# AI capex — X 热门讨论 (2026-09-29 07:07 UTC)
## @TheVictorBuilds (TheVictorBuilds) · 09-29 05:50 · ♥44 ↻16 💬40 THE AI BUBBLE IS ABOUT TO FORCE THE LARGEST CAPITAL ROTATION IN MODERN HISTORY
Bain dropped their seventh annual global tech report today and the numbers are borderline psychotic. They estimate the AI sector needs to generate $6 trillion in annual revenue by 2031 just to justify the sheer amout of capital currently being incinerated to build data centers globally. Think about that. $6 trillion. We arent talking about market cap, we are talking about actual hard revenue. Right now, existing consumer and enterprise AI applications are only pacing to hit maybe $1.8 trillion of that total. That leaves a gaping $4.2 trillion void. A void that requires entirely new markets to be invented from thin air to prevent a massive infrastructure collapse. Its a dangerous game of chicken with global liquidity. The market is blindly building physical data centers assuming the software demand will magically catch up.
Zoom out to the macro environment for a second. The analysts at Bain basically admitted that funding this infrastructure sustainably requires injecting roughly 1% of additional annual global GDP growth into the system. Where exactly is that money coming from? We are already watching central banks play a delicate balancing act with liquidity and rate cycles. You simply cant siphon trillions of dollars into physical concrete, steel and specialized chips without draining liquidity from other risk markets. And on the micro level, the centralized model is showing massive cracks. The Capex required to build these mega data centers is shutting out everyone except the top three tech monopolies. This is exactly why DePIN is no longer just a Web3 narrative but a structural necessity. We are seeing decentralized networks crowdsourcing idle GPU compute and localized power grids, using token mechanics to completely sidestep traditional debt and equity financing. When the centralized system hits a wall of capital inefficiency, distributed ledger systems become the natural release valve.
The actionable takeaway here is to fade the noise. The masses are blindly bidding up AI software wrappers and consumer apps fighting for scraps in that $1.8 trillion pool. The real edge lies in betting on the $4.2 trillion infrastructure gap. Capital always flows to the path of least resistance, and soon it will aggressively rotate out of overhyped centralized SaaS models and into decentralized physical infrastructure. Look heavily into decentralized GPU marketplaces, verifiable compute protocols, and tokenized energy grids. The bottleneck for AI is no longer the intelligence itself, but the physical constraints of power and hardware deployment. Position your portfolio in the base layer protocols that are actively decentralizing the Capex burden. Because if we actually need a wave of innovation that dwarfs the cloud era to break even, it wont be built in a closed ecosystem. It will be built onchain. > 引用 @TheVictorBuilds: COSTCO JUST HANDLED A $184 MILLION WINDFALL\
Costco reported their Q4 2026 numbers and dropped an absolute flex on the entire retail sector. After successfully challenging federal import tariffs they secured a massive $184M refund. That is just the first tranche of an expected $500M+ total recovery. But here is where it gets interesting. Instead of keeping that cash to pump their quarterly margins or aggressively buy back stock to please Wall Street insiders CEO Ron Vachris did what Costco does best. He immediately slashed prices across fresh produce meat home goods and hardware to give the value straight back to the shoppers. It’s a $0.15 per share non recurring earnings beat, but the real alpha is how they treat their community. They took a pure accounting windfall and turned it into tangible everyday relief at the checkout line right when consumer wallets are still feeling the heavy hangover of sticky inflation and high borrowing costs.
Let's break this down through a macro and Web3 lens. Costco is essentially a real world protocol operating on a highly optimized subscription staking model. The membership card is your access pass to their liquidity pool. When they recovered that $184M in MEV or treasury surplus they faced a distinct choice. Extract the value for passive holders or redistribute it to the active network participants. By choosing to lower prices and essentially cutting protocol fees they reinforced their ultimate flywheel. Lower costs mean higher volume which drives sticky membership renewals which secures their actual core revenue stream.
