Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
QF Research Profile picture
QF Research
@ResearchQf
PM at large institutions over 20Y. Investments decisions are your responsibility. DYODD. Analyze distributionally but non-deterministically. RTs ≠ endorsements.
3 subscribers
Oct 20 • 4 tweets • 2 min read
1) Mainstream DRAM and NAND spot prices were up 10% last night alone. DDR5 spot just doubled. I've never seen this before.

$MU and SK Hynix almost doubled since local bearish peak for DRAM & HBM in Aug. Large parts of buy & sell-side as well as Fintwit.

Do math for "commodity" DRAM margins. Assume spot goes sideways and contract (which lags spot) ~ catches up. If 100 price/50 cost = 50% GM. 200 price/50 cost = 75% GM (!).

It's not surprising every time I turn around, any incremental price negotiations for HBM3e and HBM4 sound better. Have discussed varying fungibility timeframes between DDR and HBM as well as flash.

Peak EPS power looks a little nutty. Again, normalized earnings matter as much. I.e. when there is a future downturn, will trough EPS also be higher?Image 2) Then consider that this is happening in what just could be the very beginning of several AI drivers that are very storage & memory intensive.

Maybe not. But could this cycle also be unlike pre-AI cycles that investors reflexively pattern match to?
Sep 18 • 4 tweets • 3 min read
1) $COHR risk/reward improved IT, and I rotated a part of LITE back into COHR. It's still a smaller position.

$LITE almost tripled past year and outperformed COHR 2.2x. For good reasons. LITE was, and still is, better positioned across a range of current and future interconnect technologies & products.

As a result, LITE's forward earnings cons rose > 30%. COHR's forward cons was almost flat after last 2 Q's.

But based on data today, COHR numbers also have greater upside vs downside while valuation compressed. LITE also became much too big in the portfolio, but that's portfolio dependent.Image 2) A number of interconnect stocks have been explosive. So have stocks in other sectors discussed since spring. Cool to hotter storage, tier 2 & 3 infra & power capacity, physical AI, rare earths etc.

Not that classic AI infra leaders haven't done well. And my confidence has grown recently that chance of bigger $NVDA upside(s) by Jan Q is higher than normal, even as the stock lagged some recently.

A common denominator is high leverage (in a good way) to AI growth. It's interesting that the range of sectors and stocks that are, or will soon be, levered AI beneficiaries, or losers, is growing rapidly.

Not a surprise. AI is touching everything we do and has started to transform entire industries.Image
Sep 14 • 4 tweets • 3 min read
1) Remember that Feb mini-panic about data center oversupply? Here's a DC & DC supplier update. 10Q's are also must reads. And perhaps Larry does believe what he said last week. $ORCL is making commitments.

Here's one of the Cowen notes that triggered that scare. $MSFT cancelling leases "points to a potential oversupply position." AI trade was all over, again.

Mentioned at the time "what DC industry is seeing globally is the opposite." Cowen had taken -1 + 2 = +1, but only pointed out -1. And this was before the huge surge in inference, which is continuing.

What were MSFT & ORCL lease commitments in Q1?

MSFT. Commitments dropped from $117B (CQ2 2024 peak) to $99.2B (CQ1 2025).

ORCL. Commitments rose from $22.9B to $48.4B in same period (Feb Q ≈ CQ1). ORCL's gains alone more than offset MSFT, even though it was a smaller player.

And MSFT stated "the next spend will begin to shift back to short-lived assets, which are more correlated to revenue growth" on its earnings call Jan 29th.

At most, MSFT lease commitments might've overshot some (lease commitments had risen $74.9B in just one year to CQ2 2024 peak, or from $42.1B CQ2 2023 to $117B CQ2 2024!) while they shifted focus to filling data centers with AI infrastructure.

And finally, MSFT and ORCL, and most other CSPs, guided to significantly higher capex in latest calls.Image 2) What are the most recent numbers? ORCL's lease commitments shot up by $56.4B Q/Q or more than doubling Q/Q! This doesn't include $6.6B in further commitments 10 days into Nov Q.

