- \n
- A token\u2019s liquidity is only as good as its worst venue. CoinMarketCap tracked 239 spot exchanges on 18 September 2026, and the ten largest alone list 16,045 trading pairs; a typical mid-cap token is live on five to twelve of them, each with its own book, fee schedule and API.<\/li>\n
- Cross exchange market making replaces ten independent bots with one engine: quote on the maker venues, hedge on the hedge venue, rebalance inventory on a schedule, and run risk on the sum of all positions rather than on each one.<\/li>\n
- Exchange integration is the hard part. Each venue has different rate limits (Binance allows 6,000 request weight per minute per IP and 50 orders per 10 seconds per account; Bybit allows 600 requests per 5 seconds per IP), different server regions (Tokyo, Singapore, Hong Kong), different symbol conventions and different fee tiers. A serious multi-exchange setup is measured in months of engineering, not days.<\/li>\n
- When makers cannot reach one venue, that venue\u2019s prices detach from everyone else\u2019s. On 10 October 2025, USDe printed $0.65 and wBETH $430 on Binance while both held near reference value on other venues; Binance later paid about $283 million in compensation.<\/li>\n
- Exchanges now police the relationship. Since March 2026, Binance requires token issuers to disclose their market makers, contract terms and token-loan arrangements, and prohibits one-sided selling and volume inflation. Multi-venue quoting has to be real, two-sided and reportable.<\/li>\n
- Since 2015, TDMM<\/a> has supported liquidity across CEX and DEX markets, with $10B+ in trading volume, 200+ markets managed, and 100+ exchange integrations, operated through a unified book and risk framework.<\/li>\n<\/ul>\n
What is cross exchange market making?<\/b><\/h1>\n
Ordinary market making means placing a bid and an ask on one trading pair, on one exchange, and earning the spread between them while managing the inventory that builds up as trades fill. The earlier article in this series, How does market making work in crypto exchanges?<\/span><\/a>, covers that mechanism in detail.<\/p>\n
Cross exchange market making extends the same job across venues. The market maker quotes on one or more exchanges, called the maker venues, and uses a deeper exchange, called the hedge venue or taker venue, as both the price reference and the place where fills are offset. Hummingbot, the open-source framework that popularised the strategy for retail bots, describes it plainly: the trader \u201cacts as a normal market maker on one exchange (the maker exchange), creating buy and sell orders\u201d and \u201celiminates the inventory risk usually associated with market making by using another exchange (the taker exchange) to simultaneously offset (or hedge) any trades filled on the maker exchange.\u201d<\/p>\n
The pricing rule is one line:<\/p>\n
Maker ask = hedge-venue best ask + margin, and maker bid = hedge-venue best bid \u2212 margin<\/b><\/p>\n
where the margin has to cover the maker fee on the quoting venue, the taker fee on the hedge venue, expected slippage on the hedge, the cost of moving inventory between venues, and a profit buffer. Hummingbot calls the sum of these the minimum profitability, and the same arithmetic sits inside every professional engine.<\/p>\n
Two things separate a professional cross exchange market maker from a retail bot running that formula. The first is scale: a firm such as TDMM quotes the same token on many maker venues at once, not one, and often hedges across several deep venues and perpetual futures rather than a single taker book. The second is that the inventory is treated as one portfolio. Balances on Binance, OKX, Bybit, KuCoin, Gate and a Uniswap pool are one position with one net exposure, one set of caps and one kill switch, even though the exchanges themselves have no idea the others exist.<\/p>\n
The three roles a venue can play<\/b><\/h2>\n
\n\n
