AI Stock Trading — Where Does the Data Actually Come From? From Zero Yuan to 39,000 a Year, I Tested Every Route
Quite a few people are using AI to trade stocks and do investment analysis, and the more technical ones are doing quant work. But for this kind of job, how smart the model is comes second — the information source is the thing your life depends on. Whatever price AI is holding, it won't tell you what time it's from, and it won't raise an error either. I tested a free MCP on Moutai and got the previous day's close. On the same Hong Kong interface, two letters in the symbol were the difference between a live price and one 22 minutes old.
This episode works through every route for feeding trading data to AI. How the three-layer architecture (channel / method / query) divides the work, why the three free sources can cross-verify each other, what each rung of the price ladder from zero yuan to 39,000 a year actually buys you, and three rules that make data trustworthy. The strategy is calculated by humans and AI only speeds up execution — this episode is technical only and is not investment advice. The complete comparison table is linked at the end for download.
Two premises first
Strategy is something humans work out. What AI does is speed up execution — helping you find data faster, watch the tape, run the process. It is not going to conjure up a strategy that makes money for you. So this episode is not investment advice. No tickers, no timing calls. Technical only: how AI actually gets data, and what it takes for that data to be trustworthy.
Information sources split into two halves, trading data and sentiment. Sentiment runs deeper and I won’t get into it here. Trading data only.
Three-layer architecture: pick the wrong layer and it’s all wasted
There are several completely different routes for feeding stock market data to AI. A lot of people assume getting data just means finding an interface. It really isn’t like that — there are three layers here.
The channel layer: a command-line interface or a direct API, giving you tick-level to second-level quotes, and taking up almost none of AI’s attention. The public quote interfaces from Tencent, Sina, and East Money all live at this layer.
The method layer: all the various Skills packages — technical analysis, Chan Theory, the Greeks. A skills package doesn’t generate data itself. How fresh it is depends entirely on which channel it’s wired to. Wire it to a second-level channel and it can compute second-level things; wire it to next-day data and what it computes is next-day.
The query layer: the various MCPs. Financials, shareholders, and macro data are its strong suit, but in testing most free MCPs give you 15-minute-delayed or even next-day data. What it’s good at is breadth, not real time.
The free route: one command, and it can even cross-verify itself
Second-level A-share data, zero cost, one command to connect, the barrier is absurdly low. In testing, three sources hit the same stock (Moutai) simultaneously, and all three prices came back not a cent apart — free stuff that can cross-verify itself, which was the biggest surprise of this whole investigation.
The Hong Kong trap is the one worth remembering. Same domain: add two letters in front of the symbol and you get the real-time price (tested intraday against the clock on the wall, same second). Leave those two letters off and you get a price from 22 minutes ago, and it raises no error. Ask a free MCP about Moutai and it gives you the previous day’s close, no error raised, no sense that anything is wrong — this “no error but stale data” pattern is more dangerous than an outright error.
There’s also a hidden cost. Tushare’s official MCP has over 200 tools and Longbridge’s has over 140, all of them resident in AI’s context. The more you install, the easier it is for AI to pick the wrong tool, and the faster tokens burn — the “tool tax” I covered last episode.
The price ladder: what each rung buys
- Zero yuan: second-level A-shares plus cross-verification across three free sources, enough for watching prices intraday
- 19 to 29 dollars a month: buys history and breadth (for backtesting), still on 15-minute-delayed data
- 99 dollars a month: the minimum for true real-time across all US exchanges
- 199 dollars a month: tick-level data plus 20 years of history plus options
- Tens of thousands of yuan a year: a terminal like Wind (39,800 a year for a single seat, per media reporting from a few years back) — what you’re buying isn’t just data, it’s licensing, stability, and a compliant basis you can cite externally
Brokers are another route. Longbridge shipped an official command-line tool, explicitly for AI use — tick-level, with order book and broker queue, able to place orders directly, and exchange-licensed. Futu takes the same approach through a local gateway. The institutional route (Anthropic’s financial edition wired into FactSet, S&P Global, LSEG, Morningstar) doesn’t publish pricing. Just know it exists.
Three rules: having data doesn’t mean you can trust it
One, every number must come back carrying its own timestamp. AI is not allowed to just say “latest price” — you want the hour, minute, and second.
Two, hit important numbers from two or three sources at once. If they agree, proceed. If they don’t, stop.
Three, plenty of interfaces don’t raise an error when you exceed your quota — they hand you a null or a stale value, which is more dangerous than an error. Use the timestamp to verify freshness yourself.
