ICE delivered its best-ever third quarter, with adjusted EPS up 10% to $1.71 on record net revenues of $2.4 billion, underpinned by 5% recurring-revenue growth across all three segments. The Exchanges segment saw open interest surge 16% year-over-year into late October, FIDS reached record revenues of $618 million, and the index business hit a record $754 billion in ETF AUM. ICE returned $674 million to shareholders and cut leverage to about 2.9x, while announcing a strategic investment in Polymarket and rolling out its ICE Aurora AI platform. Softer spots were lower CDS member-interest income on falling rates and several one-time revenue items management said would not recur.
Good morning. ICE's third quarter 2025 earnings release and presentation can be found in the investor section of ice.com. These items will be archived and our call will be available for replay. Today's call may contain forward-looking statements.
These statements, which we undertake no obligation to update, represent our current judgment and are subject to risks, assumptions, and uncertainties. For a description of the risks that could cause our results to differ materially from those described in forward-looking statements, please refer to our 2024 Form 10-K, 2025 third quarter Form 10-Q, and other filings with the SEC. In our earnings supplement, we refer to certain Non-GAAP measures. We believe our Non-GAAP measures are more reflective of our cash operations and core business performance. You'll find a reconciliation to the equivalent GAAP terms in the earnings materials. When used on this call, net revenue refers to revenue net of transaction-based expenses and adjusted earnings refers to adjusted diluted earnings per share. Throughout this presentation, unless otherwise indicated, references to revenue growth are on a constant currency basis.
Please see the explanatory notes on the second page of the earnings supplement for additional details regarding the definition of certain items. With us on the call today are Jeff Sprecher, Chair and CEO; Warren Gardiner, Chief Financial Officer; Ben Jackson, President; Lynn Martin, President of the NYSE; and Chris Edmonds, President of Fixed Income and Data Services. I'll now turn the call over to Warren.
Thanks, Katia. Good morning, everyone, and thank you for joining us today. I'll begin on Slide 4 with some of the key highlights from our record third quarter results. Third quarter adjusted earnings per share were $1.71, up 10% year-over-year and the best third quarter in our company's history. Net revenues totaled $2.4 billion and were underpinned by a 5% increase in recurring revenue. This recurring revenue growth was fueled by a 9% rise in exchange data and a 7% uplift in fixed income and data services, both reflecting sustained demand for our high value proprietary data offerings. Third quarter adjusted operating expenses totaled $981 million. Our disciplined cost management was further supported by approximately $15 million in one-time benefits, about evenly distributed across compensation, expense, and depreciation and amortization. After adjusting for these benefits, we would have been towards the low end of our guidance range.
I also want to provide some color on our third quarter adjusted tax rate of 21%, which benefited from recent prior year tax audit settlements. Excluding this benefit, the adjusted tax rate would have been within the prior 24%-26% guidance range, and as a result, we expect the fourth quarter tax rate will normalize to between 24% and 26%. Moving to capital allocation, we returned $674 million to our shareholders during the quarter, including approximately $400 million of share repurchases. In addition, we reduced debt outstanding by roughly $175 million, reducing gross leverage to just over 2.9 times EBITDA. Next, I will touch on a few fourth quarter guidance items. We expect fourth quarter adjusted operating expenses to be in the range of $1.005 billion-$1.015 billion. The sequential increase is largely driven by the aforementioned one-time expenses items not repeating in the fourth quarter.
Fourth quarter adjusted non-operating expense is expected to be between $180 million and $185 million, driven by a sequential uptick in interest expense related to our October investment in Polymarket. As a note, we funded $1 billion of that investment with CP issuance in early October and expect to fund up to an additional $1 billion in the future, also utilizing existing capacity on our commercial paper program. Now let's move to Slide 5 where I'll provide an overview of the performance of our Exchange segment. Third quarter net revenues totaled $1.3 billion, building on strong double-digit growth in the prior two years. Transaction revenue totaled $876 million.
Importantly, towards the end of October, open interest across our futures and options complex surged 16% year-over-year, with energy futures up 14% and interest rate futures climbing 37%, underscoring the growing demand for our risk management tools amid shifting macroeconomic conditions. Shifting to recurring revenues, which include our exchange data services and our NYSE listings business, revenues totaled a record $389 million, up 7% year-over-year. Underpinning growth in our record recurring revenues was 9% growth in our broader exchange data and connectivity services, which is once again led by our futures data while also benefiting from approximately $6 million of audit-related revenue that we don't anticipate will repeat in the fourth quarter. In our listings business, the NYSE helped to raise a market-leading $20 billion in new IPO proceeds through the first three quarters of 2025.
