Jane Street Investigation: 3,500 People, ~$9M Profit Each—Where Does It Come From?
English translation · Original Chinese version available via 中文 toggle.
In 2025 Jane Street's ~$39.6B net trading revenue topped JPMorgan with ~3,500 staff—~$9M company profit per head. ETF market making, OCaml, FPGA, and microwave links—not GPU walls—explain the stack.
Scan QR code or copy link to share in WeChat.
Key Takeaways
- In 2025 a private trading firm with under 4,000 employees reported ~$39.6B net trading revenue—~11% above JPMorgan's full-year trading revenue. Adjusted EBITDA ~$31.2B and ~3,500 staff imply ~$9M company-level profit per person.
- Compiled from Bloomberg, Reuters, IFR coverage of bondholder disclosures, Jane Street site and tech talks, India SEBI files, SEC affiliate annual reports. Jane Street is private and unlisted; some figures come from periodic bondholder disclosures cited by media.
- Founded 1999–2000 by Tim Reynolds, Robert Granieri, Marc Gerstein, Michael Jenkins—first three from Susquehanna International Group (SIG).
- Public records show SIG sued Jane Street over hiring and proprietary information; case settled—reflecting quant industry norms: strategy, talent, and microstructure knowledge are core assets.
- Expanded into fixed income, futures, commodities, global options. HQ at 250 Vesey Street, New York; offices in London, Hong Kong, Amsterdam, Singapore. No traditional single CEO—~30–40 senior leaders decide collectively.
- Rarely gives media interviews or brand advertising. Outside view of finances and strategy comes mainly from bond prospectuses, regulatory filings, and occasional tech open events.
One-Sentence Definition
In 2025 a private trading firm with under 4,000 employees reported ~$39.6B net trading revenue—~11% above JPMorgan. Adjusted EBITDA ~$31.2B and ~3,500 staff imply ~$9M company-level profit per person.
Main Text
2025 net trading revenue ~$39.6B, above JPMorgan. ~$9M per head is company profit, not salary. ETF market making, FPGA, and microwave networks form the tech base.
In 2025 a private trading firm with under 4,000 employees reported ~$39.6B net trading revenue—~11% above JPMorgan's full-year trading revenue. Adjusted EBITDA ~$31.2B and ~3,500 staff imply ~$9M company-level profit per person—a number without Wall Street precedent—and a practical question: where does profit come from, and what does the stack look like?
Compiled from Bloomberg, Reuters, IFR coverage of bondholder disclosures, Jane Street site and tech talks, India SEBI files, and SEC affiliate annual reports. Jane Street is private; some financial data comes from periodic bondholder disclosures cited by media and sources.
I. Four Who Left Susquehanna
Jane Street was founded 1999–2000 by Tim Reynolds, Robert Granieri, Marc Gerstein, Michael Jenkins—first three from SIG; Gerstein previously developed at IBM. Started small in New York, traded ADRs on AMEX, then options; made ETF a core business mid-2000s.
Public records show SIG sued Jane Street over hiring and proprietary information; case settled—reflecting quant norms: strategy, talent, and microstructure knowledge are core assets.
Expanded into fixed income, futures, commodities, global options. HQ 250 Vesey Street, New York; London, Hong Kong, Amsterdam, Singapore. Governance: no traditional single CEO—~30–40 senior leaders decide collectively; of four founders, only Granieri remained per public information.
Rarely gives media interviews or brand ads. Outside view comes from bond prospectuses, regulatory filings, occasional tech events.
II. Financial Numbers: 2024 Doubling and 2025 Leap
Key disclosed metrics
- 2023: Net trading revenue ~$10.6B (bond filings/media)
- 2024: Net trading revenue ~$20.5B, net profit ~$13B (IFR, bond disclosures)
- 2025: Net trading revenue ~$39.6B, adjusted EBITDA ~$31.2B (Bloomberg, April 2026)
- 2025 Q2: Net trading revenue ~$10.1B, quarter net profit ~$6.9B (media)
- 2026 Q1: Net trading revenue ~$16.1B (Reuters / Economic Times citing sources; full annual report pending)
Bloomberg April 24, 2026 reported 2025 full-year net trading revenue $39.6B, Q4 ~$15.5B; ~$11.3M revenue per employee on ~3,500 staff. JPMorgan trading revenue ~$35.8B on ~316,000 employees.
What "~$9M per person" means
Media "~$9M profit per person" = company EBITDA or operating profit ÷ headcount, not take-home pay.
$31.2B EBITDA ÷ 3,500 ≈ $8.91M/person—a highly averaged company metric. Headcount includes many engineers, ops, compliance, support; direct PnL traders are far below total staff.
