Let me extend my previous discussion on my stock investment hits in the US market.
Unlike my India portfolio, where I mostly stick to proven compounders, the US market gives me the opportunity to invest directly in technologies that I use, understand, and often witness firsthand. Not every investment worked out, but a few have more than compensated for the mistakes.
As always, I am not a financial analyst or advisor—just a techie who enjoys investing and happens to write down his thoughts publicly. These are simply my personal observations.
Nvidia – “I Don’t Want to Pass Up This Opportunity Again”
One of the biggest hits was undoubtedly Nvidia.
Back in 2022, when semiconductor stocks were being punished and AI was still a niche discussion, I wrote:
“AI and Cloud are the future.”
and later:
“I don’t want to pass up this opportunity again.”
That statement was heavily influenced by one of my biggest investing regrets—passing on Nvidia when it was a $25 billion company despite fully understanding the importance of GPUs in AI workloads.
This time I was determined not to repeat the same mistake.
While many investors focused on gaming weakness, inventory corrections, and valuation concerns, I focused on Nvidia’s unique position as the foundational infrastructure provider for AI. The arrival of ChatGPT and the subsequent explosion of AI adoption validated that thesis much sooner than I expected.
My average Nvidia purchase corresponds to roughly a $250 billion market capitalization. Since then, it has become one of the most successful investments of my life. I have already sold about 50% of my position after achieving a 10X return, while the remaining 50% continues to stay invested.
Unlike Amazon, Google, Bitcoin, Tesla, Dixon, MCX, and several other winners that slipped from my hands too early, this time I intend to stay aligned with the long-term thesis. My belief is that we are still in the early stages of AI and that the next phase—Physical AI involving robotics, autonomous systems, industrial automation, and intelligent machines—could potentially be even larger than the current wave.
One aspect of Nvidia that continues to strengthen my conviction is the power of its ecosystem. While competitors such as AMD, Google, Amazon, and other hyperscalers are building their own AI hardware, Nvidia’s moat extends far beyond the GPU itself. Over the years, Nvidia has built a comprehensive software and developer ecosystem around CUDA, which has become the default platform for AI development.
The key difference, in my view, is that CUDA benefits from a vast global developer community. Every year, millions of researchers, engineers, startups, enterprises, and open-source contributors collectively improve tools, frameworks, libraries, and workflows built on top of Nvidia’s platform. Replicating that ecosystem is significantly harder than replicating a chip.
This is why I do not view Nvidia as a traditional semiconductor company. The business reminds me more of successful platform companies that benefit from network effects. The more developers build on CUDA, the more attractive the platform becomes, which in turn attracts even more developers and customers. This is also the major difference between Nvidia open ended and Cisco’s close ended ecosystems during the dotcom era.
Historically, technology platforms with strong ecosystem advantages have proven remarkably durable. Apple’s App Store ecosystem is one example. In enterprise software, Microsoft benefited from similar dynamics. While no moat lasts forever, I believe Nvidia’s ecosystem advantage is currently underestimated by many investors who focus primarily on hardware specifications.
Whether Nvidia ultimately follows a trajectory similar to Apple remains to be seen. However, I would not be surprised if the company’s competitive position proves far more durable than many expect, precisely because its moat increasingly resides in software, developers, and ecosystem rather than silicon alone.
From a valuation perspective, my bull case for Nvidia is driven by the size of the AI opportunity rather than near-term earnings. If global AI infrastructure spending eventually reaches $2 trillion annually and Nvidia retains roughly 50% share of the ecosystem, the company could potentially generate around $1 trillion in revenue. Applying a platform-like valuation multiple to a business with Nvidia’s ecosystem strength could justify a market capitalization measured in the tens of trillions of dollars over time.
Of course, this is a long-term scenario rather than a prediction. But just as very few investors anticipated today’s multi-trillion-dollar technology giants, I believe the market may still be underestimating the ultimate scale of the AI economy and Nvidia’s position within it.