It is worth noting that only roughly half of new IPOs have met the NYSE's listing standards, and these high standards remain a critical component of our 99% retention rate. As a result of this strong performance within our exchange data business, we now expect full year growth to be towards the high end of our 4%-5% guidance range. Turning now to slide 6, I'll discuss our fixed income and data services segment. Third quarter revenues totaled a record $618 million, including transaction revenues of $123 million. On a year-over-year basis, ICE Bonds revenues increased 15%, driven by 41% growth in our muni business, which was in part driven by growing institutional adoption. Within our CDS business, results were largely driven by lower member interest, a direct result of the lower fed funds rate.
When compared to the year-ago period, recurring revenues totaled a record $495 million and grew by 7% year-over-year. In our fixed income data and analytics business, record third quarter revenues of $311 million increased 5% year-over-year, driven by growth in pricing and reference data and our index business, which reached a record $754 billion in ETF AUM as of the end of the third quarter. Data and network technology revenues were a record and increased by 10% in the quarter, an acceleration from 7% growth in the first half and 5% growth in 2024, driven by heightened demand for our ICE Global Network. Our strategic investments in data center infrastructure are paying off, driven by increasing demand for data and increased capacity as well as clients preparing to integrate AI into trading workflows.
We also continue to drive high single-digit growth across our Consolidated Feeds business and our desktop solutions as we continue to realize the benefits of investments to enhance our platform. Worth noting, the third quarter included a few million dollars of one-time revenue that we don't expect will repeat. That said, we still anticipate fourth quarter revenue growth in Data and Network Technology to be in the high single-digit range and for total segment recurring revenue to be between 5% and 6% for both the fourth quarter and the full year. Please look to slide 7 where I'll discuss our Mortgage Technology results. Third quarter revenues totaled $528 million, up 4% year-over-year. Recurring revenues total $391 million and increased on a year-over-year basis. The year-over-year improvement was largely driven by our Data and Analytics business and MSP within our Servicing business.
Shifting to the fourth quarter, we expect revenues to remain at these levels, primarily driven by Mr. Cooper's acquisition of Flagstar and customers resetting their minimums on Encompass, which I note is paired with the benefit of higher transaction fees. We expect these items to largely be offset by revenue from new customers coming online. Transaction revenues totaled $137 million, up 12% year-over-year, driven by double-digit revenue growth related to Encompass closed loans and high single-digit growth from MERS registrations. As you look to the fourth quarter, it's important to remember typical seasonal impact on purchase volumes, which tend to be lighter in the fourth quarter relative to the second and third quarters. In summary, the third quarter once again grew revenues, adjusted operating income, and adjusted earnings per share, building upon our record first half results and representing the best year-to-date performance in our company's history.
As we continue to strategically invest in our future, we've also returned over $1.7 billion to shareholders year to date. As we look to the end of the year and into 2026, we remain focused on extending our track record of growth and on creating value for our shareholders. I'll be happy to take your questions during Q&A, but for now I'll hand it over to Ben.
Thank you, Warren, and thank you all for joining us this morning. Please turn to slide 8. Technology and innovation have been foundational to Intercontinental Exchange since our inception. Our approach to AI is a natural extension of that legacy. We are using it to accelerate our existing 25-year automation journey by building and implementing tools to drive efficiency and deliver enhanced analytical insights for Intercontinental Exchange and our customers. We are now taking the next step by combining our pursuit of workflow automation across our business processes with the solutions we provide to our clients through generative and agentic AI under the name of ICE Aurora. As we continue to expand our AI capabilities, we're leveraging three core strengths: deep operational and complex workflow expertise, highly differentiated proprietary data which we believe will only grow in value, and the powerful network effects of our platform.
We started with a deep understanding of our data, workflows, task and document management, as well as the rules and compliance frameworks of our businesses. We then conducted a risk assessment of how much automation can be applied to executing these workflows based on the impact, technical maturity, accuracy, and model explainability in the AI tools available, balanced against the risks of automation, similar to benchmarks used across industries. To measure the scale of automation, we rank our automation within processes on a scale of 0-5. At 0, the process is entirely manual. At 5, the process is fully automated, including exception handling without requiring human input.
