Exchange Flows Report — Week 23, 2026: Distribution Dominates as Sell Pressure Doubles Buy Pressure Across Ten Venues

📊 Exchange Flows Report — Week 23

Week 23 of 2026 was not ambiguous. Across ten major trading venues, 1,882 discrete market events generated a combined notional flow that told one consistent story: the market is distributing, not accumulating. Total dump volume clocked in at $6,314.4M against $4,119.8M in pump volume — a $2.2 billion net imbalance toward the sell side. That alone would be notable. What makes Week 23 structurally significant is that the directional confirmation extends all the way down to order flow: sell pressure hit $5,210.7M versus just $2,548.8M in buy pressure. Sellers outgunned buyers by more than two-to-one at the order level.

Binance Futures retained its position as the undisputed volume leader, generating $7,643.9M across 980 events — nearly double the next-largest venue by volume. But the week's most structurally interesting data point came from Hyperliquid, the decentralized perpetuals platform, which logged only 229 events yet produced $3,178.2M in volume. That works out to approximately $13.88M per event, a figure that dwarfs every other venue in the dataset and raises immediate questions about where sophisticated capital is concentrating its firepower. At a time when the overall market is selling down, the efficiency of Hyperliquid's order flow suggests a cohort of large, deliberate actors — not retail noise.

The narrative of Week 23, then, has two layers. The macro layer is straightforwardly bearish: distribution is ongoing, sell pressure is structurally dominant, and the pump-dump ratio at 39.5% / 60.5% leaves little room for bullish interpretation. The micro layer is about routing: where is the smart money executing, how efficiently is it doing so, and what does the divergence between high-event-count retail venues and low-event-count institutional venues tell us about market structure going into Week 24? This report answers both questions, venue by venue, dollar by dollar.

🏆 Exchange Leaderboard

The exchange leaderboard for Week 23, ranked by notional volume, reveals a sharp divergence between volume efficiency and raw event count. The top five venues by volume account for approximately 88% of all notional flow, while the bottom five — despite generating 2,192 combined events — contribute only $1,343.6M, or about 7.4% of total volume. This concentration pattern is the single most important structural observation of the week.

• 1. Binance Futures — $7,643.9M | 980 events | $7.80M per event | Derivatives king
• 2. Hyperliquid — $3,178.2M | 229 events | $13.88M per event | Institutional efficiency outlier
• 3. OKX — $3,102.1M | 441 events | $7.03M per event | Premium futures flow
• 4. Bitget — $1,715.4M | 708 events | $2.42M per event | Mid-tier futures active
• 5. Binance Spot — $922.3M | 388 events | $2.38M per event | Spot market secondary
• 6. Bitunix — $420.4M | 730 events | $0.58M per event | High frequency, low size
• 7. KuCoin — $378.4M | 659 events | $0.57M per event | Retail-heavy, thin events
• 8. Coinbase — $316.6M | 407 events | $0.78M per event | Western premium spot
• 9. OKX Spot — $270.2M | 197 events | $1.37M per event | Cleaner spot routing
• 10. Gate Futures — $228.2M | 646 events | $0.35M per event | Lowest efficiency in dataset

The contrast between Bitunix (730 events, $420.4M) and Hyperliquid (229 events, $3,178.2M) is the leaderboard's defining tension. Bitunix generated 3.2 times more events but only 13% of Hyperliquid's volume. This is not a rounding error — it reflects fundamentally different user bases. Bitunix and Gate Futures, both sitting below $0.60M per event, are operating as retail venues where position sizing is small and turnover is high. Hyperliquid, OKX, and Binance Futures are operating as institutional routing venues where individual orders are large enough to move the needle.

Coinbase at eighth place with $316.6M and 407 events is the notable underperformer relative to its brand recognition and regulatory standing. Its $0.78M per event figure is marginal — better than the pure retail venues but well below what one would expect from a venue that services institutional clients in the US market. Either institutional flow is routing away from Coinbase this week, or the event detection methodology is capturing a different slice of its order book. Either way, the number warrants watching in subsequent weeks.