I also hold a meaningful position in ASML, which has been another AI winner for me. While Nvidia became the face of the AI boom, ASML remains one of the most important enablers behind the scenes. My thesis was simple: if AI requires increasingly advanced chips, then someone has to build the machines that make those chips possible. In many ways, Nvidia supplies the brains of AI, while ASML supplies the tools required to manufacture those brains.
Takeaway: My biggest investing mistakes were rarely buying too late; they were usually selling too early.
Palantir – The Enterprise AI Infrastructure Layer
My second major AI winner has been Palantir. In fact, Palantir has generated an even larger return (15% of my portfolio) for me than Nvidia.
While Nvidia became my AI hardware winner, Palantir became my AI software infrastructure winner.
My average purchase corresponds to roughly a $22 billion market capitalization (another 10X+ investment from my US portfolio). At the time, most investors viewed Palantir as a slow-growing government contractor with excessive stock-based compensation. I focused instead on the rapid growth in commercial customers and the increasing adoption of Foundry and AIP.
Coming from a technology background, I understood early that large language models are fundamentally probabilistic systems. They are incredibly powerful, but they remain non-deterministic by nature (except coding where models can code/compile/test/refine to make it somewhat deterministics). Even the best models struggle to achieve perfect accuracy in many real-world tasks.
This creates a challenge for enterprises. CEOs want to adopt AI, boards want AI strategies, and employees want AI tools. However, enterprises cannot run critical business processes on systems that are inherently unpredictable.
That is where Palantir’s AIP platform caught my attention.
My thesis was that enterprises would not simply buy AI models—they would need a layer that makes AI useful within real-world workflows. AIP sits above foundation models and integrates them with enterprise data, permissions, governance, and operational systems. In simple terms, it helps transform probabilistic AI outputs into deterministic business actions.
One way I think about Palantir is through a lesson from the internet era.
Microsoft owned the operating system (Windows), the killer productivity suite (Office), and the browser (Internet Explorer). Yet Google came later and used that infrastructure to solve the internet’s biggest problem—search. Over time, Google arguably created more value on top of the Windows ecosystem than Microsoft did from Windows itself.
A similar dynamic may be emerging in AI. Today, hyperscalers and LLM providers are building the infrastructure layer, but enterprises still need a way to integrate AI into real-world workflows with governance, security, and reliability.
This is where Palantir stands out. While others compete to build better models, Palantir focuses on making those models useful inside actual businesses. If AI follows a path similar to the internet, the biggest value creator may not necessarily be the company building the models, but the company that helps enterprises turn those models into business outcomes.
The market initially underestimated the value of this capability. Today, AIP adoption continues to accelerate, commercial customer growth remains strong, and Palantir has managed to combine software-level gross margins with accelerating growth—something rarely seen at scale.
One of the reasons I continue to hold the stock despite its massive run-up is that I believe the market may still be underestimating the size of the opportunity. Hardware companies have already demonstrated a path to trillion-dollar valuations. My belief is that software platforms enabling enterprise AI could eventually become equally important.
Whether Palantir ultimately becomes the first pure-play enterprise AI software company to reach a $1 trillion valuation remains to be seen, but I believe the upside is significantly larger than what most investors currently envision.
Like Tesla, Palantir has attracted plenty of star short sellers and skeptics. The difference this time is that I have my own thesis and conviction. I may ultimately be wrong, but I won’t sell simply because others are bearish. If my investment thesis changes, it will be due to business fundamentals, not market noise.
Takeaway: The biggest winners in a technology revolution are often the companies that make the technology deployable at scale.
CrowdStrike – Great CEOs Don’t Waste a Crisis
“Great CEOs Don’t Waste a Crisis”, this is what I predicted and wrote here when crowdstrike was navigating the blue screen of the death crisis during early 2024.
CrowdStrike has been another successful investment where my technical background helped me separate noise from signal.
Back in 2022, during the technology bear market, I repeatedly argued that there is no recession in cybersecurity. Companies may delay cloud migrations, reduce hiring, or postpone discretionary IT spending, but they cannot afford to ignore security.
More recently, CrowdStrike faced two major challenges.
The first was the global outage that temporarily damaged the company’s reputation and triggered widespread criticism. While the market focused on the immediate fallout, my view was different. Great CEOs with great products do not waste a crisis—they embrace it, learn from it, and use it to build a stronger company for the future.
