Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
For four decades, the gap between institutional and retail investors came down to three things: better data, faster execution, and the analytical horsepower to turn the first two into decisions. Every other supposed edge (relationships, deal flow, regulatory access) was a derivative of those three. The whole premium pricing structure of the financial data industry, the entire compensation logic of buy-side research, even the architecture of how exchanges route orders, all of it rested on the assumption that a small number of firms could afford the access and the rest could not.
That assumption is breaking in real time, and the speed at which it is breaking is the part most investors have not yet priced in.
The Edge Was Three Layers, And Two Have Already Fallen
A Bloomberg Terminal in 2026 runs $31,980 per seat per year, with a two-year minimum commitment. FactSet sits between $12,000 and $50,000 per user. Refinitiv, S&P Capital IQ, and the various specialist data vendors stack on top of those. A mid-size hedge fund desk of ten analysts is burning through roughly half a million dollars a year on data subscriptions before anyone has placed a single trade.
What that spend bought, historically, was the first two layers of the moat. Real-time tape, options flow, regulatory filings the second they hit, alternative data feeds, Bloomberg IB chat as the de facto bond market negotiation venue. The third layer, intelligence, was a labor problem: hiring the analysts who could turn the data into a thesis. That layer is also dissolving, faster than the data one, because language models have collapsed the cost of reading, summarizing, and cross-referencing financial documents to roughly zero.
The first layer (the data itself) is the one most readers underestimate. The price points above have not yet collapsed, but the value those price points capture is being undercut from three directions: regulatory disclosure mandates that increasingly require public posting of what used to be private intelligence, alternative data providers selling directly to retail aggregators, and most importantly, the brokerages themselves opening up their APIs to AI agents owned by individual customers.
Retail Volume Is Now Setting Prices, Not Following Them
The reflexive response to any “retail is empowered now” argument is that retail volume is too small to matter. That argument is roughly five years out of date.
Retail order flow crossed 20% of US equity trading volume in 2025, per Jefferies, and the spikes are even more telling than the average. J.P. Morgan reports that retail activity reached 36% of total order flow on April 29, 2025, an all-time high during the tariff-driven sell-off. Vanda Research reported that retail investors purchased a net $155.3B worth of single stocks and ETFs in the first half of 2025, and J.P. Morgan estimated that retail investors purchased a net $270B in U.S. equities during H1 2025, with potential for an additional $360B over the second half. ARC Group + 2
For context, that puts retail net buying in 2025 in roughly the same order of magnitude as the combined active equity inflows of the entire long-only mutual fund industry. The institutional consensus that retail is a contrarian indicator (the dumb-money sleeve to fade) has had to be quietly retired. Sosnick at Interactive Brokers put it bluntly: institutions are now forced to play along with narratives retail has set, not the other way around.
Two structural points fall out of this. First, the marginal price-setter in a growing share of names is no longer institutional. Second, when retail concentrates around AI-mediated decisioning (which is the direction every retail brokerage is now sprinting toward), the behavioral patterns institutions have spent two decades learning to fade will stop being reliable.
The Brokerages Are Racing To Hand Over The Keys
The clearest signal that the data moat is being dismantled is that the brokerages themselves are now selling the picks and shovels.
On March 31, 2026, Public launched what it calls Agentic Brokerage, letting users build AI agents that monitor markets and execute trades from natural-language instructions. Investors on Public can now create AI Agents that monitor markets, move money and automate trades. Three weeks later, on April 23, Moomoo launched API Skills, explicitly designed to let retail investors plug their own AI agents into the platform’s execution rails. By eliminating the need for coding, Moomoo API Skills empowers users’ AI agents to serve as 24/7 trading assistants. Interactive Brokers, Alpaca, Tradier, Tiger Brokers, and Robinhood all have variants of the same offering either live or in roadmap. PR NewswireStock Titan
This is not a niche developer feature. Connect Trade, the broker-aggregation API, now offers an MCP server that lets any LLM connect to over twenty compliance-approved brokers through a single integration. In Korea, where the regulator opened brokerage APIs to retail in 2025, the volume in self-built AI bot accounts has already crossed roughly 100 trillion won (around 70 billion USD). Grand View Research now sizes the AI trading platform market at about USD 11.23 billion in 2024 and is projected to reach USD 33.45 billion by 2030, growing at a 20 percent CAGR. Appinventiv
The point is not that retail will suddenly start outperforming. Most will not, for the same reasons most active managers do not. The point is that the structural information asymmetry that justified institutional pricing is gone. A retail user with a $20 LLM subscription, a free Alpaca account, and access to the same disclosure feeds an analyst at a mid-tier hedge fund relies on can now run a fundamentally similar workflow at roughly 0.1% of the cost.
Who Loses When The Data Aggregators Get Commoditized
This is where the investment thesis sharpens. The losers in the unwind are not, in the first instance, the prime brokers or the bulge-bracket banks. Their revenue depends on volume and complexity, both of which the retail AI wave actually increases. The losers are the layer in the middle: the data aggregators and the legacy terminal businesses that priced themselves on the assumption their customer base had no realistic alternatives.
Bloomberg LP is private and will be slower to feel margin pressure than its public peers, but the public ones tell the story. FactSet’s adjusted operating margin dropped to 36.8% in Q3 2025 on a 21% increase in technology expenses, with management explicitly framing AI investment as defensive. Refinitiv (inside LSEG) and S&P Market Intelligence face the same compression curve. The end state is not zero, but it is a business that looks more like a regulated utility (selling normalized data feeds at thin margins) and less like the 35-40% operating margin software businesses Wall Street has been comfortable paying high multiples for. This is the same kind of structural margin reset I outlined when fintech multiples started compressing earlier in the cycle.
[Also see: Our Guide to Robo Advisors]
The Trades Retail Still Cannot Make (And Probably Never Will)
It is worth being honest about what is not being democratized. Block execution at scale, access to IPO allocations, prime brokerage financing terms, OTC derivatives, structured product origination, and the parts of the bond market that still clear through Bloomberg IB chat are all still institutional preserves and will remain so. The AI-and-API wave commoditizes the analytical and small-to-mid-ticket execution layers. It does not touch the parts of finance where the moat is regulatory, balance-sheet-based, or relationship-based rather than informational.
That distinction matters for positioning. Long the prime brokers and the exchanges (CME, ICE, Nasdaq) that get paid on volume regardless of who is doing the trading. Cautious on the pure data and analytics middle layer. The terminal incumbents will survive in fixed income and FX where the IB chat network is genuinely irreplaceable, but their equity-side pricing power is on a slow erosion path.
Position For The Margin Compression, Not The Headlines
The temptation with a structural story like this one is to look for the obvious trade: short the data vendors, long the AI brokerages. That is too clean. The data vendors have multi-year contracted revenue and high switching costs that will mask the underlying compression for several reporting cycles. The retail AI brokerages are operating in a regulatory grey zone that will tighten before it loosens, particularly in Europe and probably in the US under whichever administration owns the next market accident.
The more durable trade is to mark down the multiples you are willing to pay for any business whose pricing power rests on being the gatekeeper to information that is now leaking through every consumer-grade LLM. That includes parts of the wealth management industry, mid-tier sell-side research, and any platform whose value proposition reduces to “we have the data and you do not.” The information moat has been the load-bearing assumption underneath a remarkable amount of financial sector market cap. When it dries up, the repricing is not going to be priced into a single quarter’s earnings call. It is going to bleed out over years, and the first leg of the move is already visible in the cost base, even if it has not yet hit the revenue line.