🔍 Top 3 Exchange Deep Dives

Binance Futures

Binance Futures posted $7,643.9M in notional volume across 980 events, making it the single largest venue in Week 23 by a wide margin — nearly 2.4 times the volume of the second-place finisher. At 980 events out of 1,882 total recorded events, Binance Futures alone accounts for over 52% of all market structure signals in the dataset. This is a venue that is, by every measure, setting the tone for global crypto derivatives markets.

The average notional size per event on Binance Futures was $7.80M — a figure that sits squarely in the institutional range. This is not retail traders punching above their weight; these are position-level events that reflect meaningful capital deployment or withdrawal. Given the overall market context — dump volume exceeding pump volume, sell pressure dominant — the dominant directional flow on Binance Futures this week was almost certainly weighted toward the sell side. The venue's perpetual futures market has historically served as the primary hedging and shorting venue for crypto-native institutions, and Week 23's macro data is consistent with a hedging wave rather than directional accumulation.

The most likely pairs driving volume on Binance Futures this week are BTC-USDT perpetuals, ETH-USDT perpetuals, and a rotating cast of large-cap altcoin perpetuals including SOL, BNB, and potentially XRP given the continued regulatory clarity narrative in the US. Binance Futures typically concentrates 60-70% of its volume in BTC and ETH pairs during risk-off weeks, with altcoin exposure compressing as traders seek liquidity and lower slippage in major pairs. The $7.64B figure in a distribution week suggests active position management — specifically, institutions unwinding length that was built in prior weeks.

Hyperliquid

Hyperliquid's Week 23 performance is the most analytically interesting data point in the entire report. With only 229 events generating $3,178.2M in volume, the platform achieved a per-event average of $13.88M — the highest efficiency ratio in the dataset by a significant margin, outpacing even Binance Futures by 78%. To put this in context: Hyperliquid generated 41.6% of Binance Futures' volume with only 23.4% of its event count.

What does a $13.88M average event size tell us? It tells us that the entities routing through Hyperliquid are not retail participants or even mid-tier professional traders. These are capital allocators operating at a scale where a single order represents meaningful seven-figure notional exposure. Hyperliquid's architecture — an on-chain order book with no KYC requirement and full self-custody — makes it attractive to two distinct populations: sophisticated DeFi-native funds that refuse to expose capital to centralized custody risk, and offshore entities that prefer permissionless execution for regulatory or operational reasons.

In a distribution week, Hyperliquid's large average order size is particularly significant. Large actors distributing positions want deep liquidity and minimal information leakage. Hyperliquid's on-chain order book, while transparent by design, allows for complex execution strategies including cross-venue arb and layered liquidation. The $3.18B in volume at this efficiency level is consistent with a cohort of large players using Hyperliquid as a primary execution venue for significant position exits — not testing the waters, but actually moving size.

OKX

OKX's combined showing across its futures and spot venues tells a story of breadth and depth. OKX Futures generated $3,102.1M across 441 events ($7.03M per event), while OKX Spot contributed an additional $270.2M across 197 events ($1.37M per event). Combined, OKX-branded venues account for $3,372.3M in total notional flow — placing the combined OKX ecosystem in a clear second position if consolidated, ahead of even Hyperliquid on pure volume terms.

OKX Futures' $7.03M per event figure places it in the same institutional tier as Binance Futures, confirming that the Seychelles-headquartered exchange has successfully built a dual retail and institutional customer base. Its futures market has historically attracted Asian institutional flow — particularly from trading firms and family offices in Hong Kong, Singapore, and Dubai — and the Week 23 numbers are consistent with that profile. With 441 events at this average size, OKX Futures is processing roughly one significant market event every 26 minutes across the trading week.

The OKX Spot venue's $1.37M per event average is notably higher than what one typically sees at comparable spot venues, suggesting OKX's spot market is attracting less retail noise and more professional spot trading. This is consistent with OKX's product focus on advanced traders and its relatively high account minimums compared to venues like KuCoin or Gate. In aggregate, OKX's Week 23 performance positions it as the most complete venue in the dataset — institutional-grade futures, cleaner spot flow, and a combined $3.37B in notional that places it firmly in the global top three.

⚡ CEX vs DEX Analysis

The CEX versus DEX comparison in Week 23 is one of the cleanest datasets for this analysis in recent memory. Hyperliquid, as the only decentralized venue in the dataset, generated $3,178.2M in volume across 229 events. All nine centralized exchanges combined produced approximately $14,997.5M across roughly 5,156 events (using the per-venue event counts). This gives DEX a 17.5% share of total notional volume in the dataset — a substantial figure for a single decentralized venue competing against nine established centralized exchanges.

