Be ready to see a cambrian explosion of verticalized apps on web3 rails 🏗️
Asia birthed giant super apps like WeChat and Grab, but the next wave of winners will initially wedge into niches 🎯 and scale vertically through DeFi 💰 + verifiable data 📊. Let me explain why 🧵
CoreWeave got lucky
They were in a tight spot. Rates are elevated, capital markets are cautious on neoclouds, and their unsecured bonds are yielding mid-teens. Issuing new debt would be expensive
But strong compute demand bailed them out. According to @SemiAnalysis_, H100 rental rates rose 7% QoQ and B200 27%. With compute that tight, customers were willing to prepay on long-term contracts to lock-in capacity
Though it's a balancing act between short-term vs long-term contracts. Long-term contracts bring stability but short-term contracts bring the margin as they rent for ~2x currently
$CRWV likely gave up some price to get those customer prepayments. On yesterday's Q3 earnings call, they mentioned:
"We take an economic approach to prepayments, weighing the impact to pricing against the cost of financing."
And fortunately giving up some margin right now can cost less than issuing new mid-teens debt
They've used prepayments to solve near-term liquidity but the pressure isn't gone. It just shifts the obligations to their delivery schedule which 2/3rds of the prepayments are dependent on
The shift away from financing GPU clusters with offtake agreements to prepayments means that OCF to EBITDA is the most important metric.
Moves the diligence burden from private credit funds evaluating the creditworthiness of offtaker to the actual customer of the GPU making the prepayment.
Dock to live and historic cluster stability levels will be the most important metrics to customers who are increasingly paying nearly 100% upfront.
Also significantly improves the payback period on GPUs.
Looks like today may be a record day for token volume % of open models on Vercel AI Gateway:
🟦 Open 78.4% 🟨 Closed 21.6%
While spend 💲 usually tells a different story, #3 and #4 today are Moonshot AI & DeepSeek. Adding Z.ai, their combined spend surpasses OpenAI (#2).
(Do note that's the spend for inference of the model across providers (mostly in the US), not revenue going directly to the open weight labs.)
2/ Inside this Month:
• Open compute as self-custody for intelligence
• Open models closing the gap with frontier models
• Why the agentic bank run risk is about correlation, not speed
• Agents coordinating with cryptography instead of handing over our data
8/ Even with the risks, a v2 in this direction could gamify TCG supply and turn every collector into a source of gacha inventory 🃏
Dusty collections become productive
Normies touch web3 rails
TCG marketplaces get more supply
Who builds the drop-off + certification network?
7/ However there are some risks with the current mechanism as it may be capital inefficient with the ETH collateral requirement
And high-value cards face a russian roulette situation where they can get pulled quickly if the ETH backing is light
6/ This could be a strong onboarding moment for web3
It brings normies into web3 through something they already own, gives gacha platforms a steady stream of fresh inventory, and turns dead cardboard into yield-bearing positions
5/ DePIN principles could fit nicely for TCGs
1. Collectors drop cards at GameStop or local card shops
2. PSA/BGS/CGC grades them
3. Cards are vaulted, tokenized, and deposited in pools
Now any collector can become supply. TCG marketplaces reduce reliance on brokers/wholesalers
4/ Most card collectors have binders full of commons and mid-tier cards with basically zero secondary market 💤
Those cards just sit there collecting dust when they could be productive
Their binders are just unclaimed yield
3/ Currently, the FWA pool is one giant mixed pool of random NFTs + tokenized TCGs
The mixed pool works for v1, but v2 should verticalize the pools into categories 🗂️
Imagine separate gachas for Pokémon TCG, One Piece TCG, NBA cards, Pudgies, Miladys, Nouns, and general NFTs
2/ FWA is an on-chain gacha where anyone can be the house (or the player) 🎰
To be the house, deposit a supported NFT + ETH backing to fund a standing buyback bid if selected
Selection is inverse-weighted by ETH backing (i.e. more ETH backing = lower odds of being drawn)
FWA is one of the more interesting recent web3 experiments. h/t
@Rhynotic / @token_works for shipping $FWA
Since 7/20 launch:
6,200 NFTs
14.5K ETH (~$27M) volume
$10M fees
$2.8M revenue
Post-hype, last 7D still hit $502K fees & $260K rev
v1 proved it. v2 should expand IRL 🧵