AI stock up 500%? Time to trim and buy into chickens
South Korea’s high-flying KOSPI index dropped 10% in a day - along with SK Hynix (KRX: 000660) and Samsung (KRX: 005930) falling more than 12% each - while the Philadelphia Semiconductor Index took a 7.8% hit and Micron (NASDAQ: MU) fell 13%.
So, should investors be running for the hills and dumping their AI holdings? According to Arms Rosenberg, Co-Founder and Portfolio Manager at Minotaur Capital, the answer is no.
That’s probably not much of a surprise given how lucrative Minotaur’s positions in SK Hynix and Micron have been, but it also doesn’t mean she is standing pat and letting it ride.

In this Q&A, Rosenberg explains why its concentration risk rather than a changing thesis that caused her to trim exposure to Micron, the rationale behind buying into a sector that is as uncorrelated from AI as it gets, and why a picks-and-shovels play is on her watchlist.
What’s your most recent investment and why?
Pilgrim's Pride (NASDAQ: PPC). After a week of violent swings in semis and memory, we deliberately wanted something with zero correlation to the AI trade. Pilgrim's is one of the largest chicken producers in the US and Europe: cheap, cash-generative, and benefiting from favourable protein economics as feed costs ease and demand holds up.
It's exactly the kind of unglamorous, defensive cash-flow business that gets ignored in an AI-obsessed market, which is why it's attractively priced. A useful counterweight to a book that's long the memory cycle.
Which investment did you add to your watchlist this week?
Vertiv (NASDAQ: VRT). Everything we saw this week reinforced that the bottleneck in AI isn't just chips: it's power and cooling. Vertiv sells the picks-and-shovels: thermal management and power infrastructure for data centres.
As hyperscaler capex keeps climbing and compute density rises, demand for what Vertiv makes scales with it, without us having to pick the winning chip.
It's on the watchlist while we get comfortable with valuation as it’s one we’ve also owned before.
What is the most recent investment you have trimmed or sold and what drove this decision?
We trimmed Micron. It had been one of our best performers, up over 6x since we bought it in October last year, but Vigil, our AI analyst agent that runs point on portfolio-level risk, had been flagging it hard.
At one stage Micron was contributing over 12% of the entire fund's variance on less than 4% of NAV, and together with SK Hynix the "memory pair" accounted for roughly a quarter of Minotaur's total portfolio risk. Two stocks, one HBM/DRAM pricing cycle.
After a run like that the risk-reward had turned asymmetric, so we top-sliced to manage concentration, not because we'd soured on memory. SK Hynix remains our largest position.
What’s your favourite chart or data point from this week?
I was tempted to put in one of the “Golden Goose” slides from Softbank’s AGM (if you haven’t seen it look it up, it’s hilarious).
But my favourite is this one, from Coinbase's Brian Armstrong. The bars are Coinbase's AI spend; the black line is total token usage. It shows token usage keeps climbing to fresh highs while the dollar cost has decoupled and actually come down. Armstrong's argument is that you don't keep AI spend under control with friction, usage caps and spend alerts; you do it with better engineering: cheaper (often open-weight) defaults, routing each task to the right model, aggressive caching, and keeping context lean. Doing that has cut Coinbase's AI bill nearly in half even as usage keeps compounding.
It resonates because we do exactly this at Minotaur. We've spent serious time doing our own benchmarking of models and building routing logic so a frontier model does the hard reasoning while cheaper models handle routine extraction and summarisation.
To add a bonus chart, here's one which shows our own benchmarking of how each LLM performs writing the first phase of our initiation reports on companies. You can see that Claude Opus 4-8 scores the highest but costs $33 to run vs. Minimax-m3 which is only a slightly lower score but costs $4.12 to run.
The broader signal for investors is hiding in plain sight: the cost of intelligence is collapsing. For all the hand-wringing about an AI-capex bubble, this chart shows adoption can keep growing exponentially without cost scaling alongside it and increasingly the edge isn't access to a model anyone can rent, it's the unglamorous plumbing wrapped around it.
What was your weekly high – a standout market moment or highlight?