More revealing than the raw share is the efficiency gap. CEX venues averaged approximately $2.91M per event when measured across all nine platforms. Hyperliquid averaged $13.88M per event — a 4.8x premium on per-event notional size. This gap is not random. It reflects deliberate routing decisions by large capital allocators. Entities trading at the $10M+ per event scale have options: they can route through Binance Futures or OKX with their institutional APIs, or they can route through Hyperliquid with its self-custody model and on-chain transparency. The fact that Hyperliquid is attracting volume at this efficiency level suggests that the custody risk tradeoff is, for a meaningful cohort of market participants, worth making.

The trend direction is unmistakably toward DEX at the institutional end of the market. Hyperliquid's growth trajectory over the past twelve months has been steep, and Week 23's numbers are consistent with continued institutional adoption of on-chain perpetuals as a legitimate primary execution venue rather than an experimental alternative. The question is not whether DEX perpetuals will capture more institutional market share — they clearly are. The question is what the ceiling looks like. Custody norms, regulatory clarity, and liquidity depth will determine whether Hyperliquid's share can expand beyond 20-25% of total derivatives notional.

On the retail side, the picture is reversed. CEX venues like Bitunix (730 events, $0.58M/event), KuCoin (659 events, $0.57M/event), and Gate Futures (646 events, $0.35M/event) are clearly processing high-frequency, low-size retail flow. These venues are not meaningfully competing with Hyperliquid at the institutional level — they are serving a different market segment entirely. The risk for these venues is not losing ground to DEX in their core demographic, but rather margin compression as competition for retail futures flow intensifies among the four or five platforms that serve that segment.

🌏 Regional Flow Patterns

Regional analysis of Week 23 flows reveals a clear Asian dominance in derivatives volume, a muted Western performance relative to historical norms, and a Binance-as-global-neutral baseline that continues to function as the price reference point for all other venues.

Asian exchanges — defined here as OKX (including OKX Spot), Bitget, KuCoin, and Gate Futures — collectively generated $5,694.3M in notional volume across 2,051 events. OKX alone accounts for $3,372.3M of that total, making it the de facto Asian institutional champion this week. Bitget's $1,715.4M across 708 events positions it as the most active mid-tier Asian venue, and its $2.42M per event average suggests a mix of professional and semi-professional traders — above retail noise, below institutional scale. KuCoin ($378.4M, 659 events, $0.57M/event) and Gate Futures ($228.2M, 646 events, $0.35M/event) are operating in retail territory, likely driven by Asian retail participants active during UTC+8 trading hours.

The Asian time zone typically produces peak activity windows between 02:00-06:00 UTC (Tokyo/Singapore morning session) and 08:00-12:00 UTC (overlap with European open). The high event counts on KuCoin and Gate Futures are consistent with these windows being particularly active in Week 23. Both venues have historically seen elevated activity during Asian morning sessions when retail participants in South Korea, Japan, and Southeast Asia are most active. The distribution signal during these hours reinforces the macro narrative: Asian retail is selling into strength, not buying dips.

Western exchanges present a more nuanced picture. Coinbase — the primary Western institutional venue — generated only $316.6M across 407 events ($0.78M/event). For a platform with Coinbase's regulatory standing, custody infrastructure, and institutional client base, this is a subdued week. The $0.78M per event average is too low for pure institutional flow and too high for mass retail — it suggests a mix of professional retail and smaller institutions, with the really large US-based capital allocators either sitting on their hands or routing through Binance Futures and OKX directly.

Binance's dual presence — Binance Futures ($7,643.9M) and Binance Spot ($922.3M) — represents the global neutral baseline. Binance is not Asian or Western; it is effectively the global reference market. Its futures dominance reflects its status as the deepest liquidity pool in crypto derivatives globally. The Binance Spot venue's $922.3M across 388 events ($2.38M/event) shows professional but not institutional-scale spot activity, consistent with arbitrageurs and cross-exchange basis traders using Binance Spot as one leg of multi-venue strategies.

