I spoke with IEX and I believe I understand now why they don't use (and need) Thor. First because their technology is much better than it was at RBC. RBC being a broker dealer didn't had that much incentive in getting the fastest connections to the exchanges or the best hardware. They weren't in the "arms race" to zero latency, IEX is. Secondly, because their router is centrally located amongst data centers in New Jersey, they don't need to spread out orders in a Thor like fashion because they are quite close to the other data centers (and the exchanges housed within). At RBC, they were routing off lower Manhattan. Signals had to travel up the Hudson River, and across into NJ, before reaching each of the data centers sequentially: first Weekawken (BATS), then Secaucus (Direct Edge), then Cartaret (NASDAQ) and finally Mahwah (NYSE). The time it took to reach each data center was relatively consistent, and predictable. This enabled the HFTs to play their games. By having their router close to the other data centers in NJ, there is no need for Thor anymore. Any difference in reaching out different destinations will be so small, it won't matter.
I wrote this article from the information that I learned from talking to IEX: " Ever since Michael Lewis published his big best seller Flash Boys, an interesting exchange has gotten a lot of the public’s attention. IEX, or The Investors Exchange, was created by Brad Katsuyama and his colleagues from the Royal Bank of Canada(RBC) after they realized that the US stock market was severely lacking an exchange that would actually respect and be fair to the investors and traders who were bringing their business there. The main idea is that they will not allow special advantages to HFTs that are routing their orders to their exchange; the way they achieve that is through several methods: Limited amount of order types. Unlike the main exchanges that allow almost any kind of order type (even if unfair) to be created. No maker/taker rebates – a pricing system designed to give “kickbacks” to certain firms at the expense of other firms. No collocated HFTs. They deny HFTs the privilege of being collocated inside their exchange. A 350 microsecond latency for any order action taken by a participant (enter, cancel, revise an order), which ensures the IEX Matching Engine has sufficient time to receive and process changes to the NBBO (national best bid and offer) before a fast participant can receive and act upon the change on IEX. This last point is a very important one. If you check out Eric Hunsader’s twitter feed and website (@nanexllc and nanex.net) he talks about all the time how HFTs can cancel orders faster than order routers can spread out their orders to different exchanges. This lets the HFTs widen out their spreads and make a little more on the transaction than otherwise would be the case. How can a trader use IEX to its advantage and improve its order execution? You see, the way IEX works, you can route an order, say buy 10,000 shares of PG, to IEX (which currently is an ATS, a dark pool, but soon it might become a displayed exchange) and if there are orders to sell PG there, you will get executed for whatever amount is resting there for up to 10,000 shares. If there isn’t the full amount that you want, they will re-route the balance of your order to the 11 registered displayed stock exchanges (3 from NYSE, 3 from NASDAQ, 4 from BATS and the Chicago Stock Exchange) plus the LavaFlow ECN. The thing is, Katsuyama and his colleagues are the folks who created Thor, the system that is used to combat HFTs games by getting orders to hit all the exchanges/ECNs at the same exact time to prevent them from cancelling or changing their orders after they discover your interest. Because IEX is run by folks who know a lot about these HFT games and who were creative about how to work around it, you can expect that they will be intelligent about how IEX routes out your orders. This creates an interesting tool to traders: using IEX as a smart router. They have a proprietary method in which they re-route orders to other pools of liquidity (they have filed patent applications covering it). What they do disclose is the following: After leaving IEX, the order does not go through dark pools or internalizers first (which is the case with quite a number of retail brokerages’ smart routers). Frequently orders that hit dark pools tip off HFTs and allow them to cancel their orders in other places. Because routing to a dark pool is a shot in the dark, you don’t know how many or if there are any shares of your stock bid or offered there at all, your upside can be limited and the downside of alerting HFTs is significant. They have direct data feeds from the exchanges that they are trying to hit. This means that their order router is looking at the present rather than looking at the past which is the case with the SIP. They will hit those 11 exchanges plus