\n \n Role<\/p>\n<\/td>\n
\n What the engine does there<\/p>\n<\/td>\n
\n Typical venue for a mid-cap token<\/p>\n<\/td>\n<\/tr>\n
\n \n Hedge venue (reference)<\/p>\n<\/td>\n
\n Reads the book to compute fair value; takes liquidity to offset fills from elsewhere<\/p>\n<\/td>\n
\n The exchange with the deepest book for the token, often the primary listing<\/p>\n<\/td>\n<\/tr>\n
\n \n Maker venue<\/p>\n<\/td>\n
\n Posts two-sided quotes at reference \u00b1 margin; cancels and replaces as the reference moves<\/p>\n<\/td>\n
\n Secondary CEX listings: Gate, KuCoin, MEXC, Bitget, HTX, regional exchanges<\/p>\n<\/td>\n<\/tr>\n
\n \n AMM venue<\/p>\n<\/td>\n
\n Holds a concentrated-liquidity range and rebalances it; hedges the range\u2019s delta on a CEX<\/p>\n<\/td>\n
\n Uniswap, PancakeSwap, Raydium, depending on chain<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
The same venue can play more than one role for different tokens, and the roles can swap when depth migrates. A token that launches on a DEX and later lists on a large CEX usually sees its hedge venue move from the pool to the order book within weeks.<\/p>\n
Why crypto exchanges make cross exchange market making necessary<\/b><\/h2>\n
Liquidity in crypto is fragmented by design. There is no consolidated tape, no national best bid and offer, and no obligation on any exchange to match another\u2019s price. That produces three structural facts that shape every market-making decision.<\/p>\n
![\"Bar](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/01-exchange-fragmentation-markets-coins-300x169.png\")
The same tokens trade on ten order books that do not talk to each other. Trading pairs and coins listed on the ten largest spot exchanges, 18 September 2026.<\/figcaption><\/figure>\n The venue count is large and the books are separate.<\/b> On 18 September 2026, CoinMarketCap tracked 239 spot exchanges. The ten largest by volume listed between 520 (Coinbase) and 2,796 (MEXC) trading pairs each, 16,045 in total. CoinGecko\u2019s 2026 Spot CEX report counts 14,485 pairs across the top twelve centralized exchanges in 2025, of which 9,870 were stablecoin pairs and 97.7% of those were quoted in USDT or USDC. Most tokens with any real following are listed on several of these venues, and every listing is a separate order book with its own bids, asks, fees and API.<\/p>\n
Depth is concentrated, but the concentration moves.<\/b> Kaiko\u2019s liquidity concentration report found the eight largest exchanges holding 91.7% of global order-book depth and 89.5% of volume, with Binance alone at 30.7% of depth. CoinGecko\u2019s Q2 2026 report puts Binance at 38.7% of top-ten CEX spot volume, with Bybit second at 10.0% after MEXC\u2019s volume more than halved, from $275.2 billion to $121.2 billion, in a single quarter. Rankings below the top spot reshuffle constantly, and a token whose liquidity plan assumed one venue would stay dominant is exposed when the volume moves.<\/p>\n
On-chain venues are now part of the map.<\/b> CoinGecko\u2019s CEX and DEX activity report shows the DEX share of spot volume doubling from 6.9% in January 2024 to 13.6% in January 2026, with a peak of 24.5% in June 2025, and the DEX share of perpetual volume rising from 2.0% to 10.2%. Between August 2025 and January 2026, PancakeSwap and Uniswap each handled roughly $0.55 trillion of spot volume, volumes on the scale of the largest centralized exchanges after Binance. A token\u2019s price is now discovered across order books and automated market makers at once, and a maker that only covers one kind of venue is covering part of the market.<\/p>\n
![\"Six](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/03-multi-exchange-market-at-a-glance-2026-300x169.png\")
The multi-exchange market at a glance, September 2026.<\/figcaption><\/figure>\n What fragmentation does to prices<\/b><\/h2>\n