On May 6, 2010, US stocks fell nearly 1,000 points in nine minutes, and there’s a detail in the regulatory report: NYSE quotes reaching the national market system were delayed by as much as five minutes, yet the time stamped on those quotes read “current.” (The report’s conclusion was that this delay wasn’t the primary cause of that day’s crash, but it confirmed the phenomenon really did occur.) If even an official system can have timestamps that don’t mean what they say, how can you not verify data you wire up yourself.
FinanceBench (a 2023 benchmark) found that with the same model, feeding it an entire raw financial report versus having it retrieve on its own produced results that differed by a factor of four. The model didn’t change. The only thing that changed was how it got its information.
This methodology isn’t just about stocks. Exchange rates, oil prices, competitor pricing, inventory — for any scenario where you feed real-time data to AI, the thinking is the same. Map out the layers of your data sources, pick the route that fits the use case, then set the rules for how the data gets used.
The commercial red line
Free public interfaces have no exchange licensing and no service commitment, and one redesign at the source site breaks them — only suitable for your own research and prototypes. To put something in a product or publish externally, go through a broker or a commercial data source, and settle the license first.
Source: EP0070 digital-avatar voiceover (pinpinvoice0717b voice) · text = final voiceover script (ground truth) · timecodes aligned sentence-by-sentence by gemini-2.5-pro (chunked parallel ASR, 98.2% coverage)
[00:00] AI stock trading — where does the data actually come from? I spent a few days recently going through this question from the ground up. Every route on the market for getting data, I tested one by one, and I verified every price on the official site. In the end I built a web comparison table. Price, real-time performance, accuracy — three dimensions laid out side by side, posted on my blog, and you’re welcome to go grab it. This video is me laying out what came out of that research and sharing it with you. Before we officially start, there are two things I have to be clear about.
[00:27] First: the strategy is worked out by a human. What AI does is make execution more efficient. It is not going to conjure up a money-making strategy for you. You need to internalize this. Second: nothing in this episode is investment advice. I don’t pitch tickers and I don’t call entry or exit timing. We’re only talking about technology — only about how to get AI to handle the data side more reliably. Also, information sources actually split into two parts, trading data and sentiment. Sentiment runs much deeper, and I’m not opening that up today. We’re only talking about trading data. Now, feeding stock market data to AI — there are several completely different routes out there,
[00:54] and the differences are huge. A lot of people think getting data just means finding an interface. It really doesn’t work that way. I’ve broken it into a three-layer architecture — think of it as three different levels. The bottom layer I call the channel layer. That’s the command-line interfaces, or direct APIs. What you get is tick-level to second-level quotes, and it takes up almost none of AI’s head space. The public quote interfaces from Tencent, Sina, and East Money all fall into this category. Their advantage is speed, high real-time performance. One layer up is the method layer — all the various Skills packs.
[01:21] Technical analysis, Chan theory, the option Greeks — these skill packs don’t produce data themselves. How fresh their data is depends entirely on which channel is wired in behind them. Hook one up to a second-level channel and it can compute second-level things. Hook it to next-day data and what it computes is next-day. Above that is the query layer — all the MCPs. Financial statements, shareholders, macro data: that’s where MCP is strong. But free MCPs, from my testing, mostly give you fifteen-minute-delayed or even next-day data.
[01:48] What it’s good at is breadth, not being real-time. So you see, it’s all “getting data,” but across those three layers the latency, the depth, the licensing, and the price differ enormously. If you don’t build that understanding first, it’s very easy to fall into traps later. Let’s start with the most surprising route: the free one. Second-level A-share data, zero cost, one line of command to connect — the barrier is absurdly low. I tested three sources: the public quote interfaces from Tencent, Sina, and East Money. Fire off one line of command and the data comes back with a second-level timestamp,
[02:16] plus five levels of order book. I hit all three at once on the same ticker, Moutai, and the prices from all three didn’t differ by a single cent. Something free that can also cross-verify itself — that was the biggest surprise of this whole research run. Honestly, before I tested it I had no idea free interfaces could get this far. Now, the Hong Kong stock trap is more interesting. Same domain: add two extra letters in front of the code and you get the real-time price. I tested during trading hours and it matched the clock on the wall to the second. Leave those two letters off and what it gives you is the price from twenty-two minutes ago,
[02:43] and it throws no error. If you don’t check the timestamp, you have no idea the data is stale. I fell into this one, so I have to tell you about it. What about the free MCP? I asked it about Moutai and what it gave me was the previous day’s close — next-day data. It doesn’t error out either, and it doesn’t think anything’s wrong. That situation, no error but stale data, is more dangerous than an outright error, because you have no clue anything went wrong. And there’s another problem: tool specs eating up head space. Tushare’s official MCP, for instance, has over two hundred tools;