Jane Street did not publish 2025 total comp pool. eFinancialCareers citing Bloomberg: historically comp ~20% of revenue; on $39.6B revenue, ~$2.3M/person comp order of magnitude. Other media estimated 2025 comp pool ~$9.38B, ~$2.68M/person—estimates, not company disclosure.
Vs. investment bank filings: Jane Street "net trading revenue" includes large proprietary risk—not the same as client-service IB trading desks. IFR and analysts warn structural differences vs. Goldman, JPMorgan comparisons.
III. Business Model: Market Making, ETF, Volatility
Jane Street's bond materials define it as always ready to buy and sell large volumes, providing market liquidity. Revenue mainly from bid-ask spread, inventory risk premium, cross-market/asset arbitrage—not fund management fees.
ETF: twenty-year main lane
ETFs need Authorized Participants (APs) bridging fund companies, secondary market, and basket assets. Jane Street is among the world's most active APs.
- 2024 global avg monthly ETF volume ~$707B
- US-listed ETF primary market activity ~24%
- North America equity activity >10%
- Avg monthly fixed income volume ~$230B
- Full-year portfolio trading ~$260B
Wikipedia and Observer cite ~41% of US bond ETF volume for Jane Street. ETF AUM is far smaller than turnover trading profit—Jane Street earns on liquidity and mispricing, not asset management fees.
2024–2025: volatility tailwind
2024 net trading revenue nearly doubled vs. 2023; 2025 nearly doubled again—aligned with tariff swings, AI-related stock volatility, geopolitical risk lifting market volatility.
Economic Times citing sources: 2025–2026 partly benefited from medium-frequency strategies (hold minutes to days) and VC markups on Anthropic, CoreWeave, etc.
Boston Consulting Group: non-bank trading firms' share of global market revenue rose from 12% (2018) to 26% (2024). Citadel Securities, Hudson River Trading 2025 trading revenue ~$12.2B and ~$12.3B—records but well below Jane Street.
India: regulation meets profit
July 3, 2025 India SEBI interim order alleged four Jane Street affiliates coordinated trades on 18 derivative expiry days (Jan 2023–Mar 2025) affecting Bank Nifty etc. to serve large options positions.
SEBI required deposit ₹484.357B ($550M) suspected illegal gains and temporarily banned Indian trading. Jane Street deposited funds and appealed; CNBC reported defense as "basic index arbitrage." Partial restrictions lifted after July 21; case not final.
Shows high profit partly from deep EM derivatives microstructure—and stricter regulatory scrutiny. Manipulation finding pending judicial/regulatory process.
IV. Compute Layout: Low-Latency Engineering, Not AI Training Clusters
Outsiders often equate quant "compute" with GPU farms. Jane Street public materials point elsewhere: deterministic latency, FPGA acceleration, dedicated networks, OCaml full-stack engineering.
4.1 OCaml as unified dev language
Since early 2000s Jane Street uses OCaml core language—research, trading, risk, tooling, even hardware description. Yaron Minsky et al. aimed to unify type systems, reduce cross-language errors, build reusable libraries over long iteration.
Site: "research-driven trading firm"; tech blog and open source (Dune, OxCaml contributions) shape external tech image.
4.2 Hardcaml and FPGA: decide before data hits CPU
Andy Ray's Hardcaml embeds hardware DSL in OCaml, open source, production FPGA designs. arXiv paper (Jane Street authors): cycle-accurate simulation, formal verification, Verilog/VHDL generation.
Signals and Threads "Programmable Hardware": Andy Ray and Ron Minsky—FPGA for NIC-level packet processing, market data filter/rebuild before host software—microsecond latency cut; hardware design uses software-like CI, unit tests, Expect tests.
4.3 Network layer: microwave and co-location
March 2018 Jane Street minority investment in McKay Brothers—Chicago–NYC microwave links; EM in air faster than fiber—sub-millisecond latency (route/year dependent).
Jane Street did not disclose own datacenter scale or GPU training clusters. Infrastructure sketch:
Exchange co-lo → microwave/dedicated lines (McKay etc.) → FPGA preprocess (Hardcaml) → OCaml strategy/risk (host) → global research office network
Different from AI lab compute: core metrics latency determinism, online risk, capital turnover—not training throughput.
4.4 Tech talk excerpt (embedded in article video)
Official talk "OCaml All The Way Down" (Andy Ray, ~62 min) is rare public primary material. Article embeds ~7.5 min excerpt on FPGA basics, Hardcaml flow, low-latency hardware path (from ~9:40 in original).