We are applying this model to every workflow across Intercontinental Exchange bottom up, measuring exactly where we are today in terms of the maturity of AI models, automating workflows with or without human intervention, and where we can get to based on the current state of the technology. Currently, most generative or agentic AI models at their core are best at pattern recognition, and this recognition continues to evolve. This means there is a stochastic and probabilistic accuracy to them, measuring the reliability and predictability of the outcomes AI models produce for the highly regulated businesses that we and our customers operate. There has to be an acknowledgment of how much accuracy a probabilistic outcome must have in order to be considered acceptable for full automation versus when some level of human interaction remains necessary, especially in exception handling.
Today we have clear visibility of where we can go and are executing on this in many areas, balanced by the risk I just outlined. That is our strategy and what our ICE Aurora platform is all about, and we're already seeing results across ICE. AI is streamlining and automating workflows across systems, accelerating product development and dramatically accelerating the speed with which we can deliver the modernization of multiple tech stacks within ICE. Importantly, we aim to do this without compromising our adherence to information security, data management, and privacy. In our energy markets, the macro AI and data center expansion trend is expected to drive significant energy demand over the next decade. We believe our trading and clearing platform, which offers deep liquidity and price transparency across the full energy spectrum, is uniquely positioned to support customers despite lower overall market volatility.
The third quarter of this year was the second strongest third quarter in our history following the record quarter of a year ago, led by continued strength in our global gas and power markets, with third quarter volumes up 8% and 18% year-over-year, respectively. As we've consistently said, open interest is a leading indicator of future growth, and we're pleased to see it continue trending higher with record Futures Energy OI in October up 14% year-over-year, including 25% and 30% growth in our Brent and TTF benchmarks, respectively. This reflects the value of our diversified energy platform, the depth of our liquidity, and the confidence customers place in our benchmarks, which serve as global price reference points across thousands of related contracts, providing trusted price transparency across geographies.
Across our global gas portfolio, which spans North America, Europe, and Asia, volumes have increased 20% year to date. Importantly, the strong year to date performance has been underpinned by broad-based strength, including a 16% increase in our North American complex, a 26% increase in our European portfolio, and a 27% increase in our Asian JKM market. In parallel, our power markets have seen continued growth with volumes up 21% year to date and 18% in the quarter. This reinforces the synergy between our gas and power markets and the need for comprehensive risk management tools that offer transparency, flexibility, and choice in fixed income and data services. Driven by multi-year investments, our comprehensive platform delivered another quarter of record revenues, which grew 5% year-over-year, including 7% growth in recurring revenue and 10% growth in our data and network technology business.
Our proprietary data is the cornerstone of our business and a key differentiator in the evolving AI landscape. With over 50 years of experience, our high-quality pricing and reference data serves as the foundation for what is today one of the largest providers of fixed income indices globally. From benchmark indices and analytics to custom solutions, we support the full ETF ecosystem. As AI becomes embedded in trading strategies across all areas of investing, we expect our proprietary data to grow in strategic importance, with our data sets providing a competitive edge to users of AI models that depend on precision, depth, and large quantities of historical data. Our data is securely managed within ICE's infrastructure, protected by firewalls and entitlements. Our commercial agreements tightly control access and only permit specific use cases through authorized delivery channels.
This approach helps ensure our data remains exclusive and strategically deployed, especially as models increasingly rely on high-quality inputs to drive performance. In our reference data business, we're leveraging AI to process and validate documents from hundreds of sources using AI models that we thoroughly test for fit for purpose and high probabilistic outcomes from Google, Meta, Amazon, and several other AI models, achieving over 95% accuracy in extracting reference data from fixed income prospectus. This capability is a critical part of the collection process, improving both efficiency and speed of delivery, enabling us to do more with the same resources. Today, within our reference data business alone, we are processing roughly 40,000 documents on average per month using AI.
