Company
About Careers Events Technology Research Performance Syphon OS Syphon Evo Syphon Connect White Label Syphon Aether
Sign In Request Access
Think Different · Trade Different

Institutional throughput, minimal infrastructure

$93,550,000,000 traded on three machines — driven by autonomous AI agents.

Throughout the competition, our matching engine ran on deliberately constrained infrastructure — just three machines with 72 vCPUs, which processed $93,550,000,000 in notional volume and 1.57M execution reports, driven by 242 autonomous trading agents built and deployed by participants.
Platform-layer CPU never exceeded 9%. That's not the limit — that's headroom.

$0
Cumulative Traded Notional
3D abstract visualization
scroll
AI-Native

Not a gimmick — participants built and deployed genuinely autonomous trading agents.

The agents participants built made their own decisions, called tools, placed orders and ran risk control autonomously on the platform — the figures below are real activity captured in live production.

0M tokens
Total LLM Token Consumption
234,422,327 tokens — the total compute agents consumed in autonomous reasoning, decision-making and tool orchestration, all within a real trading loop.
0
Live AI Strategies
Autonomous trading agents deployed in production
0
Autonomous Tool Calls
AI-initiated calls to trading / analytics tools
0
Risk-Control Events
Autonomous risk-monitoring events
0
Agent Sessions
Total agent runtime sessions
0
Agent Run Steps
Total agent execution steps
0+
Scheduled Autonomous Runs
Scheduled autonomous batch runs
Throughput

What the engine actually processed.

Cumulative figures measured throughout in production (6/21 21:00 → 6/26 21:00 UTC, 5 trading days) — notional volume, matched trades and full-lifecycle execution reports, plus the WAL records used for state recovery and consistency auditing.

$0
Cumulative Notional
Cumulative traded notional ≈ $93.55B (converted to USD at fill rates)
25%50%75%100%
0
WAL Apply Records
State recovery and consistency auditing — recoverable, traceable, auditable
0
Execution Reports
Execution reports · carrying order status and fill facts
0
24h Trades
Trades in the last 24 hours
0
Trades Matched
Cumulative matched trades · 793,985
$0
24h Notional
Notional in the last 24 hours · peak competition capacity
Peak Load

The load is spiky — and the engine absorbs it.

The highest instantaneous rates appeared during the pre-close surge. The real test isn't sustained pressure — it's these instantaneous spikes. And platform-layer CPU stayed below 9% throughout.

0msg/s
Peak Exec Reports / s
Execution reports peak (per second)
6/25 08:49:09 UTC
0trades/s
Peak Trades / s
Trades peak (per second) · pre-close surge
6/26 20:50:29 UTC
0/min
Peak Exec Reports / min
Peak minute (exec reports / 679 trades)
6/26 20:00 UTC · final hour before close
0trades/s
Instantaneous peak
95 trades/s
0trades/s
Peak-minute average
679/min ÷ 60 · 6/26 20:00
0–30×peak / avg
spike ratio

Peak rates run roughly 25–30× the active average — the load is extremely spiky. What the system really has to absorb is the instantaneous spike, not sustained saturation.

Engineering by Constraint

We didn't scale the infrastructure up — we constrained it on purpose.

Real engineering strength isn't "we can hold up if we add machines" — it's "we have ample room without adding any." This competition was a public stress test: we held the infrastructure well below industry norms and let the engine stand on its own, letting the results speak.

Industry Default
Throw more servers at the load.
Autoscaling, redundant clusters, peak-provisioned capacity — cost rises linearly with scale, performance is masked by hardware, and you can't see the engine's true efficiency.
Our Approach
No autoscaling. No extra machines. Just the engine.
A single 16-vCPU platform machine carried the entire competition frontend, native matching, leaderboard and cache; the trading engine completed matching and hedging on just 2 nodes. Result: ample headroom.
Hardware Footprint

This is the entire footprint behind it.

All production-measured instance types. The platform layer carried the whole competition system on a single machine; the trading engine completed matching and hedging on a minimal topology, running low throughout.