[03:10] Longbridge’s has over a hundred and forty. All of them sit resident in AI’s context in full. The more you install, the easier it is for AI to pick the wrong tool, and the faster you burn tokens. I covered this specifically in my last video — I call it the tool tax. So more tools isn’t better. You connect what you need. Okay, free is covered. Now let’s look at the whole price ladder, and how you should actually choose for different situations. Watching A-share prices intraday: use the three free sources to cross-verify. Not a cent spent, second-level data. If you need history and breadth, say for backtesting, then the price is
[03:37] 19 to 29 US dollars a month, but the data is still fifteen minutes delayed. For genuinely real-time US equity quotes, the minimum entry is 99 US dollars a month — that’s the lowest tier with true real-time across all exchanges. Above that, 199 US dollars a month gets you tick-level data plus twenty years of history, and options data too. And above that you’re into terminals costing tens of thousands of yuan a year — the Wind kind, 39,800 yuan a year for one seat, a number the media reported a few years back. What you’re buying there isn’t just data. It’s licensing, stability, and a compliant basis you can cite publicly. Now, the broker route:
[04:05] Longbridge shipped an official command-line tool and said outright it’s for AI. Tick-level, with order book and broker queue, and it can place orders directly, with exchange licensing. Futu goes through a local gateway, same idea. Deep order book for Hong Kong and US stocks costs money — everyone should know that. As for the institutional route, Anthropic has its own finance edition, launched this May, wired into FactSet, S&P Global, LSEG, Morningstar and the like. Institutional pricing isn’t public, and that one isn’t for individuals. Just know the route exists. So, for different situations,
[04:31] the route you pick is completely different. Here are a few typical decision references. Watching A-share prices intraday: three free sources cross-checked, zero cost. Genuinely real-time US quotes: the 99-dollar tier. Tick-level and options: the 199-dollar tier. Order-book depth on Hong Kong and US stocks plus placing orders: go through a broker. Bulk research on A-share financials, shareholders, and macro: Tushare’s points system, which the community puts at roughly 200 to 500 yuan. Now someone might ask, why is there such a big spread in price? The core behind it
[04:57] is licensing, the depth and breadth of the data, and the service commitment. Free interfaces come with no license and no commitment, and they break the moment the source site gets redesigned. With a commercial data source, what your money buys is stability and compliance. Now, getting the data doesn’t mean you can trust it. You have to lay down rules. Coming out of this research I set three ironclad rules for myself. Rule one: every number has to come back with its own timestamp. AI isn’t allowed to just say “latest price” — I want to see that timestamp so I can check freshness myself. Rule two: for important numbers, hit two or three sources at once,
[05:25] and only move on if they agree. The three-free-source cross-check is this rule in practice. Rule three: plenty of interfaces don’t throw an error when you blow past your quota — they hand you a null value, or a stale one. That’s more dangerous than an error. So you absolutely have to verify with timestamps yourself. Which reminds me of a case. On May 6, 2010, US equities dropped nearly a thousand points in nine minutes, and there’s a detail in the regulatory report that came later: NYSE quotes reaching the national market system were up to five minutes behind, yet the time stamped on the quote said current. The report concluded that this delay wasn’t the
[05:51] main cause of the crash that day, but it confirmed the phenomenon really did happen. Look — even official systems have timestamps that don’t match reality. How can we wire up data ourselves and not verify it? There’s also a study called FinanceBench, a 2023 benchmark. It found that with the same model, feeding it the full original financial statement versus letting it retrieve on its own produced a fourfold gap in scores. The model didn’t change. The only thing that changed was how it got the information. Which is exactly the point: how you feed the data is itself part of the capability. If the data you feed is inaccurate or out of date,
[06:18] no matter how smart AI is, it can’t give you reliable results. Last point: the commercial red line. Free public interfaces have no exchange license, no service commitment, and they break the moment the source site gets redesigned. They’re only suitable for your own research and prototyping. If you’re going into a product, shipping it publicly, you have to go through a broker or a commercial data source and negotiate the license first. Don’t step on this mine. Get this whole thing straight and AI can genuinely do work for you. And this methodology carries over to all real-time data. Exchange rates, oil prices, competitor pricing, inventory —
[06:46] any situation that needs real-time information, the thinking is the same. First get clear on the layers your data source sits in, then pick your route by situation, and finally lay down your rules for using the data. The full comparison table, with price, real-time performance, and accuracy side by side, is on my blog — just Google Zhang Pinpin and you’ll find it. That’s it for this episode. See you next time, bye.