# AI capex — X 热门讨论 (2026-09-29 23:30 UTC)
## @GrindeOptions (Cole Grinde) · 09-29 21:16 · ♥36 ↻3 💬10 Remember folks, $TSLA will be spending $30 billion per year on CapEx.
Where the money is going (management’s stated priorities):
• AI compute / data centers and training infrastructure • Robotaxi / Cybercab fleet and dedicated production • Optimus manufacturing capacity (including converting Fremont S/X lines) • Semiconductor fab (Terafab JV with SpaceX) • Solar manufacturing • Battery / lithium supply chain (refinery, LFP Megapack lines) • Broader auto capacity, Superchargers, service, and retail
This new $30 billion credit facility will help bridge the gap between funds from operations and cash on hand.
It’s GOOD debt and not BAD debt as it makes it easier to fund expenditures and they’re able to do it in a shareholder friendly manner. https://x.com/GrindeOptions/status/2105044019361501557
## @botblastcap (botblastcap) · 09-29 10:31 · ♥35 ↻4 💬5 what interested me the most on this call was the part on Venice's owned data center.
based on my understanding so far, Erik & co are trying to move from renting/leasing inference to owning GPUs and data center capacity.
basically, this means the same product (private, uncensored AI) for Venice as business but with a different cost stack. frontloading CapEx would lower unit costs and more provide way more control later instead of paying someone else’s rack forever.
this enables a few things: 1. lower long-run inference unit costs 2. better gross margins once owned boxes are utilized 3. less rental/lease exposure when GPU markets tighten 4. more control over what runs where
margins ↑ → free cash flow ↑ → room for discretionary buy + burns and/or higher programmatic burns.
if they execute, this is extremely bullish both Venice and base:0xacfe6019ed1a7dc6f7b508c02d1b04ec88cc21bf
oh, and it's still at ~$1.3b mc. > 引用 @Bankless: September $VVV Livestream https://t.co/INyuljWu8G https://x.com/botblastcap/status/2104881699943559411
## @VanquishTrader (VanquishTrader) · 09-29 18:41 · ♥36 ↻2 💬1 WHY I THINK MUSE COULD TAKE $META TO $5T
The biggest change in my Meta thesis is that Muse gives the company a path to become much more than an advertising business. Meta already has billions of users, hundreds of millions of businesses and one of the best distribution networks on the planet, so it does not need to spend years finding customers for a new AI product. Muse can be pushed directly into Instagram, WhatsApp, Facebook, Messenger and eventually glasses, which means Meta can distribute a personal agent at a scale almost nobody else can match while the core ad engine keeps compounding underneath it.
The monetization opportunity is what makes the $5T path interesting to me. Zuckerberg has already said Meta expects to make Muse free for a huge number of tokens and eventually take “a small fee from transactions,” which moves Meta closer to the moment where it can participate in the actual commerce happening after discovery. Subscriptions are the easy first layer, but commerce is much larger because Muse can increasingly help users shop, book travel, order food and pay across services through connectors like $SHOP, $PYPL, $EXPE and others. A tiny take rate across enormous transaction volume can become a very large business without Meta needing to replace advertising to get there.
Muse also strengthens the businesses Meta already owns. For consumers, glasses and Charm give the agent more ways to stay with you throughout the day instead of living inside another app. For businesses, Meta Enterprise Platform, Business Agent, Muse API and Muse Code create another monetization layer on top of the same models and infrastructure. That is what I think the market is still beginning to understand because the same AI spend can improve ads, power a consumer agent, support commerce, create enterprise products and eventually monetize new hardware. Meta is no longer spending all this money for one AI use case.
That is why I think Muse could ultimately be the product that takes Meta toward a $5T valuation. The core advertising business already gives Meta an enormous earnings base, but Muse dramatically expands what those billions of users and businesses can be monetized through. The biggest risk is still execution because commerce, enterprise and hardware all need to prove they can become real businesses, but the setup is very different now. I used to look at Meta’s AI capex mainly as spending required to protect the ad engine. I increasingly see it as the infrastructure behind an entirely new consumer and enterprise platform, and if Muse becomes the layer people use before they shop, communicate, work or transact, $5T starts becoming much easier for me to imagine. https://x.com/VanquishTrader/status/2105005084363174141
## @MrBrondorDeFi (Mr Brondor) · 09-29 05:22 · ♥31 ↻3 💬0 AI SAID SLOW DOWN
ethereum:0x4a220e6096b25eadb88358cb44068a3248254675 SAID SAME THING WITH THE CHART
QNT went from ~$70 to a spike near $350–375 in a few days after The Clearing House picked Quant for tokenized deposits.
$2T a day rails. 25 banks.
Then profit taking hit.
Down ~30% off the local high, now floating around $250
Same week the frontier labs started talking about pacing the race.
Hyperscalers still staring at ~$700B of 2026 capex while cash flow gets chewed.
Essays about slowing down. Markets already doing it.
Nobody cancelled the deal.
Rollout is still H1 2027.
The plumbing didn’t vanish because one candle took a nap.
The race paused.
That’s what races do after a vertical.
Nfa-dyor -Brondor https://x.com/MrBrondorDeFi/status/2104804072356303345