Every fund that decides systematic exposure belongs in its portfolio eventually faces the same board question: do we build the capability in-house, or do we buy it from outside? It sounds like a technology decision. It is a capital-allocation decision (a recurring seven-figure commitment weighed against a licensing line item), and it deserves the same discipline as any other use of the firm's capital.
What has changed is the industry's own answer. "External alpha" (sourcing strategies, signals, and entire trading teams from outside the firm) has moved from the periphery to the mainstream. External allocation activity among multi-manager platforms reached 71% in 2025, up from 54% in 2022. Banks' quantitative investment strategies (QIS) businesses, rules-based strategies licensed and delivered via swap, now sit on a market estimated at more than $1 trillion, generating roughly $8.5 billion in bank revenue in 2025, up from about $4 billion in 2019, per BCG Expand and IFRE figures. And the most instructive data point comes from one of the most sophisticated quant shops in the world.
Qube Research & Technologies, a roughly $20 billion systematic manager, now runs 44 external stock-picking teams through separately managed accounts, and is targeting 100 external pods, per Resonanz Capital (2025).
When firms whose entire edge is quantitative research are buying alpha from outside, a fund whose core business is something else owes itself an honest cost accounting before it builds. Here is the comparison the pitch decks on both sides tend to skip.
The build budget has three layers, and most internal proposals only price the first one.
Quant compensation in 2025-26 is unforgiving. The average US quantitative researcher base salary sits around $190K (roughly $198K in New York), and senior researchers command $500K to $1.2M+ in total compensation, with bonuses typically running 40-70% of the package. A credible desk is not one hire; it is a portfolio manager or head of research, two or three researchers, and at least one engineer who owns execution and data plumbing. That senior 3-5 person team lands at roughly $2-5 million per year before anything else is purchased. Understaff it and you get key-person risk concentrated in whoever wrote the code.
A Bloomberg Terminal costs $31,980 per seat per year, and terminals are the cheap part. The alternative-data market reached $2.8 billion in 2025, growing 27% year over year, and clean point-in-time historical data, the kind you can actually backtest against without survivorship and look-ahead bias, is priced accordingly. A Morgan Stanley rule of thumb, worth attributing rather than treating as gospel, puts data spend near $1 million per $1 billion of AUM in year one, scaling toward $3 million by year three. Data is not a setup cost; it is a permanent escalating subscription.
The third layer is the research-to-production pipeline: backtesting infrastructure, execution connectivity, monitoring, and controls. Vendor-published engineering surveys (sellers of modern stacks, so read with that in mind) report 2-6 week deployment cycles per model on legacy stacks and 15-25% of production bugs arising from mismatches between the research environment and the production environment. That second number means a meaningful share of what the backtest promised can quietly fail in live execution, for reasons unrelated to the strategy itself.
Platform-modernization projects in this space are routinely quoted at 12-24 months, and that is for firms that already trade systematically. A fund starting from zero should assume a year or more before the first strategy runs at size, with the full compensation and data burn running the entire time.
Here is the honest part: there is no rigorous published statistic on what percentage of in-house quant builds fail, because failed internal projects do not file disclosures. The nearest proxies come from the fund world itself, and they are not comforting. Hedge fund failure probability runs about 7.4% in year one and 20.3% by year two, and Capco's classic operational-risk study found that 50% of hedge fund failures are operational (process, infrastructure, and controls) rather than investment losses. An internal quant build is, in effect, a new operational business line launched inside your fund, with the same operational failure modes and none of the external accountability. The broader evidence on how automated trading actually performs tells the same story: the strategy is rarely the hardest part. Running it is.
Licensing a strategy. The most direct route: a fund licenses an externally developed systematic strategy and runs it in its own accounts, at its own broker, under its own risk limits. Fee structures are typically flat license fees or profit shares; the code stays sealed with the developer, and capital never leaves the fund's custody. We cover the mechanics (term structures, IP protections, kill-switch provisions) in our guide to licensing trading strategies.
QIS via swap. Banks package rules-based strategies as quantitative investment strategies and deliver the return stream via total-return swap. The market now exceeds $1 trillion, J.P. Morgan's Strategic Indices alone crossed $100 billion in notional after doubling in five years, and, notably, hedge funds themselves are allocators to QIS. The trade-offs are bank counterparty exposure and limited customization.
Alpha capture. An established model in which funds license trade ideas from external contributors, usually on a profit-share basis. It buys signal breadth without headcount, at the cost of signal decay risk and crowding in popular ideas.