LavaFlow over a custom-built dark fibre network (which is proprietary) that is designed as to not allow the HFTs there to cancel or modify orders in other places when they see the buy or sell interest. Their own 350 microsecond delay prevents HFTs from outracing IEX’s smart order router to the other displayed exchanges based on “signaling” from buy or sell interest on IEX. IEX’s new order routing approach doesn’t use (or need) Thor for several reasons: First because their technology is much better than it was at RBC. RBC, being a broker dealer, didn’t have much incentive to get the fastest connections to the exchanges or the best hardware. They weren’t in the “arms race” to zero latency, whereas IEX’s hardware and connections are as good or almost as good as the top HFT firms these days. Secondly, because their router is centrally located among data centers in New Jersey, they don’t need to spread out orders in a Thor-like fashion because they are quite close to the other data centers (and the exchanges housed within). At RBC, they were routing off lower Manhattan. Signals had to travel up the Hudson River, and across into New Jersey (NJ), before reaching each of the data centers sequentially: first Weekawken (BATS), then Secaucus (Direct Edge), then Cartaret (NASDAQ) and finally Mahwah (NYSE). The time it took to reach each data center was relatively consistent, and predictable. This enabled the HFTs to play their games. By having their router close to the other data centers in NJ, there is no need for Thor anymore. Any difference in reaching different destinations will be so tiny that it won’t matter. With this unique architecture, you can expect significant improvements over other routers. The results from their own statistics show that it is indeed the case. They have a 98% fill rate (data from June 2014). Where they route is tightly correlated to each exchange’s market share, and not to the exchange’s pricing schedules (like most routers). They also report a ~15% fill rate on their venue, where 70-80% of their trading is done at the midpoint. What this means is that if you send an order for 10,000 PG, it has a real opportunity to first execute at the midpoint on IEX (and getting executed at midpoint is a “price improvement” compared to the NBBO, it puts a bit of money in your pocket through better priced fills), and in the case where IEX does not have all the offsetting liquidity, the balance is sent out to the displayed exchanges. If there are 10,000 shares (or whatever balance that wasn’t filled at IEX) in those displayed exchanges (at the NBBO), you are likely to get 98% of it filled. This resolves the problems that a lot of traders have experienced: the market disappears on you when you are trying to work out a big order. Eric Hunsader from Nanex recently did an exposé on this issue. It should be a required reading to anyone trading US equities in any significant size. In it he showed a chart of the ideal results while routing a big order, the results from a regular router and the results from the IEX router. The different in performance in remarkable: The cost of using the IEX smart router is also not significant, just $0.0001 per share (about $1 for a 10,000 share order), plus the exchange fee/rebate incurred when orders are routed out to other exchanges (that you will pay or receive anyway). When the orders are matched at IEX, the cost is $0.0009. Its low when you consider that occasionally getting at least part of your order filled at midpoint will save you a significant amount ($0.005 for a 1 cent spread stock, the savings are more than five times the cost). Currently (as of August of 2014) there are three retail brokerage houses that allow order routing to IEX. They are Interactive Brokers, Tradestation Securities and Lightspeed. Having an account for bigger orders at those brokerages can decrease your costs dramatically. It’s also important to ask your brokerage (in case it’s not one of the three) to enable IEX routing: the more clients demand it, the more they will feel forced to do it. As a result, traders and investors won’t have to rely on the inefficient and frequently biased smart routers the some brokers offer. As it will be the case with a lot of advice in this site, try it out and see if your results improve." http://www.beathft.com/?p=13
I think that cost is what IEX charges your broker, but not what your broker charges you. For Tradestation: http://www.tradestation.com/en/pricing/stocks-and-etf-pricing They don't charge extra for per share pricing, but for flat-fee pricing: "Direct routing up to 1000 shares is no additional charge, $.004 per share thereafter." For Lightspeed: http://www.lightspeed.com/pricing/routing-fees/ IEX routing fee is $0.0014, which is a little higher than the amount you mentioned IB: At least for bundled pricing, I don't think they charge extra for directed orders Looks like I'll be opening a Lightspeed account.