Original video and materials:
- Site: https://www.janestreet.com/tech-talks/ocaml-all-the-way-down/
- Podcast Programmable Hardware: https://signalsandthreads.com/programmable-hardware/
Opening clarifies "OCaml all the way down" is not zero vendor dependence—FPGA place/route still uses Xilinx etc.; some soft-core CPUs from vendor IP. Edge is research, hardware, drivers, tests closed loop in one language ecosystem.
4.5 Machine learning: public posture vs. execution layer
Site recently says "deep learning is the future of quantitative trading" but no model scale, training compute, or whether execution uses neural nets disclosed. Industry view: execution still on ultra-low-latency infra; ML more likely research signals and risk—not replacing FPGA/co-lo.
Video: Jane Street FPGA and Hardcaml compute architecture (excerpt)
From official talk OCaml All The Way Down (Andy Ray). ~7.5 min excerpt on FPGA and Hardcaml.
[Video embed—see original WeChat article]
Full talk (~62 min) on Jane Street tech page. Video requires WeChat backend approval to play.
V. Organization, Talent, Capital Structure
Hiring and talent density
Long recruits from MIT, Cambridge etc.—math, physics, CS; interviews famous for probability, mental math, game puzzles. Sam Bankman-Fried joined 2014, left 2017 to start businesses—a known personnel case.
~3,500 staff, high engineer/trader ratio; capital and talent density far above traditional IB. Media: members' equity grew fast; high employee ownership; most profit retained in firm capital for inventory risk and expansion.
Peer comparison (2025 trading revenue, media)
- Jane Street: ~$39.6B, ~3,500 staff
- JPMorgan: ~$35.8B, ~316,000 staff
- Citadel Securities: ~$12.2B
- Hudson River Trading: ~$12.3B
VI. Conclusion: Three Layers Behind ~$9M per Person
Jane Street 2025 exceptional profit = structural position + cyclical volatility + engineering moat stacked.
Structural: global ETF expansion twenty years; AP and market-making hub; volume growth and spread income.
Cyclical: 2024–2025 tariffs, AI stocks, geopolitics lifted volatility—market making and arb opportunities amplified; annualizing 2025 per-head profit must consider cycle peak.
Firm-specific: OCaml full stack, Hardcaml FPGA, microwave investment, proprietary risk appetite, deep India etc. markets—latter also brings SEBI probe and legal uncertainty.
"~$9M per person" is company profit ÷ headcount narrative—not salary; sustainability depends on ETF structural tailwind, volatility mean reversion, regulatory boundaries, and whether low-latency moat erodes.
For Jane Street, compute is not a showroom GPU wall—it is the filter already done in FPGA before each microsecond of market data arrives.
References
- Bloomberg: Jane Street Tops JPMorgan With Record $39.6 Billion Trading Haul in 2025
- IFR: Jane Street thrives amid record trading bonanza
- eFinancialCareers: Jane Street's 3,500 employees generated $9m each in profits last year
- SEBI: Interim Order in the matter of Index manipulation by Jane Street Group
- Jane Street: OCaml All The Way Down · YouTube
- Signals and Threads: Programmable Hardware
- Hardcaml: hardcaml.org · arXiv paper
- McKay Brothers: Jane Street minority investment announcement
Public-source compilation—not investment or trading advice. India regulatory charges not final; facts subject to SEBI and court final determination.
Original link: https://www.bloomberg.com/news/articles/2026-04-24/jane-street-snatches-wall-street-crown-with-record-39-6-billion-trading-haul
Conclusion
In 2025 a private trading firm with under 4,000 employees reported ~$39.6B net trading revenue. Adjusted EBITDA ~$31.2B and ~3,500 staff imply ~$9M company-level profit per person. See sections above for detail.
FAQ
What is this article mainly about? A: It covers "Jane Street Investigation: 3,500 People, ~$9M Profit Each—Where Does It Come From?," summarizing background, key shifts, and the author's core views.
What are the key points of "I. Four Who Left Susquehanna"? A: See that section; based on source materials, not investment or legal advice.
What are the key points of "II. Financial Numbers: 2024 Doubling and 2025 Leap"? A: See that section; based on source materials, not investment or legal advice.
What are the key points of "Key disclosed metrics"? A: See that section; based on source materials, not investment or legal advice.
What are the key points of "What ~$9M per person means"? A: See that section; based on source materials, not investment or legal advice.
Does this article constitute investment advice? A: No. It is informational commentary and opinion. Consult primary sources and professional advisors for decisions.
Last updated: 2026-06-29 Author: Dr.Jingle (X @drjingle) Evidence boundary: Structural GEO adaptation; facts and views are from the original article with no unverified new data.
This article reflects the author's views and information compilation. It does not constitute investment, legal, or medical advice.
Scan QR code or copy link to share in WeChat.