June Didn’t Change My Portfolio. It Changed How I Look At It.
June was rough. -11.3% for the month. -2.35% for H1, while the S&P’s up over 9%.
| Comparison | June | YTD |
|---|---|---|
| My Portfolio | -11.3% | -2.35% |
| S&P 500 | -1.3% | +9.2% |
I’m not going to spend this post explaining why June happened. Higher rates, AI selloff, China weakness — plenty’s already been written about that angle. I do a portfolio review every month, but this one also happens to be the six-month mark, when I sit down and ask whether anything needs rebalancing or trimming rather than just tracking performance. June just made this particular review a lot harder to skim through than the usual monthly one.
I don’t invest in 25 random stocks. I invest in a handful of themes, and I’ve said as much in this blog for months now — robotaxis, industrials, consumer, healthcare. I’ve been building each one out on purpose, one company at a time. This review wasn’t about discovering that structure. It was about checking on it: how big each theme has grown without me watching closely, whether that size still matches what I intended when I started building it, and what June did to each one once it was under real pressure.
| Theme | Holdings |
|---|---|
| Physical AI / Autonomous Mobility | Pony, WeRide, XPeng, Uber, Grab, Hesai, Albemarle (partial exposure) |
| Alibaba | China’s ecosystem — cloud, e-commerce, AI — my largest single position |
| Healthcare Infrastructure | WuXi AppTec, WuXi Biologics, XtalPi, Novo Nordisk, Doximity, Ovid |
| Industrial / AI Backbone | Honeywell (HON/HONA), Applied Optoelectronics, Ondas |
I built the Physical AI basket because XPeng, Grab, and the rest aren’t the same bet. XPeng is chasing China’s-Tesla. Grab is closer to Uber-plus-Meituan than a robotaxi play. The design was different companies expressing the same theme from different angles, so no single execution failure would take the whole position down. June was the first real test of whether that design works the way I built it to.
I don’t think the mistake was buying any of these. I’d buy most of them again. The question this review is asking is whether owning different companies inside the same theme meant I’d diversified the investment, or just diversified the business models while leaving the money exposed to the same thing once the theme fell out of favor.
Physical AI
The logic here was sound going in. If autonomous driving works, who gets paid? The operator, the platform, the sensor maker, the raw material supplier. Real differentiation, on paper.
| Name | Role | Primary geography | My YTD Return % |
|---|---|---|---|
| Pony AI | Robotaxi operator | China, expanding to Europe, Middle East, Singapore, South Korea | -42.50% |
| WeRide | Autonomous mobility platform (robotaxi + ADAS licensing) | China, Middle East, Europe | -23.69% |
| XPeng | Consumer Physical AI (cars, robotaxi optionality, flying cars, robotics) | China, plus consumer sales across 60+ markets, and now production sites in Indonesia, Austria, and Malaysia | -22.23% |
| Grab | Regional super-app (mobility, delivery, payments) | Southeast Asia; expanding to Taiwan (pending) and US fintech via Stash Financial (closed Jul 2026) | -15.31% |
| Uber | Global mobility marketplace | North America, Europe, Middle East, LatAm | -0.86% |
| Hesai | LiDAR/sensor hardware | Global — China R&D, manufacturing expanding to Thailand, customers incl. Mercedes-Benz, Toyota, robotaxi operators worldwide | +4.08% |
| Albemarle | Lithium/battery materials, upstream EV supply chain | Global — AV is a subset of its exposure, not the whole thesis | +22.84% |
Hesai and Albemarle sell into the theme without needing any one operator to win. Both green. Uber barely moved, since its marketplace exists with or without autonomy. Grab is down a lot, but nowhere near as much as the pure plays, because there’s a profitable business underneath the AV optionality. Pony, WeRide, and XPeng sit at the bottom together: the pure bet, that autonomy has to become real on a timeline the market’s willing to pay for today.
Not diversified, then. Not really. I’d built several different paths to the same destination, and that only shows up once you check whether they move together, not whether the logos are different. They moved together, but not for identical reasons underneath, and that’s worth being precise about instead of writing off the whole basket as one story. Pony is being punished despite genuinely strong operating numbers — the market just isn’t paying for it yet. XPeng’s chart looks similar but the story underneath is different: some sentiment, but also battery costs and R&D spend genuinely eating into margins. The market isn’t wrong to price that in. What the chart doesn’t show is that XPeng is quietly ahead of most of its Chinese EV peers on the thing that determines whether it’s a Chinese exporter or a global manufacturer — it now has production running in Indonesia, Austria, and Malaysia, not just cars showing up on ships. Most of the “Chinese EVs are going global” story is still exports and announcements. XPeng’s version of it is already on the assembly line.
