1. My growth portfolio performance

2. A look back on 2025 and what I’m watching for in 2026

3. For paid subs, a deep dive into my portfolio and what I’m thinking

Subscribe now

📐 Performance relative to benchmarks

• My growth portfolio up 86% for the year

• Nasdaq 100 is up 20%

• S&P 500 is up 16%

👉 Important notes:

• All reporting is for my growth portfolio. This excludes other liquid investments like Bitcoin (still HODLing) and a more “set it and forget it” portfolio.

• My growth portfolio is concentrated around certain growth themes with a long investment time horizon. The bets are typically volatile.

• Performance is calculated using time-weighted returns, so deposits/withdrawals don’t skew results. I also track SPY and QQQ the same way to make the comparison apples-to-apples.

  ---

🕵️‍♂️ 2025 Year End Review

👉 Semiconductors (SOXX +40%) was the best performing sector, outperforming the overall market 2 to 1.

AI supply-chain bottlenecks were the dominant driver of outsized returns.

• Memory and optics were the biggest winners. Both entered the year at cyclical valuation troughs, then saw demand surge faster than supply could respond.

  • In memory, that translated into extraordinary performance: Sandisk (SNDK +560%, the best-performing stock in the S&P 500), SK Hynix (+280%), and Micron (MU +235%).

  • In optics, tight supply drove sharp gains as well: Lumentum (LITE +334%) and Coherent (COHR +86%).

• Somewhat unexpectedly, Intel outperformed other AI compute names like Nvidia and AMD. Investors are increasingly viewing the success of Intel’s foundry and advanced packaging efforts as inevitable, as TSMC remains capacity constrained and unable to expand fast enough.

• Looking ahead to 2026, investors are likely to continue chasing bottlenecks. The key question is which of these are temporary, cyclical shortages—and which represent durable, structural shifts?

👉 Since November, however, overall market performance has broadened out beyond AI.

• AI infrastructure related stocks sold off (Semiconductors 🔻1%)

• Investors rotated into big and regional banks (KBWB +10%, KBWR +8%), Healthcare (XLV +8%), and pretty much everything else.

• 2026 could be a big year for banking earnings

  • 10Y/2Y treasury spread is widening, giving bankings greater net interest margins

  • Banking deregulation is expected in 2026

  • Large banks are leaning into AI more than other industries. Some believe that banks will be big beneficiaries, allowing them to run leaner operations.

👉 Enterprise Software underperformed the broader market.

Investors believe that AI will be a disruptive force, unseating incumbents

But there is a clear delineation in the performance between infrastructure and application software

• Infrastructure software is outperforming.

  • AI is pushing more workloads into the cloud, directly increasing demand for infrastructure software, which benefits from usage-based, volume-driven revenue models.

  • Agents will need to be supported with their own cloud environments to do their jobs.

• Investors believe application software, by contrast, face meaningful disruption:

  • Agent-driven workflows break the seat-based pricing model, weakening a core revenue pillar for many incumbents.

  • AI-assisted software development dramatically lowers the cost of building products, giving new entrants the ability to recreate—and rethink—the application stack faster than ever.

  • Autonomous agents can erode workflow moats, bypassing the rigid processes and integrations that traditional application software companies rely on to stay entrenched.

• 2026 will be a pivotal year for enterprise software. After overhype, companies will have to prove that AI-driven products can deliver real, durable value at scale.

Neoclouds (and Oracle) had a great year overall, but with lots of volatility

• Corweave (CRWV) experienced the first rollercoaster climb, +360% after IPO, but soon fell by 75% from it’s peak. It still finished +85% since IPOing

• Bitcoin miners like IREN, Cipher (CIFR) and HUT8 (HUT) rode the hype wave, only to fall dramatically after investors became weary of the economics and risk. Though volatile, they still were up a lot on the year.

• Oracle (ORCL) was up 100% on the year with massive OpenAI contracts, but the market slowly started to discount OpenAI’s commitments. Oracle is probably the cleanest public market expression of OpenAI’s success and failures.

• These players will face two challenges in 2026

  • Financing: They require enormous amounts of capital to develop data centers. Expect equity dilution

  • Supply Chain: The power supply chain is constrained. Expect more missing data center delivery targets and delays.

Rivian and Tesla were big winners in the EV/autonomy theme, with the potential future getting priced in.

• Investors are pricing in a lot of future upside from Tesla’s autonomy prospects with breakthroughs in FSD

  • Tesla will need to show that FSD can translate into new car sales. Anecdotally, I think it is already happening.

• Rivian’s upcoming R2 mass market vehicle is starting to get notice from investors. Rivian has also been emphasizing their own end to end homegrown autonomy hardware/software stack.

  • Rivian will need to show they can execute. Scaling a mass market vehicle will not be easy.