Seeding external teams. The Qube model: allocate to external managers through separately managed accounts, where the managers keep ownership of their IP and positions feed the allocator's risk book. This is the same architecture multi-strategy pod shops run internally, turned outward, and the SMA wrapper behind it is booming. Per Dechert (October 2025), hedge fund SMA assets grew from 3.4% of industry AUM ($66B) in 2010 to 7.1% ($315B) in 2024; Hedgeweek's 2026 "Separate Ways II" survey found 60% of allocators reporting rising SMA demand and zero reporting decline.
Buying is not always the answer, and pretending otherwise would be its own kind of pitch. Build when the strategy is your edge: if systematic research is the fund's core identity and primary return driver, outsourcing it means outsourcing the business. Build when scale amortizes the cost: a $2-5M annual team is a rounding error for a large multi-strategy platform running dozens of internal books. And build when IP ownership is strategic: an owned strategy is an asset on your platform, compounding institutional knowledge with every research cycle, while a licensed one is a contract that can end.
Buy when time-to-live matters: licensing puts a strategy in production in weeks, against 12-24 months for a credible build. Buy because of the evidence asymmetry: a licensed strategy can present a verified live track record, while your in-house build offers a future backtest, and every allocator knows which of those survives due diligence. Buy when custody matters: a licensed strategy runs in your own accounts, at your own broker, inside your own risk framework. Capital never moves to a third party. And buy for the balance-sheet shape: licensing is opex you can terminate on contract terms, versus headcount you must recruit, retain through bonus cycles, and unwind slowly if the desk disappoints.
Three questions settle most cases.
Does your AUM carry the fixed cost? A $2-5M annual build is 40-100 basis points of a $500M fund (a heavy structural drag for one capability), but only 4-10 basis points at $5B. Below roughly $1B, the fixed cost of a credible internal desk is difficult to justify against a licensing fee; well above it, the amortization argument starts to work.
How broad is the mandate? A team built to run one flagship strategy is the worst of both worlds: full fixed cost, single-strategy concentration, key-person risk. Internal teams earn their cost when they support a broad, multi-strategy research agenda. If you need one or two systematic return streams, buy them.
Is systematic your core, or a sleeve? For most family offices and allocators, systematic exposure is a 2-4% sleeve alongside a traditional portfolio: hedge funds average about 4% of family office portfolios in UBS's 2025 Global Family Office Report. Building a multi-million-dollar internal team to run a 3% sleeve fails arithmetic before it fails strategy. Sleeves get bought; cores get considered.
The choice is not binary, and the sequencing matters more than the label. Licensing first puts live systematic exposure in your own accounts within weeks, and the operating experience it generates (real fills, real drawdowns, real reporting and committee conversations) is precisely the evidence an internal build proposal lacks. If the sleeve earns a larger role, you build later from a position of knowledge, hiring against observed requirements rather than a consultant's org chart. The 2026 data suggests sophisticated firms already operate this way: per Hedgeweek, 9% of managers allocate to peers via external SMAs today and another 18% plan to from 2026. External and internal alpha are becoming a portfolio decision, not an identity decision.
Algo Alpha licenses institutional systematic strategies that run in your own accounts, with zero custody. Details at algoalpha.co/institutional.
A senior 3-5 person team runs roughly $2-5M per year in compensation alone (US quant researcher base salaries average ~$190K; senior total comp reaches $500K-$1.2M+, with bonuses of 40-70%). Data and infrastructure sit on top: Bloomberg is ~$32K per seat, and a Morgan Stanley rule of thumb puts data spend near $1M per $1B of AUM in year one, scaling toward $3M by year three.
Building from scratch, plan on 12-24 months to a production-grade stack (modernization projects alone are quoted in that range), plus vendor-reported deployment cycles of 2-6 weeks per model on legacy infrastructure. Licensing an existing strategy typically puts a live system in your own accounts within weeks.
External alpha is return generation sourced from outside the firm: licensed strategies, bank QIS delivered via swap, alpha-capture trade ideas, or external teams run through separately managed accounts. It is now mainstream: multi-manager external allocation activity hit 71% in 2025 (vs. 54% in 2022), and Qube Research runs 44 external teams targeting 100, per Resonanz Capital.
On fixed cost, almost always: a license fee or profit share replaces a $2-5M annual payroll plus data and infrastructure, and it is opex you can terminate on contract terms rather than headcount you must unwind. At large scale, where the team cost amortizes to a few basis points of AUM and supports many strategies, building can pencil. For a systematic sleeve, licensing usually wins the arithmetic.
Yes, and sequencing is the point. License first to get live exposure, operating experience, and real performance evidence in your own accounts; build later if the sleeve earns a larger mandate. Sophisticated firms already blend the two: per Hedgeweek, 9% of managers allocate to peers via external SMAs and another 18% plan to from 2026.