Looking at ORCL & MSFT alone, data center lease commitments are up $55.8B, or by +41%, YTD.

I'd also take 50:50 odds that MSFT DC commitments starting to rise again Sep Q.Image
Sep 3 • 5 tweets • 3 min read
1) $NVDA is more likely have bigger upside(s) than usual by Jan Q.

Based on a pretty broad set of data points. It's also not impossible that Oct Q could also be quite decent. Given NVL run rate by Q4, one week in timing could mean over +/- $3B.

But timing won't matter if NVDA DC rev has larger than usual midpoint for combined Oct + Jan Q vs $101B DC cons for that period.

This is independent bottoms up & top down build. I.e. most sell side models look more conservative than typical. There's always a spread, could be wrong, but my midpoint is higher than normal vs expectations. Subject to quick change with.

More importantly, 2026 outlook and even beyond is still improving due to variety of demand drivers. AI top down is accelerating is still crazy.Image 2) One recent data point.

Few believe/model HH +300% Q/Q. HH reiterated +300% during JPM Asia tech tour days ago with 1M left in Q3. Upside risk vs models?

"Temporary" small ODM issues were Quanta & Wistron. Order strength. Keep in mind other suppliers.
Aug 12 • 4 tweets • 2 min read
1) $NBIS and $APLD now outperformed > 2x despite $CRWV bouncing. AI inference spikes overflowed to CRWV and then even 3rd tier such as APLD. NBIS is another +40% since 7/18.

NBIS raised year end ARR from $700M-$1B to $900M-$1.1B. Despite higher ARR and sales more than doubling Q/Q, NBIS didn't raise rev guide due to expected delivery of GPUs coming online by the end of the year, "particularly the Blackwell Ultras."

"Pricing is stable for Hoppers" and "we did bring on the B200s, and we are actively selling through them well."Image 2) Less overhang and "cheaper," yet still not profitable. But listen to the call and do a moderate bull case back of envelope. And consider secular environment

NBIS capacity target is 220 MW this year, with > 1 GW contracted capacity end of 2026. E.g. there is 200 MW under construction in NJ, part of which will come online this year.

Like APLD, it's a seller's market for anyone with future capacity.Image
Aug 8 • 6 tweets • 2 min read
1) Nice week for AI but software net exposure was at 5 year lows, and negative, near July end (GS PB).

Glad didn't post this then. We all wanna be contrarian.

Semi exposure was also at lows entering 2023. Remember tech (and general equity) sentiment then?

But that was right after ChatGPT moment, though market took a while recognizing its importance. Imo we're still in very early innings of understanding consequences of AI. Many are still trying to ignore it.Image 2) Semis to ratio is now at 21Y high?

Pretty chart. Is it breaking out again after consolidating?

This is only 20% serious. I've chart-mined, tho GICS do have long history. Semis include many non-AI semis. Many are down or barely up since GPT. Software has a big AI infra ... Image
Aug 8 • 6 tweets • 3 min read
1) Physical AI, rare earths & tariff macro, $MP & $OUST.

MP is up > 3.5x in ~ 2 months. That Sputnik moment went warp speed faster than one might've hoped. MP is the domestic rare earth champion.

Also rare earths = most important macro factor for at least a few weeks. Image 2) MP beat but current numbers don't mean much.

a) DoD stake and partnership guarantees $110/kg floor. That kicks in Q4. Could domestic & commercial and military demand be so high that > $110/kg is possible?

"Included in this commitment is some upside sharing with DOD if, as we suspect to occur over time, prices go materially above $110."

b) Initial magnet production target at Independence was 1000 MT. New 10x facility is targeting 10,000 MT, but that not sufficient for rare earth independence.

Worth listening to call and considering both risks & EPS power if MP executes.Image
Jun 27 • 4 tweets • 3 min read
1) $NVDA hit ATH yesterday. Other AI related names are at or near ATHs. Makes sense. AI demand & supply data keep improving almost daily within an exponential framework.

It's not impossible that CQ4 NVL racks may be ~ 12-14K vs ~ 6K in CQ2. Prior post in May was more a 20-80 guestimate. Q4 variance is still very large, but smaller than in May. Track data real time going forward. Subject to quick change.