• Uber finished the year 20% off all time highs. It almost seems like the better Tesla’s FSD does, the more investors want to discount Uber’s role in the autonomy ecosystem. It’s ironic that Uber will start 2026 the same way it did 2025, with doubt (the stock was up 35% in 2025).

  • We are still in early innings for how the autonomy ecosystem will play out. In 2026, we need to see more viable autonomy alternatives to Tesla and Waymo for Uber to benefit.

Precious metals won the year and outperformed most growth themes.

• Gold continued its climb this year as several major countries moved to diversify reserve holdings away from the U.S. dollar.

  • In 2025, gold surpassed the Euro to become the world’s second-largest reserve asset.

  • The bigger question now: could it challenge the dollar in 2026 as the dominated reserve asset? I think so. Structurally, nothing is changing the path we are on today.

• Silver, platinum, paladium and copper followed gold higher, as they often do after sustained moves in gold prices. Silver, in particular, remains a speculator’s favorite, and that showed up clearly in the late-year surge, which had all the hallmarks of speculative excess.

  

Crypto was a lot of hope and hype in 2025.

Despite Bitcoin’s institutional and regulatory acceptance, it failed to hold above $100K

• A lot of things were going right for Bitcoin—pro-crypto regulation, institutional adoption, etc.

• In reality, a lot of the fever was driven by equity fueled Digital Asset Treasury companies (like MicroStrategy). But liquidity dried up, leverage got swept out, OG Bitcoin whales took gains, and fears of quantum are rising (and will continue to rise)

• But remember: “Volatility is the price of admission. The prize inside are superior long-term returns.” It’s just another year for Bitcoin. Don’t get caught up in the hype or fear either way.

Ethereum was an utter disappointment in 2025, but it may be too early to judge.

• It suffered in many of the same ways as Bitcoin

• 2026 will need to be a show me year—less talk, more adoption of stablecoins and tokenization

Finally, US treasuries rates declined more than other developed countries

• Despite fears of tariff induced inflation, US treasury rates fell faster than other developed countries.

• Folks are expecting rates to fall and financial conditions to ease further in 2026.

• But we may not see the 10Y go down at the sam rate as the 2Y.

🔍 What’s in my portfolio?

➡️ By Theme

• 65% AI (datacenter buildout + applications)

• 19% Cash+Gold (anticipation of volatility)

• 16% Everything else — EV/Autonomy, New Fintech, and GLP-1

➡️ By Individual Investment

Note: The change column is the delta from my last portfolio update.

Commentary

AI Infrastructure

• Exited all neocloud exposure: I did very well with neoclouds this year, with big realized profits in Coreweave, Nebius, and IREN. I decided to exit these positions because there was just too much hype around these names. The narratives drove the stocks way higher than they should have gone. There are still meaningful risks that I don’t think is properly priced: 1) They will dilute their shareholders to finance development 2) The power supply chains could cause delays. I still think there is an important role for these neoclouds to play, but will happily enter when prices become more reasonable.

• Exited Power Integrations (POWI): I opened this trade because I thought POWI would be a beneficiary of Nvidia new datacenter power architecture. So I did what Druckenmiller would do: invest first, and research later. After doing more research, I think this story is not clear cut. I ended up selling for a loss unfortunately.

• Opened positions in Sandisk and Lumentum: These companies are beneficiaries of supply chain bottlenecks to support the datacenter buildout.

  • Sandisk produces NAND flash memory, which is becoming increasingly important the memory stack for hyperscalers. They need “hot” memory that is fast to recall past conversations, context, etc. in AI chats. Is this cyclical or structural? Right now, Sandisk trades like it’s cyclical, meaning investors are expecting oversupply. I think there is structural demand that will last longer than anyone thinks. At 16x fwd earnings, I actually think SNDK is not a bad price. Let me know if you want a longer write up on this one. I’m holding my position.

  • Lumentum makes Indium Phosphide lasers, which are a key component in networking equipment in the datacenter. Now that the GPU is datacenter scale, each chip needs to talk to each other at the same time to deliver the AI goods. This means there is a huge step level change in the amount of optical content (and therefore lasers) required. There is a shortage of Indium Phosphide capacity and Lumentum is one of the key players. Is this cyclical or structural? I think it’s a bit of both. The Indium Phosphide bottleneck should resolve. But datacenter networking architectures are changing rapidly to allow for more performance and Lumentum’s new technology (optical circuit switching) is getting adopted. This could support the next leg of growth beyond bottlenecks. I’m holding my position. The stock has run up A LOT. I may take some chips off the table. Will let you know.

• Opened a position in natural gas (UNL): UNL is an ETF that tracks the rolling 12 months forward contracts of natural gas. The bet here is that power constraints will force hyperscalers to build their own natural gas powered infrastructure, rather than relying on the grid. This would create huge new demand for natural gas. Natural gas is pretty cheap right now, so the risk/reward seems good. But this one is volatile (they call natural gas the widow maker). I’m holding for now, and looking for a less volatile expression of this trade.

