Context and references.

Weekly Bitcoin exchange balance flow through early May 2026 showed a net outflow of roughly 4,200 BTC — consistent with the structural pattern of BTC moving to self-custody and cold storage. Glassnode and CryptoQuant both publish the canonical series; CoinDesk references this metric routinely in market structure coverage.

The May 2026 reading continues the multi-year structural decline in exchange-held BTC. Peak exchange balance was approximately 3.0 million BTC in March 2020; current readings near 2.3 million BTC represent a 23% reduction in immediately-sellable supply. The trend is durable and reflects the broader institutional and self-custody adoption pattern.

Stablecoin flow is the corresponding companion metric. The same week saw USDT and USDC supply expand by roughly $800M combined, indicating dry powder continues to build for crypto deployment. Bloomberg Crypto and The Block both track these flows in weekly summaries.

Whale wallet activity (defined as wallets holding 1,000+ BTC) showed continued accumulation behavior in early May 2026. Glassnode's whale supply share metric rose by 0.4 percentage points week-over-week, reaching 18.2% of circulating supply — the highest reading since pre-FTX days.

Why two providers give you two different numbers.

Exchange balance is not an on-chain fact. The chain shows addresses; it does not label them. Glassnode and CryptoQuant each maintain their own map of which address clusters belong to which venue, built from deposit-address heuristics and in-house research, and neither publishes that map in full. So the 4,200 BTC net outflow above is a labelled estimate, not a measurement. Pull the same week from the other provider and expect a different figure.

That is not a reason to distrust either one. It is a reason never to mix them inside a single series. If you have been reading CryptoQuant since January, keep reading CryptoQuant — a dramatic "trend break" that turns out to be you switching dashboards is the most common self-inflicted error with this metric. Re-labelling also rewrites history: when a provider re-clusters a large venue's wallets, the historical curve moves under you, and a chart you saved last quarter will no longer match the live one.

The same caveat governs the 18.2% whale supply share. A 1,000+ BTC threshold is an address-level cut, and exchange and custodian cold wallets sit far above it. Providers strip known custodial addresses out with differing levels of aggressiveness, which is why whale-share readings from two dashboards can differ by more than the week-over-week move being discussed. The direction over several weeks is the readable part. The first decimal place is not.

Crypto assets are volatile and not suitable for every investor. This page is editorial analysis, not financial advice.