On a macro level as central banks juggle liquidity injections and rate cuts while main street still struggles Costco is actively capturing brutal market share from competitors who are too busy defending their short term margins. Contrast this exact dynamic with 90% of crypto protocols today. When a DeFi project generates excess revenue or a treasury surplus they almost always use it to buy and burn their own token to bail out early VC bags. It is pure value extraction. They bleed their actual daily active users with high fees just to sustain a synthetic chart. It’s entirely backwards.The market is violently rotating away from high extraction tokenomics toward real yield utility and sustainable user acquisition. If you want to find the next cycle winners stop looking at buy and burn mechanics. Start screning on chain data for protocols that operate exactly like Costco. Find the DEXs lending markets and L2 ecosystems that are aggressively lowering transaction costs and routing sequencer fees directly back to active liquidity providers and daily users.
When you reward the people actually interacting with your smart contracts you build sticky liquidity that does not flee the second yields drop by half a percent. Retention will aggresively outperform pure token emission games over the next 12-18 months. The incoming liquidity wars will be won by builders who realize that passing value to the user is the only unbreakable moat. https://x.com/TheVictorBuilds/status/2104810939598151767
## @growthrapidly (Joel) · 09-28 15:58 · ♥31 ↻3 💬4 $META is down today and trading around $718.
It’s taking a breather, which is completely normal after the run it just had.
Remember, this stock was trading around $545 in late August and recently crossed $750.
Muse has been a major driver of the recent surge, and the crazy part is that it launched less than three weeks ago and has already seen millions of downloads.
The bigger story, in my opinion, is 2027.
That’s when more of $META ‘s massive AI infrastructure buildout should be in place and investors should start getting a much clearer picture of the returns on all this CapEx spending.
I don’t think the biggest move has happened yet. https://x.com/growthrapidly/status/2104601648513376352
## @Kaizen_Investor (KaizenInvestor) · 09-28 09:27 · ♥31 ↻0 💬8 Some of my thoughts on the upcoming $MU earnings this Wednesday.
Top line expectation - $51,500 Bottom line - 31.6 EPS
There are 5 other structural drivers I'm watching closely
HBM yield progress & margin accretion The complex 12- and 16-high TSV stacking makes High-Bandwidth Memory exceptionally difficult to manufacture at scale. Because these dies command a massive premium, incremental yield improvements flow entirely to the bottom line. If yields scale efficiently, watch for forward EPS estimates to revise sharply higher.
Wafer cannibalization from packaging constraints Producing a single HBM wafer consumes roughly 3x the cleanroom capacity of standard DRAM due to die size and backend complexity. Surging AI demand is actively starving the conventional memory market of supply. Tracking this trade-off dictates whether the broader DRAM cycle stays structurally tight or loosens prematurely.
Pricing durability across conventional transitions While HBM dominates the narrative, traditional DDR5 and LPDDR5x still fund baseline operations. Amid concerns over softening spot prices, the durability of contract pricing will prove whether the HBM-induced supply squeeze is actually strong enough to override sluggish macro demand in the PC and smartphone segments.
Capex discipline vs. cleanroom limits The market inherently fears memory upcycles ending in undisciplined overbuilding. Capital expenditures must be scrutinized to ensure they are tethered strictly to multi-year, non-cancelable agreements rather than speculative capacity. Physical cleanroom construction bottlenecks act as a natural governor on supply growth, making long-term dynamics more predictable.
Customer inventory digestion pacing Beyond the AI data center, legacy markets are navigating uneven demand. If mobile and PC OEMs accumulated excess memory inventory to front-run earlier price hikes, an extended digestion period will stall new orders. The pace at which these specific channels clear dictates when baseline volume leverage returns.
The technical setup will react to the headline, but these underlying fundamentals dictate the multi-quarter trend. https://x.com/Kaizen_Investor/status/2104503138875027801