Why options data costs more than stock data

The options line on a data plan is usually the expensive one, and the natural guess is that the options exchanges charge more. On the per-person fees they mostly do not. The cost sits in what a vendor has to receive, store, rebuild and license before a single chain reaches your screen, and most of it is written in the processor's own documents.

This page is about that options-specific part. The fee vocabulary common to stocks and options (display, non-display, professional, redistribution) is in why real-time stock data is so expensive, and the processor itself has its own entry at OPRA.

How it works

Every US listed options exchange sends its quotes and trades to one processor. OPRA, the Options Price Reporting Authority, is a national market system plan whose participants are the exchanges that list options; the participant list on its own site names six exchange groups: BOX, Cboe, MEMX, MIAX, Nasdaq and NYSE. The Securities Industry Automation Corporation runs the system. In the SEC's words in its order of 25 February 2026, SIAC "gathers the last sale and quote information from each of the OPRA Plan members, consolidates that information, and disseminates the consolidated OPRA Data."

What the consolidated stream contains is narrower than people assume. Per the binary data recipient specification, version 6.4b of 25 August 2026, it carries last sales, quotes, open interest, an end-of-day summary, index values for index options, and administrative and control messages. Each quote is one exchange's best bid and offer in one series. On top of those, OPRA "calculates and identifies the 'Best Bid and Best Offer'" across all exchanges, the options equivalent of the NBBO.

The other half of the arrangement is the vendor. OPRA's overview defines a vendor as anyone who retransmits its data outside its own organisation, and says that "any external retransmission", down to a snapshot quote shown to a customer before an order, is enough to make you one. Every vendor signs a vendor agreement directly with OPRA, and that agreement is where the costs this page describes are set.

What a chain is built from

A chain is not a message OPRA sends. It is a table the vendor reconstructs, one row per series, from a stream that only tells you what changed.

The national best is published as a change, not as a state. The specification's appendix on the best bid and offer says OPRA recalculates it on every qualifying quote and attaches the new best to that quote as an "appendage" only when it moved: "If it is determined that the new quote does not affect the Best Bid or the Best Offer, appendages are not added." To print a bid and an ask on every row, a vendor has to hold the current state of every series in memory all day and apply each update in order. Miss a packet and that row is wrong until it is recovered from a retransmission or the next change overwrites it.

The ranking rule is specific. Price first; at the same price, the larger size; at the same price and size, the earlier quote. Some quote types never count: non-firm quotes, indicative values, rotation quotes and quotes during a halt are all listed as ineligible. The appendix also notes that the best bid and offer can be locked or crossed. A vendor that drops crossed rows, or shows one exchange's quote where the national best belongs, prints a different chain from the same feed.

There are a lot of rows. OPRA publishes a nightly series mapping file for the next session. The file generated on 2 October 2026 lists 1,925,926 series on 6,291 root symbols, and that leaves out FLEX series and anything listed during the day. One underlier is one row on the stock tape. In that file, Apple options are 3,326 series across 23 expiries and 126 strikes, and SPY options 11,984 series across 30 expiries. A "full chain" for one liquid ETF is twelve thousand rows, each updating independently. That is why cheap endpoints trim to near-dated, near-the-money strikes; the options chain entry lists the trims.

The last trade on a row can be hours old. A series that has not traded since the morning shows a last sale from the morning beside a quote from now. Open interest is on a third clock, a cleared figure from the previous session. See open interest.

Why the full feed is a bandwidth problem

Stock data scales with the number of companies. Options data scales with companies times expiries times strikes times two, and every one of those series is quoted by several exchanges at once.

The capacity projections OPRA sent to multicast subscribers on 15 September 2025 are the figures a firm is told to build for. For July 2026 they project up to 13.575 million messages and 4.403 gigabits in the busiest 100 milliseconds, 0.501 gigabits in the busiest 10 milliseconds, and 311 billion messages in a day, rising to 343 billion by July 2027. Expressed as a rate, those bursts are roughly 44 and 50 gigabits a second. That is our arithmetic; the notice states the figures per interval. It also says the projections are "for one stream only". Firms that take both redundant streams double them, and the notice adds 10% for retransmissions.

