Maximum Sustainable Daily Volume (Hyper-CLAMM)

Sustainable daily volume ($5–15B), theoretical peak ($20–50B+), limiting factors, and IMO-style multi-dimensional analysis

Overview

The maximum sustainable daily volume that Hyper-CLAMM (the concentrated liquidity AMM engine powering spot markets in the Morpheum CLOB DEX) can realistically support is in the range of $5–15 billion per day** under production conditions in early 2026 architecture, with a **theoretical peak capacity** approaching **$20–50+ billion per day during extreme surges, assuming full sharding utilization and optimal load distribution.

This estimate comes from reengineering the established Hyper-CLAMM design features (from system docs like hyper_clamm_system_design.md, spot-engine.md, phase2-todos.md, and related sharded CLOB patterns), cross-referenced with real-world analogs like Hyperliquid’s own spot/perp activity in January 2026.

---

Capacity model (notation)

Let $N$ be the number of active shards, $v_i$ the sustainable daily notional (USD) handled by shard $i$, and $\bar{v} = \frac{1}{N}\sum_i v_i$ the average per-shard throughput. Total spot notional per day is:

$$V_{\mathrm{day}} = \sum_{i=1}^{N} v_i = N\,\bar{v}.$$

Per-swap work is dominated by $O(\log n)$ tick traversal plus ReClamm glide (exponentiation in virtual balance updates). Write f_swap for sustained swaps per second per shard and $\bar{s}$ for typical trade size (USD). Over $T = 86400\,\mathrm{s}$:

$$V_{\mathrm{comp}} \approx N \cdot f_{\mathrm{swap}} \cdot \bar{s} \cdot T.$$

Sustainable capacity assumes latency, consensus, and glide/oracle margins stay within design envelopes; burst capacity allows temporary elevation of f_swap and $\bar{v}$ at the cost of higher tail latency and throttle risk.

---

Realistic maximum (sustainable, production-ready)

$5–15 billion / day

This is the level at which Hyper-CLAMM remains stable, with sub-100 ms end-to-end latency, bounded slippage $s \in [10^{-4},\,5\times 10^{-4}]$ ($0.01$–$0.05\%$) on most pairs, and no significant degradation in glide / virtual balance adjustment logic.

Why this range?

• Sharding: $N \in [100,\,200]$ (typical from shard-clob.md and clob-system-design.md), each shard executing tick iteration in $O(\log n)$ plus ReClamm glide in under $10\,\mathrm{ms}$ per swap.
• Per-shard throughput: order of $\bar{v} \in [5\times 10^{7},\,1.5\times 10^{8}]$ USD/day (conservative vs major L2 AMM aggregates).
• Total:

$$V_{\mathrm{sust}} = N\,\bar{v} \in [5\times 10^{9},\,1.5\times 10^{10}] \;\text{USD/day} \quad (\text{equiv}\ \$5‐15\,\mathrm{B/day}).$$

• DAG parallelism (MorphDAG-BFT sub-DAGs per shard) keeps consensus overhead low at this scale.

Current real-world benchmark (Jan 2026):

• Hyperliquid spot volume hovers $\sim 3\times 10^{8}‐6\times 10^{8}$ USD/day (e.g. $352M, $325M on different days).
• Broader Hyperliquid L1 DEX aggregates reach $\sim 6\times 10^{8}‐6.4\times 10^{8}$ USD/day.
• Morpheum targets higher headroom via deeper sharding and hybrid CLOB-AMM routing, so $10$–$25\times$ current Hyperliquid spot is a reasonable engineering target:

$$\frac{V_{\mathrm{sust}}}{V_{\mathrm{HL,spot}}} \approx \frac{N\,\bar{v}}{V_{\mathrm{HL,spot}}} \in [10,\,25].$$

---

Theoretical peak capacity (stress / surge scenario)

$20–50+ billion / day (short bursts, 1–3 days)

Let $\beta \gt 1$ be a burst multiplier on per-shard daily notional (caching, batching, CLOB-heavy routing). With N_burst $\gtrsim 200$ and $\bar{v}$_burst $\in [2.5\times 10^{8},\,5\times 10^{8}]$ USD/day:

$$V_{\mathrm{burst}} = N_{\mathrm{burst}}\,\bar{v}_{\mathrm{burst}} \approx \beta\,N\,\bar{v} \;\in\; [2\times 10^{10},\,5\times 10^{10}] \;\text{USD/day}.$$

• Full activation: N_burst $\gtrsim 200$, aggressive caching (in-memory + transient virtual states), SIMD/vectorized batching in reclamm_math.go, and CLOB-hybrid routing for $\gt 80\%$ of volume.
• Per-shard burst: up to $\sim 2.5\times 10^{8}‐5\times 10^{8}$ USD/day during frenzies (analogous to concentrated HIP-3 style sub-markets posting $\mathcal{O}(10^{9})$ USD single-pair volume in short windows).
• Total burst: aligns with top perp DEX days ($\sim 9\times 10^{9}‐1.4\times 10^{10}$ USD perp spikes) or Uniswap-scale monthly peaks annualized to multi–billion-dollar daily bursts.