💰 Arbitrage Routes Analysis

Week 23's distribution environment created fertile conditions for cross-venue arbitrage, particularly between venues with divergent per-event efficiency. When price discovery happens primarily on high-efficiency venues (Binance Futures, OKX, Hyperliquid) while retail-heavy venues lag, systematic spreads open up that disciplined arb bots can exploit. The data suggests several high-activity arb corridors were in play this week.

The most structurally productive arb route in Week 23 was almost certainly the Binance Futures / Hyperliquid basis trade. Both venues operate liquid perpetual markets with deep BTC and ETH books. During distribution events — where large sell orders on one venue push funding rates and mark prices temporarily out of sync — arb bots can establish offsetting positions across the two venues and collect the spread as it closes. The $10.8 billion in combined notional between these two venues represents an enormous liquidity base, and the efficiency differential (Hyperliquid at $13.88M/event vs Binance Futures at $7.80M/event) suggests price formation was not perfectly synchronized throughout the week.

The second major arb corridor was the OKX Futures / Binance Futures pair. With comparable per-event sizes ($7.03M vs $7.80M) and combined volume of $10.7 billion, these two venues are the most liquid institutional perpetuals pair in the dataset. Intra-day funding rate divergences between OKX and Binance Futures are a well-established arb opportunity, and Week 23's elevated distribution activity would have widened those spreads during peak sell-off periods. Execution desks running cross-venue strategies likely harvested several hundred basis points of annualized yield across the week simply by maintaining delta-neutral positions between these two books.

The retail-to-institutional arb routes — specifically Bitunix/KuCoin/Gate Futures against the major venues — offer lower per-trade yield but higher frequency. With Bitunix at $0.58M/event and Gate Futures at $0.35M/event operating in the same market as $7-14M/event venues, price discrepancies are a structural feature rather than an anomaly. The challenge for arb bots on these routes is execution speed: retail venues with high event counts but shallow books can see price dislocations reverse before an arb can be filled. The viable strategy here is latency arbitrage — co-located bots that capture the spread in milliseconds before the market self-corrects.

Estimated spread conditions for Week 23: Given the distribution environment and elevated sell pressure, funding rates on perpetuals across major venues were likely in negative territory for extended periods — meaning shorts were being paid to hold positions. This creates a structural arb opportunity for entities willing to go long spot on Binance or Coinbase while holding an offsetting short perpetual on Binance Futures or OKX, collecting the negative funding rate as income. The $5,210.7M in total sell pressure and $6,314.4M in dump volume is consistent with funding rates dipping into the -0.01% to -0.05% per 8-hour range on BTC and ETH perpetuals — enough to generate meaningful yield for delta-neutral funding farmers operating at institutional scale.

📈 Market Share Shifts

Without prior-week data for direct week-over-week comparison, the structural analysis of Week 23 market share must rely on the internal composition of the dataset and longer-term trends visible in the venue rankings. Several patterns stand out as likely share shifts rather than stable equilibria.

Hyperliquid's 17.5% share of total notional in the dataset is almost certainly above its six-month average. Decentralized perpetuals platforms have been on a consistent growth trajectory throughout 2025 and into 2026, and Hyperliquid's per-event efficiency advantage over CEX venues suggests it is gaining share at the institutional end of the market specifically. If Hyperliquid's market share six months ago was in the 10-12% range, a move to 17.5% represents meaningful structural shift in institutional routing preferences. The trend is directional: more institutional flow to Hyperliquid, less to mid-tier CEX venues.

The venues most likely losing share are in the mid-tier retail futures category: Bitunix, KuCoin, and Gate Futures. These three platforms combined for 2,035 events and $1,027.0M in volume — $0.50M per event average. In a market where Hyperliquid is capturing whale flow and Binance Futures is capturing the institutional middle, the retail futures tier faces margin compression from two directions. Competing on fees is unsustainable, competing on features requires significant investment, and competing on liquidity depth is structurally impossible given Binance Futures' dominance.

Coinbase's position deserves particular attention in the context of market share trends. The ongoing regulatory clarification in the US — which has historically been Coinbase's primary competitive moat — has not translated into dominant market share in Week 23's dataset. At $316.6M, Coinbase is eighth by volume, behind even Bitunix in raw event count. One hypothesis: US institutional capital that previously routed through Coinbase for regulatory safety is now comfortable routing through Binance Futures and OKX directly, reducing Coinbase's structural advantage. If this is true, it represents a meaningful long-term share loss for the US exchange incumbent.

