Going independent in 2026

Why a one-person shop is conceivable now: two of the three arenas have low access barriers and gettable data, and AI tooling has collapsed the cost of building the stack. Hard and capital-constrained, but real.

The idea

Can you realistically do this alone in 2026? (the honest verdict)

It depends entirely on the market. In equities and futures, no: the fixed-cost wall (colocation, proprietary data, microwave links, the consolidation into giants) is insurmountable for an individual. In crypto and prediction markets, yes, genuinely: open data, your own infrastructure, no membership or PFOF moat. The honest answer is market-specific, and most of this page is that breakdown.

Intuition first: the question is not "are you smart enough" (many readers are). It is "does the venue you want to trade have a fixed-cost wall, and can you clear it?" In the classic arenas the wall is the whole point of the economics page: a moving break-even volume that has already closed the door. In the open arenas the wall was never built. We refuse both lies the internet tells you. The hype lie ("anyone can run an HFT bot and print money") ignores the wall and the edge. The doom lie ("it is hopeless, the giants own everything") ignores that the giants own the classic venues, not the open ones. The truth is a map, not a slogan.

The structure of the rest of the page: first the market-by-market verdict (where the door is open), then what AI actually changes (and does not), then the clean separation of what you genuinely need from what is now cheap. By the end you should be able to point at a market and say honestly whether it is worth your year. For the segment-by-segment companion to this market-by-market view, see is HFT still profitable in 2026.

Equities and futures: effectively closed (and why)

For a newcomer, the classic lit venues are closed. Competing means colocating next to the matching engine, licensing proprietary depth-of-book data, renting microwave routes between venues, and out-running firms that have spent two decades and tens of millions on exactly that. The break-even volume is now beyond any individual. You would be the slow money, someone else's profit.

Intuition first: in equities and futures the edge that remains at the top is latency and scale, and both are bought, not coded. A solo operator with a brilliant model and a cloud server is, structurally, the stale-quote target the colocated firms pick off. The model does not matter if you arrive a millisecond after the people you are trading against. The specific walls, each a page elsewhere on the site: colocation and FPGA (you must be in the building, on the fast hardware); market-data fees (proprietary feeds plus non-display licences, a substantial fixed cost before you trade once); the consolidation (a handful of giants who define the speed frontier); and regulation (Reg NMS, MiFID II, with membership, reporting and registration overhead). Any one is daunting; together they are a moat.

The honest exceptions, kept honest: there are slower niches in these asset classes (longer-horizon statistical arbitrage, event strategies that are not a pure speed race) where a small, smart operator can play. But that is quant trading, not HFT, and it competes on research, not latency. If your edge is speed, equities and futures are closed. If your edge is a slower signal, you are in a different, broader contest, and even there, data and execution cost real money. For the readiness map below: Capital amber, Data red, Infrastructure red, Edge red-for-speed and amber-for-slow-signal.

Crypto: a genuine opening (and where it is hard)

Crypto is the real opening. Venue data is free over public WebSocket and REST APIs, there is no colocation or exchange membership to buy, your infrastructure is ordinary cloud, and many venues run the same maker-taker rebate model you would profit from. A small team can run the full research-to-production pipeline end to end. The fixed-cost wall that protects equity incumbents is mostly absent.

Intuition first: everything that makes equities closed is missing in crypto. No proprietary data tax; the order book streams free. No colocation arms race you can lose before you start; latency matters but the playing field is cloud, not a microwave dish you cannot afford. No prime broker or membership gate. The same microstructure maths (microprice, inventory management, adverse selection) transplants directly onto a venue you can actually access. See market making in crypto.

Where it is genuinely hard, stated plainly: it is 24/7 (no overnight risk-off; your kill-switch and monitoring must be real, see kill-switches); the flow is toxic (you are trading against other bots and informed whales with thin retail cushioning); venues carry operational and counterparty risk (outages, withdrawal freezes, exchange failure, a real and non-trivial tail); and the edges decay fast because everyone has the same open data. High Sharpe, low capacity, fast decay is the canonical crypto-MM signature. The honest sizing: this is a real business at small-to-medium scale, not a path to a giant. Capacity is limited and edges fade, so it rewards a fast research loop and continuous renewal over a single brilliant model. But "a real business a small team can actually run and be paid for" is exactly what the equities arena denies you, and crypto grants it. For the readiness map: Capital amber-to-green (you can start small), Data green, Infrastructure green, Edge amber (real but fast-decaying, so you must keep researching).

Prediction markets (Polymarket): open, but small

Prediction markets are the most open venue of all (public data, an order-driven book, no institutional gatekeeping, bounded payoffs) and the least crowded by sophisticated market makers. The microstructure maths applies cleanly. The catch is capacity: books are thin and events resolve, so the total money you can deploy is small. A genuine edge in a small pond.

Intuition first: Polymarket-style venues are where the canonical microstructure (queue position, adverse selection, inventory risk) holds but the institutions do not. There are fewer sophisticated competitors per book than crypto, let alone equities. For a quant who understands the maths and is willing to do event-specific work, that is a rare combination: real inefficiency, low competition, open access. The structural specifics: bounded payoffs (a share resolves to 0 or 1, which changes inventory and impact dynamics versus an unbounded asset); event-driven (the book lives and dies on an event's resolution, so timing and information about the event matter as much as microstructure); and thin books (small size moves the price, so tiny capacity per market). See prediction-market microstructure for the full treatment.