The second challenge came during the recent wave of frontier AI model releases and the so-called “SaaS apocalypse” narrative. Investors suddenly began questioning whether AI would disrupt traditional software companies, including cybersecurity vendors.
I disagreed.
My technical understanding is that cybersecurity is probably the last domain where enterprises would be comfortable relying entirely on probabilistic AI systems. False positives create operational chaos; false negatives can create catastrophic consequences. Reliability, governance, and trust matter far more than flashy demos. On top of that CRWD possessed the cleanest cloud based architecture for cybersecurity solutions.
In my view, AI strengthens cybersecurity leaders rather than replacing them. As attackers gain access to increasingly powerful AI tools, defenders will require even more sophisticated security platforms to keep pace.
That conviction led me to add to my CrowdStrike position during both periods of uncertainty rather than reduce it. So far, the thesis has worked out well, and I am currently sitting on roughly a 5X gain from my average purchase price.
Takeaway: Great businesses don’t avoid crises—they emerge stronger from them.
Cloud Infrastructure Winners – The Picks and Shovels of the Digital Economy
Alongside my AI investments, I also built positions in cloud infrastructure companies such as Snowflake, Cloudflare, MongoDB, Datadog, UIPath and Zoom as I believed AI and Cloud are complementary solutions for scaling the model usage.
My thesis was straightforward: regardless of which AI applications ultimately win, the world will need more data, more cloud infrastructure, more observability, more databases, and more real-time data processing for scaling the model/AI usage.
These companies may not receive the same attention as Nvidia or Palantir, but they occupy critical positions within the modern technology stack.
-
Snowflake became a key data platform for enterprises.
-
Cloudflare evolved from a CDN provider into a global application and security platform.
-
MongoDB continued benefiting from the shift toward cloud-native development.
-
Datadog established itself as a critical observability platform.
Not every investment generated Nvidia-like returns, but the underlying business trends largely evolved as expected.
Takeaway: Infrastructure businesses may not always be the most exciting, but they are often the biggest beneficiaries of long-term technology adoption.
What Am I Doing Now?
Today, my approach is fairly simple. Rather than trying to predict individual AI winners in isolation, I look at the entire AI value chain and invest only in the layers that I understand well.
I have consciously avoided the power generation and data center infrastructure layer because I do not have sufficient domain expertise there. While attractive opportunities may exist, I prefer to stay within my circle of competence.
In the computer hardware layer, I already have my winners in Nvidia and ASML.
In the hyperscaler layer, Microsoft remains my preferred way to participate in enterprise AI adoption.
In the software infrastructure layer, Palantir has been my biggest winner. I also hold UiPath and believe it could potentially benefit from a similar trend as enterprises increasingly combine AI with automation to overcome some of the limitations of large language models.
The foundation model layer is largely inaccessible to public market investors since most leading AI model companies remain private.
That naturally pushes me toward the application layer, where I believe some of the biggest long-term winners are likely to emerge.
Historically, the largest value creation in technology often happens closest to the customer. Infrastructure providers build the foundation, but application companies ultimately capture much of the business value.
With that thesis in mind, I started building positions in Lemonade and Upstart as early as 2022 when I began constructing my AI-focused portfolio. These have easily been the most volatile investments in my portfolio. Despite significant drawdowns, I continued accumulating both companies all the way down—from prices below $100 to nearly $10–11 in some cases.
The reason was simple: my conviction in the long-term thesis remained intact.
Lemonade is attempting to reinvent insurance through AI-driven underwriting, claims processing, and customer acquisition, while Upstart is trying to modernize lending by moving beyond traditional FICO-based credit scoring.
Both remain controversial, both face significant execution risk, and both have experienced substantial volatility. However, if AI truly transforms decision-making across industries, these are exactly the types of businesses that could create asymmetric outcomes over the next decade.
I also continue to hold positions in Shopify and Spotify. While neither is a pure AI investment, I believe both stand to benefit significantly from AI adoption.