• Other moves:

  • Added to Teradyne (TER): Still believe there multiple growth levers yet to be realized. But majority of the % of portfolio increase was price appreciation.

  • Trimmed Broadcom (AVGO): I really struggled with this one. AVGO is a one of a kind company. I do think the valuation was getting stretched given emerging risks (MediaTek taking some of the TPU biz, OpenAI reliance, etc.).

  • Converted NVDA equity into calls: I took some long-term gains on NVDA stock and bought some leap calls. This effectively unlocked some capital I could allocate somewhere else.

AI Applications

• Added to ServiceNow: ServiceNow has reversed trend and trading near 52 week lows. I’ve been adding on the way down. I do think they sit in a unique spot compared to other application software. I think the risks are overblown, though they have been selling the agent dream hard and wonder if they oversold the dream. I will be looking at this one closely. Right now, it feels like it’s bottomed at 10x fwd revenue. Holding.

• Added Microsoft Leap Calls: Microsoft arguably has the most surface area to put AI compute to work (enterprise, cloud, GitHub, gaming, consumer, etc.). They certainly have the right strategy (watch this interview with Satya). And don’t forget, they also have access to ALL of OpenAI’s IP. Everything. Though there are some bumps the adoption of Copilot in the enterprise, I think 2026 will be an interesting year for MSFT. I decided to do a levered bet via the leaps. 28x fwd earnings seems reasonable, but they need to beat the street’s expectations on earnings growth (currently 20ish percent). I’m holding.

• Other moves

  • Added to DataDog: DDOG smashed earnings, the stock rocketed to $180, where it was expensive at almost 18x fwd revenue. Then the stock dropped to 130s at 11.5x fwd revenue, where it’s cheap. I think the market is getting swept up in narratives. DDOG is in a very good position to capitalize on an acceleration to cloud adoption. I’m holding, but could add more if we go lower.

  • Trimmed Google: Love google, just ran up a bunch. This was more of a rebalancing move.

  • Holding Snowflake: SNOW has also been executing very well. They are benefiting from a massive tailwind on data platforms. I think investors are weary because they got burned 5 years ago (the stock still hasn’t recovered from Covid highs). I’m holding.

Other

• Sold some gold: This was a mistake. Gold has room to run in 2026. I’m holding.

• Added to Uber: The autonomy story is nowhere near over. I’m holding.

• Trimmed Lilly/Added Novo Nordisk: The valuation spread between the two started to not make sense. I think the GLP-1 pill will be a different dynamic, and might give Novo some leeway to catch up.

  Looking for an exit on Copper: Made my nut on Copper, I think there might be some more room to go. Sold 1/4, will close out rest in the next few weeks.

• I regret selling Rivian: Biggest lesson learned—don’t get in too early, and don’t become impatient if you do. I am anticipating execution missteps, and will look for an entry.

• Last but not least, Ethereum was a fail in 2026: First, I will say that I still think the tokenization / stablecoin theme will play out and Ethereum should benefit. But there was definitely a timing / hype mismatch. I did take some losses to reduce exposure, but I’m holding now.

📐 Performance relative to benchmarks

• My growth portfolio up 84% YTD

• Nasdaq 100 is up 18%

• S&P 500 is up 13%

👉 Important notes:

• All reporting is for my growth portfolio. This excludes other liquid investments like Bitcoin (still HODLing) and a more “set it and forget it” portfolio.

• My growth portfolio is concentrated around certain growth themes with a long investment time horizon. The bets are typically volatile.

• Performance is calculated using time-weighted returns, so deposits/withdrawals don’t skew results. I also track SPY and QQQ the same way to make the comparison apples-to-apples.

  ---

🎈 Are we in an AI bubble?

Short answer: No.
Will there be one? Inevitably, yes.

Let’s unpack the main arguments I hear for why people think we’re already in one — and why I disagree.

1️⃣ “Investment is way ahead of ROI — just like the Dotcom boom”

AI ROI is more than just new revenue generated. It’s greater earnings power and cash flow generation. Hyperscalers have already shown unprecedented earnings growth — faster innovation cycles (Nvidia), higher ROI to advertisers (Meta, Google), and massive productivity gains (revenues soaring while headcount flatlines).

Beyond big tech, we’re seeing the fastest revenue ramps in history.

• AI coding software like Cursor and Replit hit $100M ARR in under a year. Cursor rumored to hit $500M ARR, up from $100M, in just a few months

• OpenAI is nearing 1 billion users just three years after ChatGPT launched.