Prices on separate venues drift apart, and the drift is larger and longer-lasting than most founders expect. The reference study is Makarov and Schoar\u2019s Trading and Arbitrage in Cryptocurrency Markets<\/i>, which examined 34 exchanges across 19 locations over 2017 and 2018 and found that price deviations \u201cpersisted from several hours to days and weeks\u201d. The cross-country gaps were the extreme case, with the US\u2013South Korea premium reaching 40% for days, but even within the same jurisdiction the paper measured daily arbitrage spreads that were routinely material after fees. Markets have matured since then, and the largest pairs on the largest venues now trade within a basis point or two of each other, but the mechanism has not changed: where there is no one pushing a common price into every book, the books diverge.<\/p>\n
For a mid-cap token the practical picture in 2026 looks like this. On its deepest venue the token may quote at a 10 to 30 basis point spread. On the fourth or fifth venue, with no dedicated maker, the spread is often 100 to 300 basis points and the mid-price can sit half a percent away from the deep venue for minutes at a time. Arbitrage bots close the gap eventually, but they do it by taking liquidity from the thin book, which widens it further, and they keep the profit. Cross exchange market making is the alternative: the token\u2019s own market maker posts the consistent price on every venue and captures the gap itself, while the project gets a token that trades at one price everywhere.<\/p>\n
When one venue loses its makers: 10 October 2025<\/b><\/h2>\n
The clearest demonstration of what happens when a venue is cut off from the cross-exchange price came during the largest liquidation cascade in crypto history. Between 21:36 and 22:16 UTC on 10 October 2025, three assets on Binance detached from their value everywhere else. Ethena\u2019s USDe, which stayed close to $1.00 on-chain and on other exchanges, printed $0.65 on Binance. wBETH, Binance\u2019s own wrapped staked ether, traded down to $430 while ETH\/USDT was above $3,800, an 88% discount. BNSOL printed $34.90. The prints lasted about 40 minutes.<\/p>\n
![\"Chart](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/04-binance-depeg-10-oct-2025-venue-dislocation-300x169.png\")
When one venue loses its makers: Binance low prints between 21:36 and 22:16 UTC on 10 October 2025 as a share of the reference price the same assets held elsewhere.<\/figcaption><\/figure>\n Post-mortems from 21Shares and others attribute the dislocation to the exchange\u2019s infrastructure being overwhelmed and to market makers being unable to reach the venue, leaving books so thin that \u201ceven small orders move the market\u201d. Binance\u2019s margin engine marked positions against those distressed spot prints, liquidated users, and the exchange subsequently paid about $283 million in compensation. The assets had not lost value; one venue had lost its makers.<\/p>\n
For token teams the lesson is specific. A price is only as reliable as the makers who keep it aligned with other venues. A single-venue liquidity plan has no other venue to align with.<\/p>\n
How cross exchange market making works, step by step<\/b><\/h2>\n
![\"Diagram](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/02-how-cross-exchange-market-making-works-300x169.png\")
How cross exchange market making works: one loop, repeated every few milliseconds, per pair, per venue.<\/figcaption><\/figure>\n The engine runs a five-stage loop, continuously, for every pair on every maker venue.<\/strong><\/h2>\n
1. Read the reference price<\/b><\/h3>\n
The engine subscribes to the hedge venue\u2019s order book over WebSocket and computes a fair value: usually the microprice (a mid-price weighted toward the side of the book with more size), adjusted for any perpetual-futures basis and for the short-term volatility it is observing. On tokens with several deep venues the reference is a depth-weighted blend rather than a single book, so that one exchange\u2019s hiccup does not move every quote.<\/p>\n