Let $m$ be the centeredness margin (e.g. ReClamm requires $c \geq m$ with $m = 0.8$ in stress docs) and $\delta$ the oracle deviation circuit-breaker threshold (e.g. $\delta = 0.5\%$). In extreme regimes, throttling engages when glide spam or oracle stress violates:

$$(c \geq m) \;\wedge\; (|\hat{p} - p_{\mathrm{oracle}}| / p_{\mathrm{oracle}} \leq \delta),$$

beyond which V_day is capped toward the $30$–$50\,\mathrm{B}/\mathrm{day}$ band before invariants force slowdown.

---

Key limiting factors and bottlenecks (IMO-style multi-dimensional analysis)

To assert the bound rigorously (stepping back like IMO problem decomposition), each dimension supplies an inequality on V_day.

1. Computational dimension

Per shard, feasible sustained swap rate f_swap $\in [10^{4},\,10^{5}]\,\mathrm{s}^{-1}$ (Go + fine-grained locking) and typical trade size $\bar{s} \in [10^{3},\,10^{4}]$ USD. The raw notional flux per shard is q = f_swap$\,\bar{s}$ (USD/s), so:

$$V_{\mathrm{comp}}^{\mathrm{raw}} = N\,q\,T = N\,f_{\mathrm{swap}}\,\bar{s}\,T.$$

In principle V_comp^raw is enormous at these f_swap and $\bar{s}$; the operational bound on V_day is not set here but by GC pauses, cache coherence, and the $O(\log n)$ + glide serial path, which cap realized throughput orders of magnitude below naive f_swap$\,\bar{s}$ — leaving headroom in the \10^‐$10^$/day engineering range before other dimensions dominate.

2. Consensus / async complexity dimension

With sub-DAGs per shard, throughput scales nearly linearly in $N$ until greedy shard assignment skew or coordinator limits appear. Let $\alpha(N)$ be effective throughput efficiency ( $\alpha \approx 1$ for $N \lesssim 200$–$300$ ). Cross-shard $2$PC introduces abort probability p_abort($N$); staying under p_abort $\lt 10^{-2}$ keeps:

$$V_{\mathrm{cons}} \;\lesssim\; \alpha(N)\,N\,\bar{v}, \qquad \alpha(N) \to \alpha_{\infty} \lt 1 \quad \text{where} \quad N \to \infty,$$

yielding a soft cap in the \3\times 10^‐$5\times 10^$/day range before *p*_abort exceeds$1%$.

3. State / storage dimension

Let $\Delta S$ be compressed state delta per day. For V_day $\sim 5\times 10^{10}$ USD, engineering targets keep:

$$\Delta S \;\lesssim\; 10\,\text{GB/day}$$

(compressed), so persistence and indexing remain bounded.

4. Economic / MEV dimension

Hook fees and MEV-aware routing imply an increasing effective cost per unit notional at extreme V_day, producing self-throttling in a band compatible with \2\times 10^‐$4\times 10^$/day before LP and arb externalities dominate.

5. Real-world analog dimension

• Hyperliquid perp peaks: $\sim 9\times 10^{9}‐1.4\times 10^{10}$ USD/day (single chain, not fully sharded like Morpheum).
• Uniswap V3 monthly records imply $\sim 3\times 10^{9}‐4\times 10^{9}$ USD/day burst equivalents.
• Morpheum sharding and gasless DAG imply a multiplier $\kappa \in [5,\,15]$ on analog single-chain spot figures, consistent with V_sust and V_burst above.

---

Summary table

Scenario	Daily volume range	Conditions / triggers	Confidence level
Current production (early 2026)	$0.5–2B / day	50–100 shards, typical load	Very high
Sustainable maximum	$5–15B / day	Full 100–200 shards, optimized glide/hooks	High
Theoretical peak (surge)	$20–50B+ / day	200+ shards, CLOB-hybrid >80%, burst tolerance	Medium–high
Hard invariant break risk	>$50–100B / day	Oracle deviation >0.5%, glide spam, state bloat	Low (throttled)

---

Summary

Hyper-CLAMM is therefore engineered to comfortably handle top-tier spot DEX volumes seen in 2026 (far beyond current Hyperliquid spot $\sim 3\times 10^{8}‐6\times 10^{8}$ USD/day), with headroom for $10$–$50\times$ growth as Morpheum adoption scales. The sharded, DAG-optimized, ReClamm-virtual design gives it a structural advantage over monolithic or EVM-based AMMs in this volume regime.