Worth noting how thoroughly the theme got priced as a group rather than company by company. XPeng and WeRide both had a good run of news this year — a safety partnership with Autoliv, a class-leading range result in Norway, an AI model reveal at a major research conference — and almost none of it moved either stock. It kept landing inside broader tech selloffs, where the whole theme was getting sold together, regardless of which company the news was actually about.
Alibaba
Still my biggest position. My YTD return %: -14.58% on the ADR, -35.55% on the HK line. Doesn’t sit neatly in any theme above, which is exactly why it needs its own paragraph instead of getting waved off as “China.”
Some of the decline is broad China sentiment. Some of it’s real — a headline loss last quarter, negative free cash flow from the AI capex ramp, a Pentagon designation that’s genuine political risk.
It’s not my biggest position because I decided it deserved 15% of the portfolio. I’d been trimming it earlier in the year. Then the price fell hard enough that selling more meant locking in losses near the bottom, and I wasn’t willing to do that. So I did nothing, and doing nothing is how it stayed the biggest position by default.
Position size isn’t only about conviction. Sometimes it’s just path dependency — what you did or didn’t sell before the price moved against you. Alibaba’s the clearest example of that in the whole portfolio. It’s shown signs of recovering this month. I plan to keep trimming into that, so the size matches what I believe now, not what it grew into by accident.
Healthcare
Healthcare is built the same way as Physical AI — different companies, different roles, spread across a value chain, not because they’re all healthcare but because each one expresses a different piece of the thesis.
| Name | Role | My YTD Return % |
|---|---|---|
| Ovid Therapeutics | Clinical-stage CNS biotech — the one deliberate exception | +65.96% |
| WuXi AppTec | Outsourced pharma R&D and manufacturing infrastructure | +20.01% |
| WuXi Biologics | Biologics manufacturing infrastructure | +13.10% |
| Novo Nordisk | Commercial pharma, already selling product today | +12.19% |
| XtalPi | AI-driven molecular discovery platform | +1.37% |
| Doximity | Physician engagement software, not a drug bet at all | -37.94% |
Infrastructure held up. Commercial pharma held up. The one speculative bet acted like a speculative bet — it’s my best performer of the year, which I think is more luck than skill, but it was a sized, deliberate bet, not an accident. Doximity’s the outlier, and it’s a software problem, same market-patience issue as Pony’s, nothing healthcare-specific about it.
I don’t think it’s worth guessing too hard at why the market treated the same design differently this year. What matters more is that I noticed the gap at all.
What I actually got wrong
Healthcare has been the strongest theme in the portfolio all year, and it’s smaller than Physical AI by half. I let allocation drift based on how each theme was built originally, not on which one was earning its size back.
The question underneath that isn’t really about how many stocks I own. It’s whether each one is still earning its place. Some clearly are — they give me exposure to a different part of a theme, or a genuinely different source of return. Others I’m less sure about. Am I buying another business, or just another expression of a belief I already own? I don’t think I’ve been asking that often enough.
H2
No dramatic overhaul planned. None of the theses changed in June — I still believe robotaxis get there, healthcare infrastructure keeps compounding, Alibaba’s AI spend eventually pays off. What needs to change is how I build around those beliefs, not the beliefs themselves. Concretely, that means acting on the two things from the section above: rebalancing toward what’s actually earned its size, and getting stricter about what a new position needs to add before it goes in.
On the first one: I’m deliberately not setting a fixed allocation target for either theme. I trade actively enough — sell a tech position, add a healthcare one — that a fixed percentage would be stale within a month anyway. What I want out of this review isn’t a number to hit. It’s the habit of actually asking the question every time I size something, not just noticing the pattern after the fact.
On the second: before adding to anything, am I buying a new business, or just more exposure to a belief I already own? And before trusting that a differentiated position protects me, is this sector currently being sold as a group, regardless of the individual company? That second question is the one that mattered in June, not whether the businesses themselves were good.
One thing sitting in the back of my mind through all of this: I still think there’s a real correction or bear market coming before any of this re-rates higher. I don’t know when. Could be next month, could be next year. I’m not trying to time it, and I’m not making moves based on a date I’ve made up in my head. But it’s part of why this feels more urgent to me now than it would have in January — if a bigger drawdown is still ahead, I’d rather already be sized the way I want to be than find out the hard way a second time.
The themes haven’t changed since January. What’s changed is that having the right themes isn’t the same as having them sized correctly. I’m better at finding good companies than deciding how much of my portfolio each one deserves. Right now, that’s the part of investing I’m trying to get better at. I suspect that’s a more common mistake than buying a bad business in the first place.
Disclaimer: It is not financial advice. The author may hold positions in securities discussed. Readers should conduct their own due diligence and consult with a qualified financial advisor before making investment decisions.