The honest ceiling: the total deployable capital across all live markets is small, so this is a venue for a sharp solo operator or tiny team, not a scalable firm. The edge can be excellent per unit of capital; the units are few. Treat it as a place to prove a loop and earn real (if bounded) money, and as a complement to crypto rather than a standalone business. For the readiness map: Capital green (you can start tiny), Data green, Infrastructure green, Edge green-per-unit but capacity-capped.

What does AI actually change in 2026?

AI changes velocity and leverage, not the existence of an edge. Code generation, research agents and infrastructure-as-code let one person build what used to take a team: the research and engineering legs of the pipeline collapse from weeks to days. But AI cannot manufacture a real edge, and it does not lower the equities fixed-cost wall. It makes a good operator far more capable; it does not make a sub-scale venue viable.

Research velocity. AI assistants propose, screen and prototype hypotheses far faster, so a solo quant can iterate through ideas at something like team pace. The bottleneck shifts from "can I implement this" to "is this edge real out of sample", which is exactly the performance discipline AI cannot fake. Code generation and infra-as-code. The engineering that used to need a dedicated systems hire (feed handlers, an order router, a backtest harness, monitoring) is now substantially AI-assisted and reproducible from config. A one-person shop can stand up a credible production stack in days. This is the single biggest practical change for the independent operator.

What AI does not change, stated plainly: it does not lower the equities colocation and data wall (that is hardware and licences, not code); it does not give you a faster physical link; it does not invent alpha (it accelerates the search, but the market still has to contain an inefficiency); and, because everyone has the same AI tooling, it speeds up the decay of any edge you find, since competitors discover and copy it just as fast. AI is a leverage multiplier on both sides of every trade. The net for an independent: AI turns "you need a team to build this" into "one disciplined person can build this", which genuinely tips the crypto and prediction-market verdict toward viable. It does nothing for equities and futures, where the constraint was never the code. The fuller treatment is what AI changes for HFT.

What you genuinely need vs what is now cheap

You genuinely need four things: capital, data, infrastructure and an edge. In 2026, data and infrastructure are cheap or free in the open venues (free crypto and Polymarket data, your own cloud, AI-built systems). Capital and a real edge are not, and a real, surviving edge is the one thing no tooling can hand you. The cheap things got cheap; the hard things stayed hard.

Capital: still required, but the floor dropped. In crypto and prediction markets you can start with a modest book and scale with the edge; in equities you need the multi-million fixed cost before trading once. The 2026 change is that the open-venue floor is now low enough for an individual. Capital still gates how much your edge can earn (the capacity ceiling), but it no longer gates entry in the open arenas. Data: now cheap or free in the open venues, still a wall in equities. Free public L2 over WebSocket in crypto and Polymarket versus proprietary feeds plus non-display licences in equities (transparent costs). This is the single biggest "got cheap" item and the reason the open venues are open. The paid layer this site will offer (clean, research-ready historical order-book datasets) exists precisely to remove the one remaining data friction in the open venues: good history for backtesting.

Infrastructure: now cheap, thanks to cloud plus AI. Ordinary cloud plus AI-assisted, infra-as-code systems replace what used to be a systems team. The building-a-trading-system and backtesting pages are the blueprint; in 2026 a solo operator can actually follow them. A backtest harness and reference implementations remove the rest. Edge: the one thing that stayed hard, and the only thing that matters. No amount of cheap data, free infrastructure or AI tooling substitutes for a real, surviving inefficiency. AI speeds the search and the decay in equal measure, so the edge must be genuinely yours and continuously renewed (the research loop). This is where the honest reader spends their effort, because it is the only pillar that cannot be bought or generated. If you take one thing from this page: the cheap things are cheap for everyone, so the edge is the whole game.

Worked example

The four-pillar readiness map, read out market by market, is illustrative, as of 2026, and the spine of the verdict. In equities and futures, capital is red (a multi-million-pound fixed cost before trading), data is red (proprietary feeds plus non-display licences), infrastructure is red (colocation, FPGA, microwave routes), and edge is red for speed and amber for a slow signal, so the verdict is closed to a newcomer. In crypto, capital is amber-to-green (start small, scale with edge), data is green (free public L2 over WebSocket), infrastructure is green (your own cloud, an AI-built stack), and edge is amber (real but fast-decaying), so the verdict is genuinely open, a real small-to-mid business. In prediction markets, capital is green (start tiny), data is green (public order-book data), infrastructure is green (your own cloud), and edge is green-per-unit but capacity-capped, so the verdict is open but small, a sharp edge in a small pond.

A concrete starting shape (illustrative, not advice): a crypto market-making loop on one or two liquid pairs, free venue data, a cloud-hosted, AI-assisted stack from the systems blueprint, a hard kill-switch, and a relentless research-to-production loop to replace edges as they decay, sized small, scaled only as the net Sharpe survives live. That is a real 2026 path. Equities, for the same person, is not. The clean way to prove the loop with bounded downside is a sharp position on a single prediction market, then graduate the same machinery onto crypto for capacity, and the one remaining friction, good backtest history, is what the datasets and tools waitlist is built to remove. Educational only, not investment advice; no figure here is a promise. The map is a framework for deciding where to spend your effort, not a guarantee of return.