2. Quote on the maker venue<\/b><\/h3>\n
It posts layered bids and asks on each maker venue at the reference plus and minus a margin, with sizes set by the venue\u2019s share of the token\u2019s volume and by how much inventory is available there. As the reference moves, it cancels and replaces. On a centralized exchange this happens many times a second; Hummingbot\u2019s implementation exposes parameters such as an order refresh interval, an anti-hysteresis duration to stop the engine chasing every tick, and a top-depth tolerance that decides how much size may sit ahead of the quote before it is repriced.<\/p>\n
3. A fill arrives<\/b><\/h3>\n
A trader on the maker venue crosses the quote. The engine now holds, say, 20,000 more tokens on KuCoin than it wants.<\/p>\n
4. Hedge immediately<\/b><\/h3>\n
It sells 20,000 tokens on the hedge venue at the best available price. Hummingbot\u2019s rule is that the strategy \u201cwould always immediately hedge any order fills from the maker side, regardless of how profitable the hedge is\u201d, and professional desks follow the same principle: the hedge is a risk action, not a trade to be optimised. The margin was set in step 2 to make the round trip profitable on average; the hedge\u2019s job is to make the exposure disappear.<\/p>\n
5. Rebalance<\/b><\/h3>\n
Over time the maker venue accumulates quote currency and the hedge venue accumulates tokens, or the reverse. The engine schedules transfers between venues, or offsets the drift by skewing quotes, so that every venue keeps a working balance on both sides. Transfers take minutes to hours depending on the chain and the exchange\u2019s withdrawal queue, which is why every venue holds a buffer rather than the minimum.<\/p>\n
A worked example<\/b><\/h3>\n
Suppose a token trades at $1.0000 bid and $1.0004 ask on its hedge venue, Binance, and the maker is quoting it on KuCoin.<\/p>\n
\n\n
\n \n Item<\/p>\n<\/td>\n
\n Value<\/p>\n<\/td>\n<\/tr>\n
\n \n Hedge-venue best ask (Binance)<\/p>\n<\/td>\n
\n $1.0004<\/p>\n<\/td>\n<\/tr>\n
\n \n Margin target<\/p>\n<\/td>\n
\n 0.35% (35 bps)<\/p>\n<\/td>\n<\/tr>\n
\n \n Maker ask posted on KuCoin<\/p>\n<\/td>\n
\n $1.0039<\/p>\n<\/td>\n<\/tr>\n
\n \n Fill: trader buys 20,000 tokens on KuCoin<\/p>\n<\/td>\n
\n +$20,078 quote currency received; \u221220,000 tokens<\/p>\n<\/td>\n<\/tr>\n
\n \n Hedge: buy 20,000 tokens on Binance at $1.0005 (slippage 1 bp)<\/p>\n<\/td>\n
\n \u2212$20,010<\/p>\n<\/td>\n<\/tr>\n
\n \n Gross margin<\/p>\n<\/td>\n
\n $68 (34 bps)<\/p>\n<\/td>\n<\/tr>\n
\n \n KuCoin maker fee at a market maker tier: \u22121.2 bps rebate<\/p>\n<\/td>\n
\n +$2.41<\/p>\n<\/td>\n<\/tr>\n
\n \n Binance taker fee at VIP tier: 4 bps<\/p>\n<\/td>\n
\n \u2212$8.00<\/p>\n<\/td>\n<\/tr>\n
\n \n Net on the round trip<\/p>\n<\/td>\n
\n \u2248 $62 (31 bps)<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
The engine is now flat: it has sold on KuCoin and bought on Binance, and holds the same number of tokens it started with, in different places. Multiply by hundreds of fills a day across ten venues and the model is the daily business of a cross exchange market maker. The risks are the moments when step 4 fails: the hedge venue moves before the hedge lands, the hedge venue\u2019s API rejects the order, or a withdrawal freeze stops the rebalance in step 5.<\/p>\n
The strategy parameters that matter<\/b><\/h2>\n
Hummingbot\u2019s cross exchange market making strategy is a useful public reference for the knobs every engine needs, even if proprietary systems name them differently.<\/p>\n
\n\n
\n \n Parameter<\/p>\n<\/td>\n
\n What it controls<\/p>\n<\/td>\n
\n Why it matters across venues<\/p>\n<\/td>\n<\/tr>\n
\n \n min_profitability<\/p>\n<\/td>\n
\n Minimum margin between maker quote and hedge price, after fees<\/p>\n<\/td>\n
\n Set per venue pair; a venue with a 10 bps maker fee needs a wider margin than one paying a rebate<\/p>\n<\/td>\n<\/tr>\n