And we’re still early. Models are improving, costs are falling, and new use cases are exploding across industries (see my past posts here and here to read more). Like many transformative technology waves in history, Wall Street lacks the imagination and underestimates the market size and rate of adoption.

As for the dotcom boom/bust analogy— it’s lazy.

• Telecom deployed “dark fiber” ahead of demand. Today, GPUs are melting from too much demand.

• Dotcoms launched with no business models; AI companies already have customers and revenue.

• The internet took years to scale. AI adoption rides on existing rails — cloud, broadband, mobile devices — making its diffusion unprecedented.

2️⃣ “Valuations are extreme”

Rebuttal: Not really. Earnings growth justified it.

Legendary investor Howard Marks commented that “valuations are not crazy” to qualify as mania. He goes on to say that though the S&P 500 multiples are higher than in the past, the composition of companies are just better—growth, profitability, etc.

In Q2, S&P earnings grew 12% YoY, driven by the same tech leaders commanding higher multiples. Most large-cap tech names are still well below 2021 mania levels.

3️⃣ “Deja vu; investment-revenue round-tripping”

Rebuttal: It’s not the same thing.

OpenAI’s partnerships with Nvidia, AMD, and Broadcom have drawn comparisons to the Dotcom-era “round trips.” But strategic reinvestment across supply chains is normal.

• Retailers invest in farms and logistics.

• Auto OEMs co-fund suppliers to advance battery tech.

• Pharma bankrolls contract manufacturers to secure capacity.

AI is no different. Nvidia is overflowing with cash with nowhere to put it. Rather than buy back shares or chase M&A distractions, it’s funding the ecosystem that is so hungry for capital. It’s the best thing it can do for itself and its shareholders.

For OpenAI, these are largely non-binding commitments that:

• Create competition among chipmakers

• Lock in manufacturing capacity at TSMC and memory providers

• Optimize different chips for diverse workloads and total cost of ownership (TCO)

It’s strategy, not excess.

🫧 But make no mistake — a bubble will form.

We’re just not there yet.

• Valuations have a ways to go before qualifying as mania.

• The first real stress test may come next year when Hopper GPUs hit four years old (Coreweave depreciates servers over six). Wall Street tends to think too linearly here — not all workloads need the latest silicon. Many can run profitably on older chips; Oracle still uses GPUs two generations back. Needless to say, I’m watching whether neoclouds start taking write-downs — and if GPU rental prices stay stable.

• Power is another looming constraint. There’s still headroom today, but growth will eventually hit a wall. The good news: both the government and private sector are moving fast to expand capacity. Let’s see if they can get it done.

• While the VC model has been under pressure, AI funding has revived it — nearly $200B this year alone. Given the success of startups in the last year, I think this will continue to grow to some obscene level, when dollars will start to fund bad ideas and start the unraveling.

🔍 What’s in my portfolio?

➡️ By Theme

• 64% AI (datacenter buildout + applications)

• 21% Cash+Gold (opportunistic dry powder + macro hedge)

• 15% Everything else — EV/Autonomy, New Fintech, and GLP-1

➡️ By Individual Investment

Read more

That’s what I did today. I’m going deep into a new theme and though I’m not done with my research, I think we might be at an inflection point.

So I “drucked” in…

Read more

• My growth portfolio up 66% YTD

• Nasdaq 100 is up 16%

• S&P 500 is up 13%

I’ll be sharing regular updates to give you a peek into how I manage my own money. These updates will cover:

• Performance vs. benchmarks

• Allocation across themes / individual stocks

• Work in Progress thoughts

• Major portfolio changes and the rationale

👉 Important note: all reporting is for my growth portfolio. This excludes other liquid investments like Bitcoin (still HODLing) and a more “set it and forget it” portfolio.

📈 What is my growth portfolio?

• Mostly thematic driven — investments tied to big picture ideas

• I only enter positions if I have long-term conviction (not trading in and out)

• Higher volatility and concentration than the market, but higher potential reward if themes play out

Having an investment thesis is one thing. Executing on it means constantly evaluating the ever evolving ledger of risks and reward.

Ok, let’s get into it.

📐 Performance relative to benchmarks

It’s been a great year for equities across the board:

• S&P 500 (SPY): +13% YTD

• Nasdaq 100 (QQQ): +16% YTD

• My growth portfolio: +66% YTD, driven primarily by AI exposure

Note: Performance is calculated using time-weighted returns, so deposits/withdrawals don’t skew results. I also track SPY and QQQ the same way to make the comparison apples-to-apples.

🔍 What’s in my portfolio?

➡️ By Theme

• ~64% AI (datacenter + applications)…yea, I’m bullish

• ~20% Cash+Gold (opportunistic dry powder + macro hedge)

• ~16% Everything else — EV/Autonomy, New Fintech, and GLP-1 (I hope to write more about these)

➡️ By Individual Investment

Read more