Shopify remains my preferred way to participate in the long-term evolution of digital commerce. My thesis is that AI will dramatically lower the barriers to entrepreneurship by enabling merchants to build stores, create content, manage inventory, provide customer support, and market products with far fewer resources than before. If AI creates millions of new digital businesses over the coming decade, platforms such as Shopify could become major beneficiaries.
Spotify is a different type of investment. My conviction comes primarily from the product itself. After years of using the platform, I continue to be impressed by its user experience, recommendation engine, and ability to personalize content at scale. AI should only strengthen these advantages through better discovery, personalization, creator tools, and new forms of audio content. Sometimes the best investments are simply products that become an indispensable part of everyday life.
Finally, I continue to hold Tesla. After selling it too early in the past, I am now comfortable staying invested and backing an exceptional founder who has repeatedly proven doubters wrong.
Whether these investments ultimately succeed remains uncertain, but after benefiting from winners in AI hardware, cloud, and software infrastructure, I believe the next phase of wealth creation may come from AI-native applications that directly transform real-world industries.
Final Thoughts
Looking back, the common thread behind most of my successful US investments was staying within my circle of competence.
I rarely try to predict interest rates, GDP growth, elections, commodity cycles, or geopolitical events. Instead, I focus on technologies, products, and trends that I understand from firsthand experience.
The real money is made by identifying transformative technology waves early, backing exceptional management teams, and then having the patience to let the thesis play out.
Equally important, staying invested through major macro uncertainties is often where the actual wealth creation happens. Over the last few years, markets have navigated everything from inflation scares and rate hikes to tariff-related volatility and geopolitical conflicts. Every event created a compelling reason to sell.
In my experience, it is impossible to hold through such periods unless one has high conviction in the underlying businesses.
Looking back at both my hits and misses, I have come to realize that my biggest investing mistakes were rarely buying the wrong companies. More often, they were selling the right companies too early.
The encouraging part is that investing is one of the few fields where mistakes can become assets if we learn from them.
Ironically, as a Software Engineer the biggest financial gains/wealth creation of my career have not come from just writing software code, but from investing in the very technologies that are reshaping how software code gets written.
Looking back, that may be the single biggest lesson from my journey so far: the best investment opportunities are often hiding in plain sight, closest to our own expertise.
More importantly, this journey has given me something far more valuable than investment returns—it has given me optionality. The freedom to consider retiring from the corporate world and the opportunity to spend more time building something of my own.
Let’s see where the next chapter takes me. :)
For now, I remain optimistic, patient, and humble enough to accept that luck plays a much larger role in investing than most of us would like to admit.
Takeaway: Conviction allows us to buy during panic, patience allows us to hold through uncertainty, and luck ultimately determines how much of the opportunity we capture.
P.S. Looking back, what the US market went through in 2022 feels somewhat similar to what the Indian market is experiencing in 2026. Back then, sentiment was extremely negative, technology stocks were being written off, and few expected what was about to happen. ChatGPT arrived in late 2022 and changed the conversation around AI almost overnight. The result was one of the strongest periods for technology-led wealth creation in recent memory.
I could be completely wrong, but my feeling is that the current phase in India is temporary as well. Technological disruptions often emerge when expectations are at their lowest. Something meaningful from the AI wave will eventually find its way into India too. It may come from an unexpected corner of the market—perhaps even the currently neglected IT services sector—and who knows, it may ultimately create India’s first trillion-dollar company.
As investors, we don’t need to predict exactly how it will happen. We only need to stay open to the possibility.
One principle that has served me well over the years is simple: never bet against the United States and never bet against India. Both countries have repeatedly found ways to innovate, adapt, and create enormous wealth over long periods of time. Personally, some of my best investment outcomes have come from sticking to that belief.
Happy Investing!
(this is the only stock trading related book I read) After reading the book I somehow understood how difficult it’s to create such system (in fact this team first used AI/ML to do the quant trading) and why it’s futile for me to try the same before achieving my FI:) As a result with a corporate demanding private job in hand, I went with long term investment (along with MF/ETF investment) as it demands little time to generate above average return (specially in India where leaders don’t get disrupted frequently like in US).