\n \n order_amount<\/p>\n<\/td>\n
\n Size per quote<\/p>\n<\/td>\n
\n Sized to what the hedge venue can absorb without slippage<\/p>\n<\/td>\n<\/tr>\n
\n \n slippage_buffer<\/p>\n<\/td>\n
\n Extra price allowance on the hedge order<\/p>\n<\/td>\n
\n Wider on thin hedge venues and during volatility<\/p>\n<\/td>\n<\/tr>\n
\n \n order_size_taker_volume_factor<\/p>\n<\/td>\n
\n Fraction of hedge-venue depth the engine is willing to consume<\/p>\n<\/td>\n
\n Prevents hedging into the maker\u2019s own price impact<\/p>\n<\/td>\n<\/tr>\n
\n \n order_size_taker_balance_factor<\/p>\n<\/td>\n
\n Fraction of hedge-venue balance available for hedging<\/p>\n<\/td>\n
\n Ties quoting to actual inventory on the hedge venue<\/p>\n<\/td>\n<\/tr>\n
\n \n active_order_canceling<\/p>\n<\/td>\n
\n Whether to cancel and replace actively or let orders expire<\/p>\n<\/td>\n
\n Enabled on CEXs; disabled on DEXs to save gas<\/p>\n<\/td>\n<\/tr>\n
\n \n anti_hysteresis_duration<\/p>\n<\/td>\n
\n Minimum time between quote adjustments<\/p>\n<\/td>\n
\n Stops rate-limit exhaustion on venues with tight order caps<\/p>\n<\/td>\n<\/tr>\n
\n \n top_depth_tolerance<\/p>\n<\/td>\n
\n Size allowed ahead of the quote before repricing<\/p>\n<\/td>\n
\n Trades off queue position against staleness<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
A professional stack adds venue-level inventory caps, per-venue kill switches, a cross-venue price-deviation alarm and an automated transfer scheduler, and it calibrates each parameter per venue rather than globally.<\/p>\n
Exchange integration: what it takes to connect to multiple crypto exchanges<\/b><\/h2>\n
The strategy is simple to describe and the pricing formula fits on one line. The difficulty is exchange integration, and it is the reason most token teams hire a firm that has already done it rather than building a multi-exchange stack themselves.<\/p>\n
![\"Table](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/05-exchange-api-rate-limits-integration-constraints-300x169.png\")
Exchange integration is not one job, it is one job per venue: published API limits and best-latency region for the venues most token teams start with.<\/figcaption><\/figure>\n Accounts, keys and permissions<\/b><\/h3>\n
Every venue requires a verified corporate account, usually with sub-accounts per strategy or per client, API keys with trading permission but no withdrawal permission on the quoting keys, IP whitelists, and, on most large exchanges, an application to the market maker programme to obtain higher rate limits and maker rebates. Binance\u2019s spot programme asks for evidence of 1,000 BTC of 30-day trading volume on Binance or elsewhere. Bybit\u2019s programme tiers spot makers from MM1 (30-day volume above $25 million) to MM3 (1% or more maker volume share). KuCoin\u2019s programme grants tiered maker rebates, a two-month assessment-free period and API capacity of 40,000 requests per 30 seconds. Each application is a separate process with separate documentation.<\/p>\n
Rate limits and order caps<\/b><\/h3>\n
The published limits differ by an order of magnitude between venues and they are enforced differently. Binance\u2019s spot WebSocket API allows 6,000 request weight per minute per IP, 50 orders per 10 seconds and 160,000 orders per day per account, and 300 WebSocket connections per five-minute window per IP, with every session capped at 24 hours. Bybit allows 600 REST requests per 5 seconds per IP and 20 spot orders per second per UID by default, with 500 new WebSocket connections per five minutes. OKX publishes per-endpoint limits, starting at 20 requests per 2 seconds on public market data, and scales the order limits for VIP5 and above by a fill-ratio formula that penalises accounts that cancel far more than they fill.<\/p>\n