A Platform Layer
Full-stack competition systemFrontend · native matching · GameFi · leaderboard · cache · analytics
INSTANCEc7i.4xlarge
vCPU / RAM16 / 32 GB
COUNT×1
CPU PEAK< 9%
CPU AVG/DAY~4%
Peak CPU utilization9% / 100%
B Engine
Trading engine — primary nodeMatching / hedging — primary
vCPU / RAM24 / 251 GB
SERVICES15 services
ROLEPrimary
LOADrunning low
Load levellow · ample
C Engine
Trading engine — secondary nodeMatching / hedging — secondary
INSTANCEm7a.8xlarge
vCPU / RAM32 / 126 GB
SERVICES8 services
LOADrunning low
Load levellow · ample
0 machines
|
0 total vCPU
|
carried $0 notional
Efficiency Benchmark

Same execution workload — an order of magnitude less infrastructure.

Our measured footprint placed side by side with the typical industry scale of a regional commercial bank's core execution system — comparable throughput, yet a stark gap in hardware footprint and total cost of ownership.

Hardware at a glance
Bank side is typical industry scale · illustrative comparison
Metric Syphonix (measured) Regional commercial bank (typical industry scale)
Core execution servers
3 servers (72 vCPU)
full measured footprint
20–80 production servers
excluding disaster recovery
Deployment architecture
Single production topology
lightweight deployment
Multi-tier cluster + HA + DR
high availability · geo-redundant DR
Platform TCO
≈ $386 / 5 days
measured · ≈ $3.2/hr
≈ $20k–100k / 5 days
typical industry scale · illustrative
Peak CPU utilization
< 9%
16-core platform, measured
Large redundant capacity
provisioned for peak load + failover
Throughput
200 msg/s · 95 fills/s
single minimal topology
Multi-node cluster
comparable workloads shared across nodes
Infrastructure efficiency
Institutional execution
on a minimal hardware footprint
Prioritizes reliability & compliance
HA and regulation over utilization
About $386 of equivalent cloud compute carried 5 days of institutional-grade execution load.
Syphonix cost is measured on 72 vCPU of equivalent compute (linearly scaled from the public AWS c7i.4xlarge on-demand price of $0.714/hr per 16 vCPU); the bank side reflects typical industry scale for a regional commercial bank's core execution system and is illustrative only, not representative of any specific bank's actual production cost. Neither side includes storage, networking, HA / DR, monitoring, software licensing, or operational and compliance costs.
Participant Roster

Elite university quant talent, head to head.

From Cambridge, Oxford, Imperial, UCL, KCL and 41 universities in total, alongside hedge-fund practitioners — a high-density, high-caliber real arena.

0
Universities
Featured Institutions
Cambridge Oxford Imperial UCL KCL + 36 more
Joined on the same field by professional practitioners from hedge funds — competitive intensity benchmarked against a real buy-side environment.
Execution Quality

Real fills — not simulated.

Beyond scale, what matters more is quality: every trade was processed and completed by a real external LP, with a substantial size per trade.

0%
External LP
fill completion rate
LP fill rate
$0K
Average
trade size
avg trade ≈ $67K
0trading days
Continuous live
run, no interruption
6/21–6/26 UTC
Daily Ramp

Volume doubled day over day, then spiked at the close.

Daily trade counts by trading day (21:00 UTC / NY 17:00 boundary). From 63K trades on day one all the way up to 280K on the final day — the last day, layered with the final hour before close, set both the trades peak and the per-minute peak: a textbook end-of-competition sprint.

Trades / Day

Unit: trades · trading-day boundary 21:00 UTC
62,942
Day 16/21→22
56,109
Day 26/22→23
147,199
Day 36/23→24
247,834
Day 46/24→25
279,881
Day 56/25→26 close
Day 1 63K trades → final day 280K trades 5-day growth ≈ 4.4×
The Takeaway

Scale went up. Cost stayed flat.

$0
notional
on
0
vCPU · 3 machines
  • 01$93,550,000,000 notional traded, 1.57M execution reports, 100% real LP fills.
  • 02Driven by 242 autonomous trading agents and elite university quant participants.
  • 03All running on 3 deliberately under-provisioned machines, platform CPU peaking below 9%.
  • 04Instantaneous peaks ran 25–30× the average, and the system absorbed them smoothly.
  • 05The same engine, with no rearchitecture and no scaling out, has multiples — even tens of times — of vertical headroom.

$93,550,000,000 notional, 1.57M executions, 100% real LP fills — driven by autonomous AI agents, on 3 deliberately under-provisioned machines with platform CPU never crossing 9%. The same engine has multiples of headroom, no rearchitecture required.