Documents assessed by AI that meet predefined confidence thresholds go straight into our database for clients to consume, while those falling below the threshold are flagged for manual review and intervention. This capability is a critical part of the collection process, improving both efficiency and speed of delivery, enabling us to do more with the same resources. We're also leveraging machine learning to power key components of our evaluated pricing. Our continuous evaluated pricing blends trade and quote data to predict bond pricing, complementing our deep market expertise and data quality workflows. Additional models use historical data to determine bid-ask spreads across the bond universe, with machine learning capabilities significantly improving evaluation quality when measured against actual trades in the market.
Meanwhile, our ICE Global Network continues to set the standard for resiliency, latency, and security, connecting participants to over 750 data sources and more than 150 trading venues, including ICE and the NYSE. The ICE Cloud comprises state-of-the-art data centers owned and operated by ICE and facilitates seamless integration with key third-party cloud providers, all under ICE's cybersecurity and Operational Resilience framework. This provides our clients flexibility to access AI workloads where it makes the most sense without compromising cyber and operational controls. We continue to invest in our data centers to support business growth needs and to meet growing customer demand, including to support increased adoption of AI strategies. This is to ensure we are accessing the most cost-effective, secure, and reliable infrastructure for ICE's needs and our customers' needs both now and in the future.
Cross Product Development AI is automating data analysis, pattern recognition, and repetitive processes using tools such as GitHub Copilot, freeing product managers to focus on validation and enhancement. This has already accelerated speed to market for certain products. For example, we've reduced the time to convert code for index qualification calculation and reporting by roughly 60%, demonstrating the new innovation underway across ICE. We're utilizing AI with our new sentiment indicator data sets, including Reddit, Dow Jones, and soon Polymarket, with Google and Meta AI models helping to process these data sets and identify patterns. While still in the development phase, these data sets are particularly attractive to market participants seeking an edge through differentiated data inputs. This illustrates how our proprietary data set is set to become increasingly vital to a trading community reliant on models to support trading decisions.
In our mortgage business, the use of AI is helping our efforts to streamline the homeownership experience, enhancing productivity of lending and servicing operations, improving the borrower experience with self-service workflows, reducing risk via automated compliance and quality checks across the mortgage lifecycle, all while improving recapture rates for our customers. All of this contributes to lowering the cost to originate and service the loan for our customers, a foundational part of our mortgage strategy. For example, customers using our industry standard loan servicing system MSP save roughly 20%-30% on the cost to service alone based on a recently conducted customer study, and we expect this number will increase with new innovations that we have come to market or are coming to market, such as our enhanced customer service loan boarding, ICE Business Intelligence for servicing, and our Loss Mitigation suite.
Thank you, Ben. Please turn to Slide 9. Given ICE's recently announced investment and business relationship with Polymarket, I thought it might be helpful to explain our thinking on the evolution of markets. ICE was an early investor in the crypto space, having been an early stage funder of Bakkt and Coinbase. We made these investments in order to stay close to the evolution of the market's use of blockchain. In the case of Bakkt, we thought that there could be an acceptance of a system of tokens that adhered to a high level of then existing securities and commodities regulation. We found, however, that traditional regulated financial firms were slow or unwilling to adopt tokens during a period of regulatory uncertainty, particularly where events of default would move unwanted tokens onto a financial guarantor's balance sheet. Current U.S.
Administration and Congress have been attempting to address these uncertainties, which has caused ICE to more actively lean into the knowledge that we've accumulated over the past decade. One of the significant macro trends of the past decade of blockchain investment is a rewiring of the rails of the banking system. ICE, for example, operates six clearinghouses around the world, all of which are highly regulated and which are required to operate within the limitations of local banking hours, customs, and preferences. On chain banking now operates globally with 24x7 availability, allowing for instantaneous margin calls and trade liquidations. This facilitates increasing margining and lending against assets, which some cohorts of asset holders are clearly taking advantage of with increased risk management tolerances and which places excess trade financing collateral into an omnibus stablecoin collateral pool.
This excess collateral pool is funded by traders via the forfeiture of earnings on their collateral, features that were previously unavailable to regulated clearinghouses. ICE decided to invest in Polymarket as we're impressed with the design of its underlying architecture of smart contracts that take advantage of this new banking infrastructure. Alongside our investment, we've also announced a Strategic Data Agreement under which ICE will become a global distributor of Polymarket's highly differentiated event driven data. As the leader in non-sports prediction markets, Polymarket provides real-time probabilities on events like elections, economic indicators, and cultural trends, offering a powerful new layer of insight supporting more informed decision making. We believe that we can accelerate Polymarket's acceptance into the traditional financial system by virtue of our distribution, understanding, and longtime customer relationships.