That's with BW Ultra only starting to kick in. B300 production prob ~ 100K's in Q3, with nominal GB300 shipments (given lag to racks) but some GB300 volume CQ4, followed by bigger ramp in 2026.

More important LT. AI inference demand went into higher gear months ago, and that acceleration has continued. Those cute AI insider hints about next gen models & apps show also show no signs of slowing.

Also where did all the bear stories about NVDA B200 GPU inventories go? Remember to review NVDA rev rec, which is product dependent.Image 2) Jensen's high level TAM comments at yesterday's annual meeting was nice but they're hard to distinguish from previous comments. What are durations of these TAMs and shape of deployments?

- Robots and AI infrastructure to train them will be the next multi-trillion $ industry

- 100 NVIDIA AI-powered AI factories, buildouts around the world. That's double a year ago, and they're larger. Average number of GPUs/factory has also doubled and requires tens of GW of AI infrastructure in coming years.

- Robots and AI infrastructure to train them will be the next multi-trillion $ industry

- 100 NVIDIA AI-powered AI factories, buildouts around the world. That's double a year ago, and they're larger. Average number of GPUs/factory has also doubled and requires tens of GW of AI infrastructure in coming years.Image
Jun 1 • 8 tweets • 5 min read
1) Rare earths. Trade, China and $MP

"If China doesn't speed up those approvals, companies have warned the White House, auto plants may have to idle pandemic-style stoppages." (WSJ)

Rare earths were more important for initial trade detente with China than some realized.

Chinese rare earth controls are similar in nature to US tech export controls. But why are mostly unknown elements so important?Image
Image 2) Rare earths are necessary for huge ranges of industrial, tech, automotive, robotic, defense (e.g. drones, missiles, fighters) and consumer products.

Light (LREE) & heavy (HREE) rare earths elements aren't always rare, but even when mined elsewhere, Chinese companies refine > 90% for critical markets.

E.g. NdFeB magnets are very strong, high perf and necessary for key electromechanical products, such as EVs.

US has some mining but *zero* refining capacity of such magnets. MP is the only US company that mines rare earths in volume.

MP defines 3 stages. Stage 1 ~ upstream REO mining. Stage 2 ~ midstream NdPr oxides. Stage 3 ~ downstream or *magnets*

MP mined 12K metric tons (MT) of REO in Q1, which is ok but much smaller than China's capacity. Sales were lower. MP stopped shipping to Chinese refiners due to tariffs and is trying to drive midstream, stockpiling, diversification and longer term vertical integration thru Stage 3.Image
May 26 • 5 tweets • 4 min read
1) Storage stocks (components) reacted ~ as expected Fri.

$STX $WDC (mostly DC) up Fri despite EU & $AAPL threats. Both up big YTD.

$MU & $SNDK (large CE exposure) down Fri despite strong ST data points. Hard to believe but MU is +11% YTD. Again, fears of pull-in and net demand destruction from unknown future electronics tariffs, such as 232.

High level anecdote. I have an AI output folder. Folder size is growing exponentially. Text was tiny. Multimodal i/o such as document formats, audio e.g. NotebookLM, images, video = storage rose from tiny -> 100's KB -> MB's -> 10's MB -> 100 MB or more/file. More later in thread.

This also needs inference & storage on AI cloud side.

I've never ever never seen spot DRAM (NAND-related spot, card, SSD prices are also up but not as much) do this without some stocks rising sharply.Image 2) So stocks are saying "this isn't a real cycle." DDR4 will be legacy soon but still.

Is the cycle 100% non-organic? There is some pull-in, but it's unclear how dramatic. May & June PC & SP data will help. Net demand destruction still depends on "policy" i.e. a megalomaniac.

