pre mkt thoughts 22 may 26

US yields continued their chop with a downward drift with 30y at 5.08 and 10y at 4.55 - roughly a parallel move of 10bps downwards. US equities are strong for the day with ES at 7480, nearing the previous highs of 7500s. In asian trading KOSPI is almost flat.
Given the strong performance of $SNDK in US hours yesterday, 285A Kioxia has also traded a touch firmer up at 57k JPY - for context this is the highest range that Kioxia has traded at. There are also more news coming out of workarounds from the high memory prices - notably using optical interconnects to move the GPU and HBM further away and allowing more HBM to be installed today. This comes as HBM hits the constraint from vertical stacking in a confined stack. Notable because the process difficulty goes exponentially from 20 stacks and beyond. This would decrease the difficulty in HBM manufacturing AND also allow lower tier players a foothold into the market.
Prior to this, in the note i wrote a few days ago - $NVDA was subtly warning for memory makers to rein in their greed, otherwise they would be forced to find solutions around it. This is what is happening. Human ingenuity resolving bottlenecks as they get too costly. - IMO, the photonic segment was lagging behind memory on the recent recovery, we could see names like $MRVL, $LITE trading stronger on this. $SIVE, the retail serenity pick may also get a boost from it.
Next up, PCBs are getting increasing attention from the BoM that MS shared. A few names like Victory Giant (2476 HK) traded strongly on that (up 16%). IMO this space is undercovered and we could see stronger performance in time. For context 2476 HK is a CORE tier 1 supplier for NVDA and has ~13% of mkt share for high performance PCBs. It is also a recent HK IPO - float is likely to be lower than other players and could easily see a squeeze.
In other news, we have $MSFT cancelling its internal licenses for Claude due to "escalating token expenses" and shifting teams towards it own Copilot. While personally i see this as a strategic move to funnel more revenues to their own internal models, one potential outcome might be the market realizing that the actual cost of running these models at scale is FAR more than the flat rate revenues.
This opens up 2 possible pathways, 1 - either corporates scale back AI usage to fit their token budget (BUT this will severely curtail the revenue ramp for the frontier labs) or 2 - the frontier labs eat the L to continue to show the revenue growth. However this will make the unit economics worse.
Broadly, AI is a miraculous tool no doubt, however much of the magic happens under a 5K/month subscription bill, not $200/ month. I see this as a potential blow that can hit the semi build out at some particular time.
Some readers also asked about KR. For Korea - we know there is huge levered positioning in markets here - however blindly shorting may not be a good approach unless you have other longs to buffer potential PnL variance. IMO, the easier play for KR is to buy the leveraged wipe out or to only short once the wobbles become apparent. The mental model i have for this is XAU in Jan.
For SPX, the market broadly was held up by semis. Excluding them, SPX would actually be down for the year (shocking?) - IMO the key takeaway should not be, ah lets not play in this sector. For now, it is the only game in town. However, this has lead to very interesting opportunities. For example, the cost of hedging downside exposure has never been cheaper, especially as downside skew gets repriced cheaper and upside gets bid.
Seasonality wise, interesting to note that the friday before memorial day tends to deliver strong positive performance. Dont be too shocked if we get a steady upwards drift tonight!
I also want to explain a little on hedges for those who asked what can be a good hedge. Let's start off with what leopold did in his filing (dated i know but hugely instructive) - First $NVDA and $SMH can be a good representation of the sector, however do note that 1. $NVDA as a compounder would mean a upward drift over time to a "reasonable valuation" and $SMH as a broad based basket will have upward drift just via rebalancing even. For example, at one point $MU was the largest exposure in $SMH earlier in the month, this has been rebalanced down. If semis as a sector goes up, $SMH will capture the upside.
The other approach of shorting "dogs" - is that in a bull market, it takes very little for dogs to reprice. A core example is "QCOM" - this was languishing around 135, until news of its "attempted pivot" managed to almost spark a 2x repricing.
IMO - if you are looking for a hedge, best to eat the upwards drift, at least you dont run an overnight hedge blow up.
Good luck!