We believe Polymarket's engineering team can help ICE's engineers better understand our own adoption of evolving banking technology, a relationship that is already paying dividends to both of us. ICE is in the process of rolling out an advanced clearing model for our global clearinghouses, one that we've very elegantly named ICE Risk Model 2. Our new clearing system was built on the existing local banking and regulatory infrastructure for funds movement and collateral management. However, the current regulatory environment is being confronted by collateral management using tokens, which I believe will help evolve regulatory oversight to take advantage of 24 by 7 capital movement, and ICE intends to be at the forefront of driving this evolution.
Given our own use case of operating six global clearinghouses with differing collateral and regulatory environments, such an evolution can make global clearing and trade settlement more efficient, and we've seen that the efficient use of collateral typically results in increased trading volumes and transaction revenues. One does not have to look too far to see that trading volumes in the U.S. equities markets have dramatically increased since the industry freed up collateral by moving from T+2 day to T+1 day settlement times. Beyond the rewiring of funds movement, Polymarket has pioneered the rapid listing of new markets driven by real-time consumer demand. Traditional exchanges have been subject to government approvals of our new product launches, which at best take 30 days and in many countries substantially longer. Polymarket is forcing a dialogue in the U.S. on how to minimize government regulatory burdens so as to not impede innovators.
We think this dialogue will ultimately benefit new product innovation for all markets and certainly for ICE. Now, augmenting on Ben's comments on the adoption of artificial intelligence, we see the jagged intelligence phenomena at play for both our own AI adoption and for that of our customers. Internally at ICE, we have our engineers using Copilot to help them write code more effectively, particularly where the projects involve modernizing our legacy code. However, to fully deploy production code at scale and at the latency precision which ICE operates, we still require unique skill sets that are not now available in AI. Our current experience is that AI has become a good assistant for our teams, but not a replacement. Ben also highlighted our use of AI in improving our customer service.
Artificial intelligence has made our help desk more efficient at diagnosing real-time issues as well as cataloging and summarizing customer inputs to create more efficient feedback loops. The third area where we've deployed AI is in our data gathering and data organization, such as cataloging bond and equity prospectuses, cleansing our data sets, and organizing unstructured data for our vast financial data offerings. Much of the regulation that ICE is required to oversee is surveillance in the form of pattern recognition. Here again, AI tools are making our colleagues more efficient at our oversight. In summary, our internal use cases for AI have made our colleagues better at what they do. In terms of our customer adoption of AI, we see that same jagged intelligence where AI is very helpful in some areas yet unreliable in others.
Where our customers interface with ICE products for pattern recognition or language organization, we're seeing positive uptake. For example, we've seen healthy uptake of our structured and unstructured financial data offerings. Similarly, the AI tools that we've built into our mortgage network, such as our Data and Document Automation and our Customer Engagement Suite, have strong interest with customers adopting these tools to more efficiently target new business and minimize the cost of mortgage onboarding, but not to replace underwriting decisions that are subject to regulatory oversight or to replicate the vast ICE mortgage network that links the industry together, including the U.S. Federal housing regulators' supervisory efforts, in validating GSE and Federal Home Loan Bank mortgage holdings and providing it with monthly mortgage service information. Finally, a number of people have speculated to me that the overall volumes of trading must have increased due to AI adoption.
While that's possible, I believe that a significantly larger volume impact has come from capital being freed up when moving equity settlement times one day forward, and with the expansion of retail trading leverage that's inherent in popular one day options. All in all, we think the current state of AI is helping to control costs and control new hiring and is for us at the margin, driving sales and transaction growth. Our record third quarter results, on top of our extraordinary third quarter results of last year, are another example of strong execution across our All Weather platform. We very intentionally position the company to provide customer solutions in numerous geographies and economic conditions to facilitate these All Weather results. I'd like to end our prepared remarks by thanking our customers for their continued business and thank you for your trust.
I'd also like to thank my colleagues at Intercontinental Exchange for their contribution to the very best third quarter in our company's history, following on our unsurpassed first half results and yielding the best year to date performance in the company's history. I'll now turn the call back to our moderator Lydia and we'll conduct a question-and-answer session until 9:30 A.M. Eastern Time.