An engine quoting five layers on each side of twenty pairs on Binance is generating roughly 200 resting orders and, in a moving market, hundreds of cancel-replace messages per minute. Without batching, request-weight accounting and a per-venue throttle, it will be rate-limited at exactly the moment it most needs to reprice. This is the engineering that separates a bot from an integration.<\/p>\n
Latency and server placement<\/b><\/h3>\n
![\"Bar](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/06-exchange-api-latency-by-region-300x169.png\")
Where your servers sit decides whether your hedge lands: round-trip API latency from the best and worst cloud regions.<\/figcaption><\/figure>\n Crypto exchanges are hosted in specific cloud regions and the round-trip latency to them varies by more than an order of magnitude by location. CCXT\u2019s 2026 latency survey measured Binance at about 8 milliseconds from Tokyo and 259 from S\u00e3o Paulo, Gate at 7 from Tokyo and 281 from S\u00e3o Paulo, OKX at 29 from Hong Kong and 353 from S\u00e3o Paulo, Coinbase at 10 from New York or London and 1,109 from Sydney, and Kraken at 16 from London and 268 from Sydney. KuCoin offers colocation at about 3 milliseconds for programme members.<\/p>\n
For cross exchange market making the number that matters is the sum of two legs: the time to see the reference move on the hedge venue and the time to reprice on the maker venue. If the maker venue is in Tokyo and the hedge venue in Hong Kong, the engine needs presence in both regions and a fast link between them, or the arbitrageurs will pick off its stale quotes before the hedge can be placed. Some venues also restrict access by IP range or jurisdiction, which constrains where the servers can sit at all.<\/p>\n
Symbol, precision and settlement differences<\/b><\/h3>\n
The same token is TOKEN\/USDT<\/span> on one venue, TOKEN-USDT<\/span> on another and TOKENUSDT<\/span> on a third. Price tick sizes, lot sizes, minimum notionals and decimal precision differ by venue and change without notice when an exchange adjusts a pair. Fees are charged in the quote asset on one venue and in the base asset, or in the exchange token, on another. Some venues settle instantly; others batch. Every one of these differences has to be normalised into a single internal representation before a unified book is possible, and every one of them is a source of silent errors when it drifts.<\/p>\n
Fees<\/b><\/h3>\n
![\"Chart](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/09-maker-fees-standard-vs-market-maker-programme-300x169.png\")
Maker fees differ by venue, and market maker programmes flip the sign: spot maker fee in basis points, standard account versus best published market maker tier.<\/figcaption><\/figure>\n Standard-tier spot maker fees in September 2026 are 10 basis points on Binance, Bybit, Bitget and KuCoin, 8 on OKX and 0 on MEXC. Market maker programmes turn the fee into a rebate: Bybit\u2019s MM3 tier pays \u22120.75 basis points on spot, KuCoin\u2019s Tier S pays \u22121.2. The taker fee on the hedge venue is the larger cost, and it too depends on VIP tier. The minimum viable margin on each maker venue is therefore a venue-pair-specific number, and a cross exchange engine carries a fee matrix, not a fee.<\/p>\n
DEX integration<\/b><\/h3>\n
Connecting to a decentralized exchange is a different discipline. There is no API key; there is a wallet, an RPC endpoint, gas estimation, transaction ordering risk and, on Uniswap v3-style pools, a concentrated-liquidity range that must be rebalanced as price moves. Hummingbot notes that interaction with DEXs \u201cis less reliable\u201d and that taker fees vary with gas, which makes the margin calculation unpredictable. A maker that runs both CEX order books and AMM ranges hedges the AMM position\u2019s delta on a centralized venue and treats gas as a fee line in the same matrix.<\/p>\n
The integration checklist per venue<\/b><\/h2>\n
\n\n
\n \n Item<\/p>\n<\/td>\n
\n What has to be done<\/p>\n<\/td>\n
\n Typical time<\/p>\n<\/td>\n<\/tr>\n