But are there organic elements too? Edge & DC AI both help (let's leave out HBM for now). PCs are being futureproofed for edge AI and Windows 10 expiration with higher configs. Rumors of higher DRAM content in future iPhones & SPs due to AI do make sense.Image
Feb 23 • 10 tweets • 7 min read
1) $MSFT DC scare 🧵. High level with details if have time. Some know all datapoints mentioned. Or more (if you have non-MNPI data pls DM me).

a) Conclusion 1st. Global AI capex next few years went UP recently. Even MSFT.

Is this just the latest AI scare? Again, Trump & co finally began scaring investors. Have pet theories about scenarios but a different thread.

The usual discussion about LT AI variability later in 🧵. Even Satya, Sam, Ilya, Jensen, Dario, Jeff etc don't know > 5-10 year consequences of AI. I have no idea. There's never been a tech cycle like this in human history.

But let's start with MSFT Dec Q & 10Q. Keep in mind, that as of 1/29, MSFT already embedded diffusion rule, DS, @OpenAI, Stargate, Cowen "news" and prob much more we don't know.

- Commercial bookings grew 75% Y/Y (!) but guided flat Y/Y Mar Q. Watch OAI commitments going forward

- Capex ~ flat Q/Q Mar & Jun but rises again F26, which is some info. Lower growth rate? No real info, since range is huge

- Have discussed leases being much larger % of MSFT capex last few years. It helps to analyze lease data in 10Q/K's

- "Expect capex to increase in coming years" in 10Q/K's have been stated forever and is a general statement. Has occurred before down capex years

- MSFT clearly said spend shifts back to "short-lived assets (i.e. infra) which are more correlated to sales growth. That's real infoImage 2) Next elements of Cowen note. International first.

Channel checks "indicate MSFT is also re-allocating a considerable portion of intl spend to the US, which suggests to us a material slowdown in intl leasing."

b) Are we sure?

Mentioned "pre-diffusion" for APAC capacity plans on purpose. This is one estimate (Citi) of APAC ex-China DC capacity. Only months old but it's outdated. Possible new projected would've created GWs of spikes.

Japan, Australia, Korea are Tier 1. Consider Tier 2. India, Malaysia, Indonesia etc. Comb thru diffusion rules. E.g. Malaysia was already higher (but pre-diffusion). Well over 3 GW is nowhere near happening with low GPU limits.

Who's who of hyperscalers and DC developers are involved in some projects. Could this be why capex is being reallocated back to the US or other Tier 1's?Image
Dec 20, 2024 • 6 tweets • 3 min read
1) Robots are making huge strides. Mechanical multiplied by AI 3D vision, physical models etc.

It's only one AI adjacency but its impact alone is enormous coming years/decades.

Imho this alone could be by far largest macroecon driver yet it's being ...
citigroup.com/global/insight…Image 2) generally ignored.

AI eyerolling has faded. Hmm maybe this is real. But with exception of few forward thinkers, macro analysts seem want whole thing to just go away.

Look at Citi report in 1). It's public. Payback could be 29W at $7.25/hour or ~ my earlier back of envelope. Image
Nov 14, 2024 • 4 tweets • 3 min read
1) Holy crap this is gen2 like $GOOGL NotebookLM I've wanted.

Already copy paste and analysing financial data, charts, AI papers. Grad student level. I'm talking to the damn thing and asking questions.

A famous paper by John Bell that ended up disproving Einstein's attempt ... Image
Image
Image 2) to frame an objective reality with his EPR paper (QM is correct but are there hidden variables?).

Bell's inequality was demonstrated by '22 Nobel Laureates. Almost a "what is reality" Nobel. So this sometimes shows up in pop science "what is AI & consciousness" discussions. Image
Image
Sep 15, 2024 • 16 tweets • 7 min read
1) Reasoning/inference and domain knowledge/ability stood out first right?

An interesting but less explored direction? A long thread but worth it. Conclusion first:

AI is on verge of (or at) design & engineering across major industries. Or medical diagnosis etc.

Coding/SW ...
Image 2) Ph.D. STEM problems, agents etc skills are better known and amazing, but here is just one example that combines reasoning, expertise, application (interconnects will be huge + also dear in my portfolio) in a way I haven't seen.

I'll compare Claude later. It's night and day.
Aug 28, 2024 • 5 tweets • 2 min read
1) Fate of markets depends on $1B or $2B difference in $NVDA guide.