Bitget's position as the fourth-largest venue by volume ($1,715.4M) and the third-most-active by event count (708 events) suggests it is gaining share in the mid-to-large professional trader segment. Its $2.42M per event average is in a sweet spot between retail noise and institutional scale — consistent with professional trading firms and active prop shops using Bitget as a primary or secondary execution venue. If Bitget can maintain this positioning, it is well-placed to be the primary beneficiary of flow that leaves mid-tier retail venues without the scale to route to the top three.

🔮 Next Week Watch

Week 24 sets up as a resolution week for the distribution narrative that dominated Week 23. The key question is whether the $2.2 billion net imbalance between dump and pump volume represents a completed distribution phase or an ongoing one. Market structure history suggests that when sell-side pressure sustains at 2:1 over buy-side pressure for a full week, the first few sessions of the following week often see a brief counter-trend rally as short-sellers take profits — followed by a continuation of the downtrend if underlying fundamentals support it. Watch funding rates on Binance Futures and OKX closely: if they normalize to near-zero or flip positive in the first two sessions of Week 24, it signals short covering. If they stay negative, distribution is ongoing.

• Hyperliquid: Watch whether the $13.88M/event efficiency sustains — any drop toward $8-10M/event signals that the large actors who dominated Week 23 have finished positioning.
• Binance Futures: The 980-event, $7.64B week sets a high bar. A significant drop in event count in Week 24 (below 700) with sustained volume would indicate fewer but larger institutional orders — typically a sign of coordinated directional positioning.
• Coinbase: A meaningful uptick from $316.6M would suggest US institutional capital is re-entering the market. Watch this as the leading indicator for Western demand recovery.
• Bitget: If it sustains above $1.5B in volume for Week 24, its position as the mid-tier institutional venue of choice is structurally confirmed, not just a one-week phenomenon.
• Funding rates: Sustained negative funding across BTC and ETH perpetuals on Binance Futures and OKX through Week 24 is the single most reliable indicator that distribution has not yet concluded.
• CEX vs DEX share: If Hyperliquid crosses 20% of total notional in Week 24, it marks a structural threshold — one-fifth of observable crypto derivatives volume routing through a fully decentralized venue is a market structure milestone.
• Gate Futures and KuCoin: If their per-event averages drop below $0.30M in Week 24, it signals retail withdrawal — a historically reliable leading indicator of broader market capitulation or consolidation.

The macro setup for Week 24 is not particularly bullish. $6.31B in dump volume does not reverse in one week without a significant catalyst — whether macroeconomic (Fed policy surprise, risk-on equity session) or crypto-native (major protocol unlock, ETF flow reversal, large liquidation cascade clearing the air). Absent such a catalyst, the base case is continued range compression with occasional sharp moves driven by liquidation cascades as overleveraged longs from prior weeks get flushed. Position accordingly.

Sign Off

Week 23 delivered one of the cleaner distribution signals in recent memory. The data does not require interpretation — it requires reading. $6.31B in dump volume. $5.21B in sell pressure. Ten venues, 1,882 events, and one directional conclusion: sellers are in control. The sophistication of the distribution is what makes this week analytically compelling: it is not panicked retail selling (which would show up as low per-event averages across the board), but rather large, deliberate position exits concentrated at the institutional venues. Hyperliquid at $13.88M per event, Binance Futures at $7.80M, OKX at $7.03M — these are the fingerprints of organized capital management, not emotional liquidation.

The market structure implication is that whoever is distributing is patient, well-capitalized, and has access to institutional execution infrastructure across at least three major venues simultaneously. That is not a retail actor. It is not a single fund. It is a cohort of similarly-positioned large entities responding to the same macro signals. When that cohort finishes its distribution cycle — and it will — the order flow data will show it first: pump volume will begin to exceed dump volume, buy pressure will approach parity with sell pressure, and the per-event efficiency at Hyperliquid and Binance Futures will shift from sell-side to buy-side dominance. We are not there yet. Watch the data.

Exchange Flows — Week 23