The stream is split across 96 multicast lines on a fixed symbol table today. From 23 November 2026, OPRA moves to dynamic symbol rebalancing: 97 lines, each root symbol mapped to between one and sixteen of them, and the mapping can change every night. "Data recipients are responsible for handling any symbol over any line." So you cannot cheaply take only the options you care about: the lines carrying them change, and each line carries other symbols' traffic too.

Direct access is physical as well. The fee schedule defines it as a port on the NMS Network, which "exists only within the Mahwah Data Center" in New Jersey, on a 10 or 40 gigabit connection. The connectivity is billed by the NYSE exchanges and reproduced in OPRA's schedule. On the IP Network a 10 Gb connection is 10,000 dollars to install and 12,000 dollars a month, and a 40 Gb one 20,000 a month; on the Liquidity Center Network it is 24,000 a month at either speed. OPRA's own direct access fee of 1,000 dollars a month comes on top. The SEC order of 25 February 2026 records that OPRA data used to travel over a network with "many access points outside of the Mahwah Data Center". Now everyone without a port buys from a vendor that has one.

This is the real cost behind a live options product. A vendor holding every series's state through bursts of that size, with a redundant copy, is running infrastructure an equity-only vendor never needs. Some vendors filter quotes to cope; ThetaData advertises the unfiltered NBBO as its point of difference. How much storage a day of it takes is the subject of storing tick data.

What it costs

From the OPRA fee schedule dated 26 February 2026, set against the equity consolidated tape's schedule of market data charges (last restated 6 February 2019, and still the version the CTA plan links):

Per subscriber, then, options cost about what stocks cost. The professional device fee has been 31.50 dollars since 1 January 2018, which is the last increase the schedule records. What differs is the professional-to-non-professional gap, about 25 times, and who OPRA counts as non-professional. Its overview requires someone "not employed in the securities industry or the commodity futures industry or by a bank or an insurance company or an affiliate of either" in a securities or futures role, and allows use for nobody but the subscriber and immediate family. That is a test of employment, where the equity policies in professional subscriber start from registration. An unregistered employee doing securities work at a bank or an insurer fails it.

One line has no equity equivalent. A vendor may pay per "options chain" instead of per quote packet: 3 cents for last sale and bid and ask on "up to all series of put and call options on the same underlying security". On the series counts above, an SPY chain as individual quote packets would be 11,984 packets, about 90 dollars. As one chain it is 3 cents. A non-professional's usage fees are capped at 1.25 dollars a month either way. This is why chain endpoints are metered per call rather than per contract.

These figures are set by amending the plan, not by a vendor. OPRA files a change under Rule 608 of Regulation NMS and the SEC publishes it for comment. The most recent order, on 25 February 2026, approved a new definition of direct access and the connectivity fees. It changed no amounts: "OPRA is not modifying the amounts of the Connectivity or Direct Access Fees."

Delayed, historical, and the line between them

Most retail options tools are fifteen minutes delayed or end-of-day, and the fee schedule's footnotes are the reason. OPRA data "becomes 'delayed' after at least 15 minutes have passed since the information was first transmitted". It "becomes 'historical' upon the opening of trading on the next succeeding trading day". There are three classes, priced differently:

• Current data carries everything: per-user or per-chain fees, and the redistribution fee.
• Delayed data carries no per-user fee. The overview says "OPRA does not charge usage-based or device-based fees for the use of delayed data", and delayed requests are not counted for usage fees. But it still needs a vendor agreement, and the redistribution fee is payable by any vendor that redistributes "whether on a current or delayed basis". So a free delayed chain still has a fixed monthly cost behind it: 1,500 dollars, or 650 for a query-only service.
• Historical data is the only class outside the redistribution fee, which "does not apply to a Vendor whose redistribution of OPRA Data is limited solely to 'historical' OPRA Data".

