Summerβs over π
11.10.2025 14:12 β π 7 π 0 π¬ 0 π 0@vkmacro.bsky.social
Summerβs over π
11.10.2025 14:12 β π 7 π 0 π¬ 0 π 0Oh and Iβm not even a London doomer (since I love coming back whenever I do)
11.10.2025 10:13 β π 2 π 0 π¬ 0 π 0Donβt live in Dubai, and I also donβt think I could, but Abu Dhabi is pretty great.
There are genuine pros and cons of both even putting aside the money.
Sadly Iβm not a PC guy but keen to learn more
10.10.2025 16:43 β π 1 π 0 π¬ 0 π 0Whatβs the best thing to read on the BDC issues/ drawdown?
10.10.2025 16:43 β π 2 π 0 π¬ 1 π 0βWe dont need risk managementβ or similar was the jist of the conversation
09.10.2025 08:38 β π 4 π 0 π¬ 2 π 0Dude thinks heβs Buffett - Iβve met him
09.10.2025 08:34 β π 4 π 0 π¬ 1 π 0I know itβs early but Iβm most excited for Amazonβs earnings this quarter.
- We get a sense of rebounding (or not) retail revenues including margins
- How anthropic is boosting AWS revenues
- Outlook for their new Tranium chips - which Semianalysis calls Amazon basics lol
Such an egg head
08.10.2025 15:48 β π 8 π 0 π¬ 0 π 0Yep fair point!
08.10.2025 11:00 β π 1 π 0 π¬ 0 π 0His history at Tesla tells us he will absolutely go all in with debt etc when needed so no reason for him to change tact here I suppose.
08.10.2025 10:58 β π 0 π 0 π¬ 0 π 0So openAI gets 100bn from NVDA while xAI and Elon gets 2bn as part of a follow on round plus has to take out 12.5bn as part of an SPV?
Doesnβt sound great for him
Thanks looking toward to this one
07.10.2025 17:53 β π 1 π 0 π¬ 0 π 0I am also doubting the revenues. We donβt know how itβs attributed
07.10.2025 15:58 β π 1 π 0 π¬ 0 π 0How are oracle generating revenues on gb200βs if most of their new DCs are being built?
07.10.2025 15:34 β π 2 π 0 π¬ 0 π 0They literally included deprecation ππππππ
07.10.2025 15:30 β π 1 π 0 π¬ 3 π 0This makes 0 sense.
Coreweave gross margins are 75%. Semianalysis estimates Oracleβs GMs at 80%.
All costs other than housing and electricity occur below the line.
Coreweaveβs excess capacity was rented out by Meta.
People forget how long it takes to bring online large infra projects.
Largest 1GW DCs wonβt be operational until H2 2026.
I think extreme shortage in all things infra is likely until then
This is exactly right
07.10.2025 13:34 β π 2 π 0 π¬ 0 π 0Can imagine!
I think the HF life means you get caught up in a lot of noise, but can often miss the big picture. Also this is a better way to live life when running money
In this industry having time to think and slowing is so important.
Itβs easy to go from one project to another, produce a report with some implications and then move on.
Adam Tooze doesnβt once acknowledge UK core inflation has settled around 3%, nor that low rates failed to boost investment pre-pandemic
07.10.2025 08:31 β π 8 π 1 π¬ 2 π 1Also who funds Elon now?
06.10.2025 18:10 β π 4 π 0 π¬ 0 π 0Last point - maybe 2027 will be us reading articles about alll the unfinished DCs that have been abandoned in the US al la Chinaβs ghost cities
06.10.2025 18:08 β π 5 π 2 π¬ 1 π 0If that doesnβt tell you about market dominance in tech then I donβt know what does!
06.10.2025 17:37 β π 3 π 0 π¬ 0 π 0We need more on how NVDA was able to get openAI equity while AMD got fleeced just so OpenAI tries their chips
06.10.2025 17:35 β π 4 π 0 π¬ 3 π 0The way this gets crazier is if Hyperscalers also decide to tap debt markets or issue equity to fund capex.
If this is really a race to some elusive crazy end goal then Meta and Google are most likely to go all in. Meta currently behind and hit post Sora release
Msft and Amazon more conservative
Thereβs also a potential doom loop from the unwind whatβs been a virtuous cycle in:
Higher stocks > more investment > higher stocks > more consumption > higher stocks
We have transitioned from robust foundations in Hyperscalers FCF to obscure promises of financing unlikely never materialise. What pool of capital exists to fund this?
While this may be the start, itβs a very large unstable layer now added to the complex that ensures collapse when things stop
AI demand can be exponential. Video is highly compute intensive. Doesnβt mean everyone should do it for free.
Same with Deep Research, Agents or whatever.
The Social Network analogy also falls over since the marginal costs of service were much smaller. Marginal cost is real right now.