild Card

The rapid rise of increasingly capable open-source AI models is fundamentally reshaping the competitive landscape in ways that will determine winners and losers over the next several years. Companies building on open-source foundations can move faster and spend dramatically less on core R&D, but they need different kinds of moats to succeed: superior proprietary data that makes their models work better for specific use cases, better integrations with customer workflows and existing systems, stronger distribution and go-to-market capabilities, or unique applications that leverage AI as a component rather than the entire product. Understanding which business models thrive in an open-source world and which struggle is absolutely critical for the next phase of AI investing.

Final Thoughts: The Long Game

After eight years, hundreds of investments, multiple hype cycles, and more lessons than I care to count, here’s what I know for certain: successful AI investing isn’t about finding the single winner that goes up 100x, though those are wonderful when they happen and I’ve been fortunate enough to experience a few. It’s about building a diversified portfolio of quality companies across the AI value chain, maintaining strict discipline through inevitable hype cycles and corrections, protecting your capital from permanent loss through sound risk management, and giving your thesis sufficient time to play out.

The investors who succeed in AI over the long term will be those who can consistently distinguish between genuine innovation and clever marketing, between sustainable business models built on real economics and temporary hype built on fantasy, between technology that solves actual problems people will pay to solve and impressive technology looking for problems to solve. This requires intellectual honesty, analytical rigor, and the emotional discipline to admit when you’re wrong.

We are still in the early innings of the AI transformation. I genuinely believe this technology will be far more impactful than most people currently imagine, reshaping virtually every industry and creating trillions of dollars in value. But it will take longer to unfold than the most aggressive optimists predict, with plenty of disappointments, false starts, and bankruptcies along the way. The opportunity for patient, disciplined, analytical investors is extraordinary precisely because the path forward is uncertain and most investors lack the patience to see it through.

My portfolio has delivered exceptional returns not because I was smarter than other investors, had better access to deals, or possessed some secret insight into the future. It’s because I developed a systematic approach based on first principles, learned from my many mistakes rather than repeating them, maintained strict discipline during both euphoric highs and despairing lows, and stayed relentlessly focused on fundamentals when everyone else was distracted by hype and fear.

The frameworks I’ve shared in this post—understanding the three layers of the AI value chain, conducting thorough due diligence that goes beyond surface-level analysis, successfully navigating hype cycles by being greedy when others are fearful and cautious when others are greedy, constructing an anti-fragile portfolio with intelligent diversification across multiple dimensions, and managing risk systematically through position sizing and exit discipline—are what allowed me to compound capital consistently through multiple market cycles and technology shifts.

If you’re serious about AI investing, you need to abandon the fantasy of getting rich quick through one perfect investment. Instead, focus on the more reliable path of getting rich slowly and steadily through systematic, disciplined investing. Build conviction through deep research rather than headlines. Size positions appropriately based on risk, not excitement. Diversify intelligently across layers, geographies, stages, and domains. Manage risk religiously like your financial future depends on it—because it does. And most importantly, develop the patience and emotional control to let truly transformative technology work its magic over years and decades, not weeks and months.

The AI revolution is real, not hype. The investment opportunities are substantial and will remain so for many years. But success requires discipline when others are reckless, patience when others are impatient, and a systematic approach when others are making emotional decisions. If you can bring those qualities to your AI investing, if you can learn from both successes and failures, and if you can maintain your discipline through inevitable ups and downs, the next decade could be the most financially rewarding period in your investment career.

Welcome to the future. Let’s invest wisely together.

– Ashley Brown –