Treasury desks moving stablecoins at size face a recurring question: pick up the phone to a single OTC desk, fire a Request-For-Quote to four market makers at once, or route onchain through a DEX aggregator. The right answer depends on ticket size, counterparty relationships, and how much price discovery you need before signing. This guide breaks down OTC, RFQ, CEX, and DEX execution for USDC, USDT, USDe, and PYUSD flows, with concrete provider comparisons and the trade size thresholds where each rail wins.
​

OTC (Over-The-Counter) stablecoin execution is a bilateral, principal trade between a treasury desk and a single market maker. The desk requests a price for a specific size, the market maker quotes a firm two-way market, and the trade settles directly. No order book, no public price discovery, no slippage.
​

The major stablecoin OTC desks are Cumberland (DRW), FalconX, Galaxy Digital, B2C2, and Wintermute. Cumberland and Galaxy operate the largest stablecoin OTC desks by volume, with active-day flow spanning eight figures on liquid names. Tickets typically start around $1M, with relationship desks regularly clearing $50M to $250M single fills for fintechs, payment processors, and corporate treasuries onboarding to USDC or USDT.
​

Settlement runs offchain (wire-to-wire fiat plus onchain stablecoin delivery), usually T+0 or T+1. Pricing comes as a spread off a reference index (Coinbase USDC/USD or Kraken USDT/USD). None of these desks publish a spread schedule, so the only reliable way to learn a desk's pricing is to onboard and ask for a quote on your own size.
​

RFQ (Request-For-Quote) is a structured workflow where a treasury desk submits a size and instrument to a panel of market makers simultaneously. Each market maker returns a firm quote within seconds, the desk hits the best price, and the platform handles settlement. It is competitive bidding without exposing intent to a public order book.
​

Paradigm is the dominant institutional crypto RFQ venue for options, futures, and spot blocks. Hidden Road (acquired by Ripple in April 2025 for $1.25B) operates a prime brokerage with integrated multi-dealer RFQ across 30-plus liquidity providers. Wintermute participates as a maker on both Paradigm and Hidden Road while also offering direct OTC.
​

RFQ shines in the mid-size ticket band where a desk wants price discovery without revealing flow to a single counterparty. The structural reason it prices better than a single dealer is that the maker knows it is being shopped, so the quote reflects competition rather than relationship. The trade-off is operational overhead: integrating an RFQ platform, KYB onboarding to each maker, and managing credit lines.
​

OTC wins when ticket size, relationship pricing, or operational simplicity outweigh competitive price discovery. Three patterns matter most for stablecoin flows.
​

Predictable, recurring size of $10M-plus. A payments processor converting $25M of USD to USDC every Monday gets sharper pricing from a relationship desk than from shopping the trade. Cumberland, FalconX, and Galaxy quote tighter on recurring flow because they can pre-position inventory. A single-dealer OTC line also collapses operational overhead: one KYB, one credit agreement, one settlement workflow.
​

Illiquid pairs. Converting $5M of USDe to USDT or sourcing PYUSD in size is a phone call, not an RFQ. The maker panel on Paradigm or Hidden Road may only have one or two real bidders on these names, so the RFQ collapses to single-dealer pricing anyway. Cumberland and B2C2 actively make markets in long-tail stablecoins; Wintermute is the deepest source for USDe and USDC.e.
​

Sensitive flow. Corporate treasuries unwinding a $100M USDC position before a regulatory filing or a token unlock prefer one trusted counterparty. RFQ surfaces the flow to four or five makers, any of whom could trade ahead. Single-dealer OTC under a written non-solicitation reduces that risk.
​

RFQ wins on price discovery in the $100k to $10M band, where the spread compression from competitive bidding more than covers the integration cost. Three concrete scenarios.
​

Mid-size ad-hoc conversions. A fintech needs to convert $2M of USDT to USDC once a quarter. A single dealer quotes on relationship terms; a Paradigm RFQ to four makers returns the best of four competing quotes. On a trade of that size the difference is a few hundred to a couple of thousand dollars, which is the number to weigh against the integration work.
​

Cross-stablecoin pairs with multiple natural bidders. USDC to USDT, USDT to PYUSD, USDC to USDC.e: any pair where three-plus makers have inventory on both sides. The RFQ forces makers to compete, and the winner is whoever has the most natural opposite flow that minute.
​

Desks running a price-improvement mandate. Fund administrators and B2B payment platforms with fiduciary or best-execution policies often need documented multi-dealer quotes. Paradigm and Hidden Road produce timestamped quote logs that satisfy SOC 2 and MiFID II best-execution audit trails.
​

DEX onchain wins below roughly $100k, when KYC friction is prohibitive, or when the desk needs atomic onchain settlement without an offchain wire leg. Curve's 3pool and crvUSD pools, Uniswap v4 stablecoin pairs, and aggregators like CowSwap and 1inch fill small USDC/USDT and USDC/DAI trades at a total cost, including gas, that is competitive with any offchain rail at that size. Quote the route before you send it; pool fees and gas both move.
​

Above the small-ticket band, slippage curves bite. A seven-figure USDC to USDT swap on Curve's 3pool prices off the pool's current imbalance, and when the pool is skewed the all-in cost can exceed what an RFQ panel would have quoted. Aggregators (1inch, CowSwap, Matcha) help by splitting across pools and CEX-connected solvers, but rarely beat institutional RFQ above $500k on liquid pairs.
​

The real DEX advantage is composability: a treasury desk can swap and bridge in one transaction, settle into a smart contract, or program conditional execution. For programmatic flows from AI agents, onchain bots, or treasury automations, DEX rails dominate regardless of size.
​

Stablecoin orchestration platforms abstract the rail decision. Instead of a treasury team manually deciding OTC vs RFQ vs DEX, the orchestration layer evaluates ticket size, pair liquidity, latency requirements, and counterparty credit, then routes to the rail with the best net execution.
​

Bridge.xyz (acquired by Stripe in October 2024 for $1.1B) provides orchestrated USD-to-stablecoin conversion with backend routing across OTC desks and onchain pools. BVNK operates similar orchestration for European corporates, layering RFQ across multiple makers. Eco sits at the onchain end, routing stablecoin movement across 15-plus chains with intent-based execution that lets developers express "deliver $X USDC on chain Y" and have the network solve the rail. Conduit, Sphere, and Stripe's stablecoin API offer similar abstractions tuned to specific verticals (B2B cross-border, subscription billing, merchant acceptance).
​

The orchestration layer is the natural home for execution policy: minimum quote count, maximum spread tolerance, preferred counterparty list, and chain preference. For a payments business clearing meaningful monthly volume across USDC, USDT, and PYUSD, the value of the orchestration layer is that it re-picks the rail per trade instead of committing the whole flow to one. See stablecoin settlement throughput benchmarks for how providers scale routing under load, and stablecoin SLA guarantees for the contractual side of provider selection.
​

Use this rough decision tree for stablecoin execution by ticket size and context.

Most institutional treasury desks end up using all four rails. The job is not picking a winner; it is wiring up the policy layer that picks correctly per trade.
​

Provider names and the rails each one operates come from those providers' own public sites. The two acquisitions cited above are linked to the acquirers' own announcements. This guide deliberately publishes no spread figures: no desk, RFQ venue or aggregator in it discloses pricing publicly, and a bps range no reader can check is worse than none. Best-execution references are to SOC 2 Type II controls and MiFID II RTS 27/28 reporting.
​