Micron's result. Revenue up 346% year-on-year, an 84.9% gross margin, the entire 2026 high-bandwidth-memory output already sold out, and guidance well ahead of consensus. It validated the memory supercycle thesis behind our largest position, SK Hynix, which jumped double-digits on the read-through. When a company that size prints numbers like that, it tells you the AI build-out is still early, not late.
What was your weekly low – a market disappointment or challenge from the week?
Rheinmetall's F126 cancellation. The stock fell drastically after Germany abruptly scrapped the €10 billion-plus F126 frigate programme. The decision caught a lot of investors off guard, us included.
It undercuts the rationale for Rheinmetall's recent NVL naval acquisition — F126 was meant to be the anchor order for that business — and raises an uncomfortable question about the reliability of German defence procurement.
That said, F126 was never a firm order in the backlog: it was a takeover Rheinmetall had offered to lead but hadn't yet signed. The core land, ammunition and vehicles business, i.e. the vast majority of earnings, is untouched.
There's a real cost to digest, and analysts have flagged possible writedowns on the NVL deal, but a €10 billion-plus market-cap hit looks like an over-correction relative to the actual fundamental damage. We're watching closely rather than panicking.
What first drew you to markets or this sector and what continues to keep you inspired today?
I grew up in housing commission in Western Sydney, raised by a single mother with no money. We had an ancient computer, and when I was nine I asked her to buy me a computer game. She told me there was already one on it called Lotus 1-2-3. I worked through the tutorials thinking it was the greatest game ever invented. That's how I ended up spreadsheeting at the ripe old age of nine. Then in Year 10 I played the BRW Portfolio game, where you buy and hold a portfolio of stocks over six months, and did really well. That's where my passion for investing was born.
What drives me today is a firm belief that capital allocation shapes the world and is one of society's highest-leverage decisions: it turns ideas into infrastructure, companies into employers, and innovation into lived reality. When the people making those decisions are homogeneous, the future we fund tends to reflect their incentives and blind spots.
The antidote is diversity of perspective at the capital-allocation table, so progress compounds for everyone, not just a familiar few. Coming from one of those diverse backgrounds, with an innate passion for investing, is exactly why I do what I do.
What’s one piece of advice you’d give to new investors?
Find a style that actually fits your temperament, and then respect risk enough to survive your own mistakes. The Market Wizards books taught me there's no single right way to do this: value, momentum, even short-only can all work!
But the strategies that blow people up are usually the ones they've borrowed from someone else and don’t have conviction to hold through a bad patch. Being right is overrated; what compounds wealth is position sizing, and not losing so much on the wrong calls that you're not around for the right ones.
How do you unwind when you’re not thinking about the market?
I LOVE watching Rugby League and specifically the mighty Parramatta Eels. I have been a blue and gold army member since I was really young and I love watching them play despite all the heartbreak they have put me through - They haven’t won a Grand Final since the year before I was born! This is me on a members' trip to Townsville with two absolute legends of the club, Peter Wynn and Brett Kenny.
Rapid fire! 🔥
What is your favourite investing book?
The Market Wizards series by Jack D. Schwager. My favourite is Dana Galante, who ran a short-only book and did 15% a year while the Nasdaq compounded at 32% a year from 1994 to 1999.
What is your favourite investing or finance/markets related podcast?
Invest Like the Best with Patrick O'Shaughnessy. Patrick is a phenomenal interviewer.
What’s the first thing you read each morning?
The daily market report our AI system, Taurient, generates, or I listen to the five-minute podcast version it generates. It's curated, AI-generated commentary published every trading day, covering major global markets and explaining the big moves in our own positions.
What is your favourite restaurant?
Uncut Seafood on Bondi Road. The best oysters, sashimi and prawn focaccia you'll ever eat. Best washed down with a glass of rose.
What’s something people are surprised to learn about you?
The breadth of my interests. I love classic literature, Kafka and the Brontës especially. I can talk at length about deepening the economic ties between Australia and Indonesia. I scream foul language at referees during Eels games. I love Japanese whisky. I do pole dancing and my favourite pole song is 'Cherry Pie' by Warrant.
Think there’s a better pick? Prove it. Share your rapid-fire book, podcast, and daily read in the comments.
3 topics
3 stocks mentioned
1 contributor mentioned