Back in a fresh $3350.JT long @ $602.2 Yen

Newly elected JP PM seems a massive fiscal dove. Additionally Gold has been breaking ATHs after ATHs, BTC is also starting to as well — a sign of the times we’re in.

Metaplanet is 61% off from where we exited for a ~2x gain; R/R seems decent at these levels.

$AMD +23% on deal with OpenAI

- OpenAI to deploy 6 GW of AMD
- Starts with 1 GW MI450+ from H2 2026
- OpenAI to invest in AMD up to 160m new shares (10% dilution)

6GW is insane scale — eqv to hundreds of thousands of chips. AMD probably hasn’t pre-booked this capacity at $TSM given the current backlog..

Hence, I’m assigning a higher probability to the rumor of AMD fabricating at $INTC being true.

If the 50/50 semiconductor tariff rule is indeed introduced by Trump, napkin maths would mean that 3GW of chips will have to be produced in the U.S. (or face exorbitant tariffs).

Virtually impossible to do this without engaging IFS $INTC.

Scaled into a small/mid-sized long on $AMD @ $219.64; will wait for cash open to decide whether to add or wait for a bigger dip.

This deal is huge and opens up doors in the future by validating AMD’s chips.

Closed $AMD @ $211 for -4%. Decided to be exposed from another angle

Long $MRVL and $ALAB @ $87.4 & $223.08

Think we should not view $AMD news as a standalone, but as a gradual shift towards alternative GPUs/custom silicon.

Custom silicon adoption & disaggregated SoCs, alongside an acceleration in AI inference workloads which require low latency and high throughput, all serve as tailwinds for the connectivity layer (switches, retimers, smart fabrics, protocol controllers etc). Interconnect may become a critical bottleneck as these AI clusters get larger, more heterogeneous and complicated, and further away from each other.

Markets sold off yesterday from The Information’s report that $ORCL has lost $100mn from rental of $NVDA’s Blackwell chips, thus casting doubts on the ROI of AI capex.

My view is it’s not that shocking that GM% is low during the initial ramp period of filling up the data centres and securing customers. One way to explain this is timing gaps in revenue recognition, another part is that DCs have very high fixed costs, and lower utilisation (which will increase as the DC ramps) = lower margins.

Not sure on the intent of the report but it seems to have conveniently omitted crucial aspects of the business model in exchange for a viral headline.

we’re so back

Memory players got a few sell-side upgrades this week; bears being converted slowly.

I think market is still undervaluing how important memory/storage will be, especially when considering how much of a supply glut there is for HBM and NAND (partly because most capacity got diverted to building HBM).

$SNDK $MU

Nice moves from these two companies on the first day of China’s open after Golden Week. $600089 is up 21% since entry, while $600869 is up 3%.

While the idea of gaining exposure to power grid expansion isn’t new (see the 1-year performance of U.S. power stocks), we are specifically choosing to bet on China’s power grid expansion because I believe it’s one of the edges — and potential ways — that China can catch up to the U.S. in terms of compute power. If you notice, there has been a gradual separation of AI supply chains between the Eastern and Western hemispheres. Nvidia’s AI chip exports to China has been subjected to heavy scrutiny from the Trump administration, while China itself has “banned” its corporations from buying Nvidia chips.

Why did China do that, despite knowing that its chips are currently far behind U.S.-designed ones? Obviously, there are national security and strategic reasons at play (some say it’s also for negotiation leverage). But are they really going to willingly fall behind in the AI race just to promote self-reliance? I don’t think so. I believe they’re going to try and brute-force compute power.

China’s industrial power rates are roughly 2–3× cheaper than that in the U.S. That means even if their domestic chips have poorer performance per watt (PPW), considering cheaper power and total cost of ownership (TCO), they may actually not be that far behind the U.S. in terms of compute power. Under this theory, power becomes the key factor until their domestic chips achieve near parity PPW. Power must hence remain abundant, cheap, and widely accessible.

Cut this @ $584 for -3%

Think I’m better off just buying spot $BTC rather than DATs tbh for exposure to loose fiscal policies.

Fully exited $APLD @ $28.03, +87% from entry

Made enough on the name, and want to allocate more to other areas of the portfolio. Also don't want to take the print risk on a small cap after such a rally.