There are tons of whisper surveys. Some game them. A bear might answer too high. Or a bull answers too low.

But it seems Asian investors do have higher expectations than US investors. Will Oct guide really ... 2) matter?

After hours prob. Or days or even weeks. But unless BW cycle or LT AI growth outlooks change, is it ultimately that imporant?

Not sure I'd present this framework. Implies $43B upside case Jan Q vs $31B cons (DC)! Unless something drastic ...
Jul 8, 2024 • 6 tweets • 2 min read
1) My 2¢ on "AI capex cycle has to pause soon right?" part X.

Even as an AI bull, this is a discussion I had to have with myself and people on and off this site. As soon as right after $NVDA Apr '23 Q blowout.

Conclusion 1st. Outlook improved after every broad overview of data. 2) And that wasn't about next few Q. If cyclical indicators were peaking, would be short today even if thot beat & raise next Q. But even defining cyclicality (inv, B:B, consumption vs shipments etc) needs change for AI. Much more later.

Little doubt ...
May 14, 2024 • 4 tweets • 3 min read
1) Maybe $MP did bottom. May buy more if close at this level.

One of many examples with potentially huge secular story. And is cyclicality reversing to tailwind?

Cyclical. Rare earth prices still edging up from Mar lows after -70% from '22 high (EVs renewables industrial etc).
Image 2) Now the interesting part. There's a chance MP is a robotic+ beneficiary.

What's Neodymium used for? Many applications e.g. Nd:YAG pump lasers. Nd has 1064 nm transition in Yttrium crystal.

More importantly Nd magnets are strongest rare earth magnets. Super important for ... Image
May 11, 2024 • 5 tweets • 2 min read
1) $AAPL has closed in on an agreement with @OpenAI to use the startup's technology on the iPhone, part of a broader push to bring AI features to its devices (BN).

Via @theinformation: OpenAI sees assistants with voice and visual capabilities as potentially transformative as ...
Image 2) the smartphone.

"The technology is another step in @sama's quest to ultimately develop a highly responsive AI akin to the virtual assistant in the film "Her," and enable existing voice assistants like Apple's Siri to be more useful."

OpenAI may preview upgraded AI "at an ... Image
Jan 25, 2023 • 5 tweets • 2 min read
1) $324M of $3.9B of $TSLA OI was FSD recognition.

Most didn't model. +1.5% GM impact. And there goes EPS beat vs "company compiled" consensus.

Inventory $12.8B vs $10.3B Q3 and $5.8B Q4 '21. Big jump Q/Q and huge jump Y/Y in days.

Many ...

$SPY $QQQ $TLT $GLD #Commodities 2) more details ... forward comments usually don't matter much. Especially if Elon says bullish things (just calibrate his past comments).

If he concedes any downside metrics? Pay attention.

Independent analysis with incoming data by far most relevant.
Jan 25, 2023 • 5 tweets • 3 min read
1) Total global ETF holdings of #gold has gone sideways despite gold +19% since early Nov.

System akin to compressed spring. Less pressure?
Spring bounces.

Δ largely $GLD driven. What if US investors traders actually buy?

1 more chart today.

$SPY $QQQ $TLT $GDX #Commodities 2) Cost of $SLV puts vs calls (3M 25 delta) 10D MA not budging despite #silver +33% in several M.

Incredible (to me). And never happened in history of this data.

Whenever silver rose as much since '15 secular low, cost of calls shifted ~8-10 (implied ...
Jan 24, 2023 • 7 tweets • 3 min read
1) Largest day of $SLV inflows (+20M oz) by far since #silver squeeze!

Positioning was more bearish despite #gold +19% #silver +31% in few months.

Little buying power was used. If anything there was large selling vs price reference points ...

$SPY $QQQ $TLT $GLD #Commodities 2) over last few Y, as well as variety of unusual technical readings that just don't occur after significant rallies.

"Yes there WILL be pullbacks, but I've mumbled could be shorter and shallower than one might usually expect."

Data implies unusually ...