\n \n Corporate KYB and sub-accounts<\/p>\n<\/td>\n
\n Entity verification, sub-account structure, signatories<\/p>\n<\/td>\n
\n 3 to 15 business days<\/p>\n<\/td>\n<\/tr>\n
\n \n Market maker programme<\/p>\n<\/td>\n
\n Application, volume evidence, pair list, KPI acceptance<\/p>\n<\/td>\n
\n 1 to 4 weeks<\/p>\n<\/td>\n<\/tr>\n
\n \n API connectivity<\/p>\n<\/td>\n
\n REST and WebSocket clients, authentication, heartbeat, reconnect logic<\/p>\n<\/td>\n
\n 1 to 3 weeks of engineering per new venue<\/p>\n<\/td>\n<\/tr>\n
\n \n Rate-limit and order-cap model<\/p>\n<\/td>\n
\n Weight accounting, batching, throttles, backoff<\/p>\n<\/td>\n
\n Part of the above<\/p>\n<\/td>\n<\/tr>\n
\n \n Symbol and precision normalisation<\/p>\n<\/td>\n
\n Mapping, tick and lot rules, fee currency, refresh on change<\/p>\n<\/td>\n
\n Ongoing<\/p>\n<\/td>\n<\/tr>\n
\n \n Latency placement<\/p>\n<\/td>\n
\n Region selection, cross-region links, IP compliance<\/p>\n<\/td>\n
\n 1 week<\/p>\n<\/td>\n<\/tr>\n
\n \n Inventory and transfer routes<\/p>\n<\/td>\n
\n Deposit addresses, whitelists, chain selection, buffers<\/p>\n<\/td>\n
\n 1 week<\/p>\n<\/td>\n<\/tr>\n
\n \n Risk controls<\/p>\n<\/td>\n
\n Per-venue caps, kill switches, deviation alarms, reporting<\/p>\n<\/td>\n
\n Before go-live<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
A firm with these steps already complete for 100+ venues can add a token to all of them in days. A team starting from zero is looking at a quarter of engineering before the first hedge lands.<\/p>\n
Inventory, hedging and risk across venues<\/b><\/h2>\n
![\"Diagram](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/08-unified-inventory-cross-exchange-book-300x169.png\")
One inventory, many venues: each venue holds a working balance of token and quote asset; the risk engine sees the sum, not the parts.<\/figcaption><\/figure>\n The unified book is what turns a set of integrations into cross exchange market making. Every venue holds a working balance of the token and the quote asset. The risk engine sees the total: net token exposure across all venues, net stablecoin, unrealised P&L, and the distance of each venue\u2019s price from the reference. Five rules keep it healthy.<\/p>\n
Venue caps.<\/b> Each venue has a maximum inventory it may hold, sized to the venue\u2019s counterparty risk and its withdrawal reliability. Exchange failures are not hypothetical; the limit exists so that no single venue outage can take a material share of the client\u2019s inventory with it.<\/p>\n
Basis risk is measured, not ignored.<\/b> Between the fill on the maker venue and the hedge on the hedge venue, the reference can move. Over thousands of round trips this is a cost line, and the engine reports it. When it grows, the margin widens or the venue is paused.<\/p>\n
Transfers are scheduled, not reactive.<\/b> Moving tokens between exchanges takes a chain confirmation and an exchange withdrawal queue, and both slow down precisely when markets are fast. The engine keeps buffers on every venue and rebalances on a schedule so that it is never forced to move inventory in a crisis.<\/p>\n
Quotes skew before they stop.<\/b> When inventory drifts long on a venue, bids move down and asks move down with them, so the venue sells more readily until it is back at target. Withdrawal of quotes is the last resort, triggered by a kill switch on stale data, abnormal fills or a price deviation beyond the alarm threshold.<\/p>\n
![\"Comparison](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/07-single-vs-cross-exchange-market-making-comparison-300x169.png\")
![\"Timeline](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/12-multi-exchange-market-making-setup-30-day-plan-300x169.png\")
![\"TDMM](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/11-tdmm-exchange-integration-list-300x169.png\")
![\"Bar](\"https:\/\/tdmm.io\/insights\/wp-content\/uploads\/2026\/09\/10-market-maker-exchange-coverage-comparison-300x169.png\")