One consequence: an end-of-day file published the same evening is still delayed data under these definitions. It becomes historical at the next morning's open. That is the cheapest class on the schedule, and it is why options history is sold by the file and the day, the shape of Cboe DataShop.

What OPRA does not carry, and the feeds that do

Depth. OPRA carries each exchange's best quote in each series. It does not carry the orders resting behind it. Depth comes from the exchange's own feed, on its own licence. On Cboe's price list, version dated 30 July 2026, Cboe Options Top is 15.50 dollars a month per professional user and 0.30 per non-professional. Cboe Options Depth is 50.00 per professional user. Distributor fees for Top are 9,000 dollars a month internal and 5,000 external.

Complex orders. Quotes for spreads, meaning multi-leg orders resting in an exchange's complex order book, have no message category in OPRA's specification. OPRA reports a complex order only when it trades, as leg prints with their own trade-type codes. The quote lives on the exchange's complex feed. Cboe Options Complex is 25.00 dollars per professional user and 1.00 per non-professional, with 3,000 a month for internal distribution.

The whole of any one exchange. A single venue's top-of-book feed is cheaper per user than OPRA, but it shows only that venue. Cboe's own aggregate of its four options exchanges, Cboe One Options, is 30.50 dollars per professional user and 0.60 per non-professional, in the range of OPRA's own rates, and it still covers only four of the exchanges OPRA consolidates. OPRA's vendor agreement goes the other way: according to the overview, a vendor that includes one participant's information for a series "must also include comparable information in respect of that same series from every other Participant Exchange's market".

The stock price. OPRA carries index values for index options, and nothing equivalent for the stock under an equity option. Every implied volatility and greek on an equity option needs that price, so it is bought separately, from the equity tape or an exchange feed, on a separate licence.

Implied volatility and greeks. None of the specification's fields is a calculated volatility or a sensitivity. The one delta in it is a negotiated term on a FLEX trade type, not a calculation. Everything sold as IV or greeks is the vendor's model over OPRA quotes and a stock price from somewhere else, and the reasons two vendors disagree are in implied volatility, greeks and why indicator values differ. The options-specific part is the timestamp: an option quote and a stock price from two feeds, two licences and two clocks have to be paired before the model runs, and nothing in either feed says how the vendor paired them.

So the options line on a data plan pays for roughly four things. The OPRA fees, which are close to the stock tape's. The infrastructure to rebuild two million series from a 40-plus-gigabit stream. A second feed for the stock price. And the modelling on top. Of the four, only the first is on a public schedule.

What you can do about it

1. Decide which of the three classes you actually need. Screening, research, backtesting and anything published to other people work on historical data, which carries no per-user and no redistribution fee. If you need the chain during the session but not the second, delayed data removes the per-user charge. Pay for current data only when a decision depends on the live quote.

2. If you publish anything, budget the redistribution fee before you build. A delayed chain on a public page or a bot is still redistribution at 1,500 dollars a month. Files of yesterday's data are not. If you are prototyping something public, build it on historical data and switch later.

3. Check your status against OPRA's test, not the equity one. OPRA's non-professional definition turns on employment in the securities, futures, banking or insurance industries. If you work in any of them, price the 31.50-dollar professional device fee, or a vendor's usage-based equivalent, before the 1.25-dollar one.

4. Ask a vendor three questions before comparing prices. Is the quote on each row the national best or one exchange's? Are quotes filtered, and by what rule? Where does the stock price for the greeks come from, and is it timestamped against the option quote? Then compare. ThetaData and Databento sell the raw OPRA side and leave the modelling to you. ORATS sells the modelling, a smoothed surface over years. Those are different purchases even when the pages look alike.

5. Buy a single exchange's feed only for what OPRA lacks. Depth of book and complex order quotes are the reasons to pay a venue directly. A cheap single-venue top-of-book feed is not a substitute for the national best.

6. Do not take the full feed yourself unless latency is your product. A Mahwah port, two redundant streams and storage for hundreds of billions of messages a day cost more than any subscription in options data and flow analytics. The wider list is the options data collection, and how to get an options chain walks the cheapest routes to a chain that is good enough.