Nov 10 we called out Fed is NOT going to HIKE in Dec when market had priced 84% chances...apparently the conviction stem from insane herding trade taking place in the market pre-emptively
2
106
J Cohen retweeted
$NBIS
U.S. Suicide Prevention & Mental Health Resources:
988 Suicide & Crisis Lifeline – Call or text 988 (Free, 24/7)
Website: 988lifeline.org
Crisis Text Line – Text HELLO to 741741 (Free, 24/7)
Website: crisistextline.org
National Alliance on Mental Illness (NAMI) – 1-800-950-NAMI (6264) Monday–Friday, 10 a.m. – 10 p.m. ET
Website: nami.org/help
Fed/Warsh needs to stfu completely in its upcoming dec meeting since infinite fedspeaks have already caused market wide fire-selling
4
131
J Cohen retweeted
BESSENT: SEEMS PERIPATETIC FOR FED TO WANT A BIGGER HIKE -CNBC INTERVIEW
BESSENT: FED SHOULD LOOK FORWARD NOT BACKWARDS
J Cohen retweeted
Andrew Garfield, the Oscar is yours
Artificial - Official Trailer - In Select Theaters Christmas
J Cohen retweeted
🚨$NVDA cuts neocloud support plans down $1B q/q and adds tighter language around conditions; removes revenue share language from last quarter; adds new language that certain commitments "could be reduced or eliminated" if contingencies are not satisfactorily resolved. Cuts previously disclosed 2028 capex $1B to zero (sounds like a project pushout or cancellation).
Now says commitments take effect "only if the neoclouds secure financing to purchase the infrastructure products and the related compute capacity is deployed, made operational, and accepted by us."
Previous language: "We continue to make commitments....Under these agreements we will earn revenue on upfront sale of our infra and if certain criteria are met, we will participate in revenue share."
Re: secondary equity investments, future commitments fall from $25B to $20B but 2028 plans rise from $3B --> $11B. Unclear if this is pushouts from 2027 investments not made (such as Figure AI).
Bottom line: Total commitments grew 41% QoQ, driven almost entirely by a surge in 2028 supply/capacity commitments and long-dated data center leases, but the new contingency clause means a large portion is conditional on neoclouds securing external financing.
Also 10-Q to arrive "in coming weeks" seems like a delay.
J Cohen retweeted
The people who have made themselves very rich by managing others' money very badly prove the adage:
the only truly necessary skill of a money manager is sales.
J Cohen retweeted
SemiAnalysis Short Report
An industry wide Capex reduction would be damaging to AI stocks so it's worth my time to read through the more notable short reports. SemiAnalysis does have great reports for the most part which is why they get paid for their hit pieces. Hit pieces only work if most of the firm's other work is reputable.
SA - Oversupply
SemiAnalysis is projecting a large shortfall in AI demand due to:
1. A slowdown in frontier labs growth.
2. Tightening capital markets.
3. US election results pushing out DC supply from 2H27 to 1H208.
4. Reduced Nvidia support of their ecosystem of AI startups and Neoclouds.
SA - Semiconductors
According to SA, the hardest hit will be HBM and Nvidia in 2029 when Feynman Ultra prices will decline to a astonishing tune of -44.2% reflecting loss margins from memory makers and Nvidia.
Broadcom will be the least hit since they have less margins to begin with and have the lower TCO. MediaTek will be hit slightly harder as they do not have the vendor financing capabilities of Broadcom.
AMD will be moderately because they didn't have as much extra margin to give as Nvidia. Amazon Trainiums will be fine because they have a very strong data center pipeline that will not face delays.
Power demand will go down as with coming in 2029 and 2030.
SA - Neoclouds
SA states Neoclouds are a far higher risk because of execution and ability to finance. SA points out there is a sizable amount of Neocloud offtake by Chinese end customers and that these are utterly done for.
Jim's Reflections - Semiconductors
I do not have the data to support or refute SemiAnalysis's exact claims but I do see the risk that Nvidia and memory makers 75% and 85% margins are very difficult to sustain.
Not called out by SA but SNDK is at a higher risk because storage is something China can actually produce at high yields and flood the market, however SNDK does not have the sky high market caps of MU, SKHY to begin with.
I have avoided in investing in semiconductor because they have to keep re-selling year after year while Neoclouds can lock in 3-5 year contracts. LTA's only account for 35% of MU's revenue through 2030 but most of IREN's GPUs are contracted for the GPU payoff time or more.
Semiconductors revenue is the first derivative of AI demand while Clouds and Neoclouds are the instantaneous demand in AI usage. AI usage may expand slower but it's unlikely to go down. This is why Cloud providers do not have cyclic ups and downs unlike semiconductor industry.
Jim's Reflection in Neoclouds
I see Neoclouds as a spread trade between GPU pricing and GPU rental pricing. This spread can widen due to power constraints but it seems that power as a limitation will be reduced in 2029.
I do not see open source inference runtimes as a moat because Chinese labs require a 30% revenue share (1). That's revenue share, not margin share. You might as well be supplying bare metal compute to Anthropic which doesn't have political risk.
The high prices you see today for Open Source Inference Tokens is due to the fact that Token APIs are by definition the shortest term duration contracts spanning literally tokens. When demand contracts, the shorter term contracts get hit the hardest because they have to renew at lower pricing but the GPUs were bought at higher pricing.
IREN has for the most part has locked in mid to long term duration contracts. It has even matched it's GPU purchases to contract signing and even held off on some GPU purchases because Dan see's that GPU prices may fall.
The other upside in Neoclouds is extended GPU lifetime and scaled operations at large sites. With this reduction in build, the chances that GPU lifetime is extended is increased.
Overall IREN can more than justify it's market cap from colocation and bare metal services and is one of the safest investments in the AI trade at this time besides SPCX, META, ANT.
J Cohen retweeted
Druckenmiller 13F summary:
- bought STX (and calls), NEM, APP, SPCX, PLTR,
- added AMZN, U, FOXA,
- reduced TSMC, STM, INSM, PANW
- dumped SNDK, EWZ, GOOGL, UAL






























