Nov 18: ARK Trims Nvidia While Buying Block, Circle and Coinbase
Wallets Linked to Binance Theft Move $136.8 Million in Ether · Coinbase Borrow Opens in New York With Crypto-Backed USDC Loans · OpenAI’s GPT-6.2 Models Compete on Price and Selected Benchmarks
Market Snapshot
Bitcoin traded at $64,080.01, down 1.512% over the preceding 24 hours, while ether stood at $1,795.23, down 1.551%. Solana diverged from both, rising 0.828% to $74.28.
Beyond spot prices, Coinbase Borrow has processed more than $3.4 billion in cumulative volume since its launch in January 2025, according to the supplied research. That figure measures historical borrowing activity, not the current amount outstanding; it provides context for the product’s expansion into New York.
Top Stories
- ARK Trims Nvidia While Buying Block, Circle and Coinbase
- Wallets Linked to Binance Theft Move $136.8 Million in Ether
- Coinbase Borrow Opens in New York With Crypto-Backed USDC Loans
- OpenAI’s GPT-6.2 Models Compete on Price and Selected Benchmarks
Key Takeaways
- ARK’s disclosed trades favor selected crypto-linked companies, but its remaining Nvidia holdings and staking ETF sales complicate any claim of a clean shift between sectors.
- Binance-linked wallet activity illustrates the gap between identifying stolen funds, restricting their movement and ultimately recovering assets.
- Coinbase’s New York launch extends access to onchain borrowing, while its cumulative volume says little by itself about current loan balances or credit performance.
- AI model competition increasingly involves both cost and task-specific performance; a model’s overall benchmark position does not settle which provider is best for every application.
What to Watch
Watch for further disclosures about transfers from Binance-linked addresses and whether authorities or service providers can freeze additional funds as assets move between networks. For Coinbase Borrow, the next useful signals will be disclosed New York adoption and loan performance; for AI developers, independent evaluations and real-world costs will help establish how GPT-6.2 performs across different workloads.