Will see post-earnings reaction to determine if I re-enter.

Somewhere in the recent months, Sama has developed an unwavering conviction in the capabilities of future models and the economic value they’ll create.

Hence, they’re now scaling the infras ahead of time to support that model — the infras decision is not based on today’s model requirements, but that of tomorrow’s model.

All these headlines with mind-blowing numbers being thrown around every few days. Everyone was laughing last year when Sama was trying to raise a trillion dollars. Now, it’s like just a trillion.. really..?

At the scale that OpenAI plans to build at:
1) they’re going to force all of the tech giants into an AI race — if AGI really is the end-all be-all, it probably at best would be a duopoly. Who will be the 2 players?

2) they won’t just use some suppliers within an industry, rather they’ll need WHOLE industries to support their infras buildout. Can you imagine several industries running at full capacity for extended years? You can already see it happening in power, memory, and fabs.

It’s a career-defining move; even if capital allocators think that AI is a bubble that will one day burst, can they really afford to stay sidelined now?

Would highly recommend watching the full video

https://x.com/a16z/status/1975976064464892120?s=46

$INTC embargo from ITT 2025 is lifted today.. will be interesting. Lots of unverified rumours (both good and bad) circulating already; tbh I have no idea what’s going to be released, but will be keeping an eye out on deets mentioned above.

Not much news about Panther Lake clock speed, but seems like 18A yields are improving and on track. Not a bad sign for 18AP and 14A as well.

I view today more as a clearing event more so than the merit of the metrics itself.

$INTC

Added fresh $UAMY long @ $10.51

As the separation of the U.S./China value chain continues, rare earth/strategic minerals will become more important. I can see the US govt taking more stakes in US rare earth producers beyond $MP.

Antimony is used for munition and other military applications, and China produces ~80% of it. UAMY is the only significant pure-play and vertically integrated antimony company.

Spent some time digesting the $INTC news today. As expected, performance gains were announced for Panther Lake versus prior-gen chips, driven by 18A’s GAA and BSPDN technology. This was already anticipated, so while it’s welcomed, it’s not exactly material.

The most incremental takeaway, in my view, is that Fab 52 is now fully operational and set to enter HVM in Q4’25, with broad availability in January 2026.

This helps derisk some of the concerns about potential 18A production delays into 2026, which is materially incremental imo. It also means Intel is now producing the most advanced leading-edge node fully manufactured in the U.S. TSMC Arizona currently produces N4, while 18A is already competitive with N3. Obviously, that carries strong political value (maybe Trump even congratulates Intel with a tweet — who knows).

There wasn’t much concrete detail on yields, so it’s hard to analyze further. Realistically, while there’s some positive progress, the road ahead in catching up to TSMC’s technology remains very tough. It will require perfect execution, a deep cultural transformation, taking calculated but necessary leaps of faith that pays off (e.g., moving first to high-NA), and of course, industry and USG support.

TSMC’s moat extends well beyond lithography expertise. It is also deeply rooted in their design enablement and co-optimization capabilities with customers. Intel, on the other hand, has never served external customers at scale, which underscores the magnitude of the cultural shift required to make IFS succeed.

That said, revival is not impossible. Incremental datapoints and news over the coming months will reveal how likely IFS’s revival is.

Over the course of this week, there’s been a subtle escalation in U.S.-China hostility, especially from China’s side. Rare earth export bans, enforcing stricter checks on NVDA chip imports, implementing special port fees on U.S. ships.. I’m guessing trade talks are not going fantastic atm and the Chinese want some leverage.

If it worsens, $BABA might chop around for a short while.. I view $UAMY and $INTC as indirect hedges to escalating tensions.

Lol, called it. Should have derisked. $UAMY is reacting positively to the escalating tensions as expected. $INTC also holding in there, for now.

And totally forgot about our $FIVE short also being an indirect hedge.

What a sell-off. I’m de-risking low-conviction ideas for the weekend and keeping some ammo for next week. If you want to panic, be the first one out the door.

Tho, I don’t think markets will crater that badly over the next few months, unlike April, now that markets know about TACO Trump. Could be famous last words, idk.