The Zhitong Finance App learned that after the US stock market closed on Wednesday, IBM (IBM.US) announced financial results for the third quarter of 2026. According to the data, the company's revenue for the quarter was US$17.103 billion, up 5% year on year, up 6% year on year; GAAP net profit from continuing operations was US$2.99 billion, up 26% year on year, and diluted earnings per share was US$2.31, up 26% year on year. Non-GAAP adjusted net profit from continuing operations was $2,812 million, up 12% year over year, and adjusted earnings per share were $2.95, up 11% year over year. IBM shares rose more than 2% after the market.
In terms of profitability, IBM's GAAP gross margin for the third quarter was 56.8%, down 0.5 percentage points year on year; non-GAAP adjusted gross margin was 58.5%, down 0.2 percentage points year on year. The adjusted profit before tax was US$3.295 billion, up about 9% year on year. The adjusted profit margin before tax was 19.3%, up 0.7 percentage points year on year.
By business, the software division's revenue increased 8% year over year to 7.804 billion US dollars. Among them, Red Hat's hybrid cloud business grew 11%, data business grew 19%, transaction processing business increased 9%, but automation business revenue fell 3%. The revenue from the consulting business was US$5.335 billion, which is basically the same as the same period last year; revenue from the infrastructure business increased 4% year over year to US$3.707 billion, of which the distributed infrastructure business increased 31%, while the IBM Z mainframe business decreased by 20%.
In terms of cash flow, IBM's cash flow from operating activities in the third quarter was US$4,025 million, an increase of about US$900 million over the previous year; free cash flow was US$3,027 billion, an increase of about US$700 million over the previous year. In the first nine months of this year, the company's cumulative free cash flow reached US$7.787 billion. By the end of the third quarter, IBM held $11.1 billion in cash, restricted cash and marketable securities, with total debt of $63.8 billion.
Looking ahead to the full year of 2026, IBM maintains the forecast of 4% to 5% revenue growth at a fixed exchange rate, and the impact of the exchange rate on the annual revenue growth rate is basically neutral. At the same time, the company raised its free cash flow forecast, and it is expected that the full year's free cash flow will increase by more than 1 billion US dollars compared to 2025.
Below is the full IBM results call (AI tool assisted translation):
I. Executive speeches
Chairman and CEO Avind Krishna
Third-quarter revenue growth accelerated to 6%, above the benchmark scenario set in July. This performance reflects an overall acceleration in various business segments: software growth of 9%, infrastructure growth of 5%, and moderate improvements in consulting. Disciplined enforcement also led to strong earnings and free cash flow performance.
Software sector: Red Hat maintained continuous growth of 11%, and the order growth rate returned to high double digits; OpenShift's annual recurring revenue (ARR) growth rate accelerated to more than 30%. Transaction processing (TP) increased by 10%, confirming that the weakness in the second quarter was a matter of pace rather than a demand issue. The data business grew 20%, and Confluent continues to be strong. Infrastructure grew by 5%, with distributed infrastructure growing by 32%.
AI and storage: AI continues to drive storage demand, and the company is expanding its share through differentiated solutions in Flash, Fusion, tape, and IBM Z storage. Consulting revenue grew 2%, and clients modernized to deploy AI safely and at scale.
z17 console: Continuing the strongest innovation cycle in the platform's history, bringing in over $1 billion in incremental revenue over 6 quarters, benefiting from security, reliability, and total cost of ownership advantages (2 to 15 times superior to non-platform solutions). Launched in July, the single-rack/rackmount Z system recorded the strongest mid-range model initial revenue in nearly 20 years this quarter.
Lightwell: This is a multi-billion dollar opportunity for open source software security. Currently, the pipeline is close to $700 million, and more than 40,000 fix packages have been released to the web.
AI Strategy: Enterprise AI is entering a new phase, and “deploying more models or agents” is not enough. The key is whether AI can be run safely, efficiently, and sustainably within the enterprise. IBM integrates capabilities across four platforms: Confluent provides a real-time data foundation; Watsonx (including Bob and Orchestrate) supports AI-driven development and agent orchestration; HashiCorp and Apptio automate identity, access, and policies; and Red Hat provides an open hybrid architecture. Data is at the core of the AI operating model, and 72% of enterprises report that real-time data infrastructure is inadequate. To this end, we launched Confluent's Streamhouse capability to continuously capture, transform, and query real-time data to unify operations and analysis data.
Cost and governance: Businesses are most concerned about “how much it costs and who controls it”. Apptio already connects AI spending with business results; llm-d, developed in collaboration with Red Hat and Google, can reduce the cost of running large models by 5 to 10 times; Bob now supports localized deployment. At the governance level, HashiCorp Vault can register, authorize, audit, and revoke agents. Agent Identity is cooperating with Nvidia's Open Agent Safety Platform, and we believe that “proxy identity” will become the foundation of enterprise AI like “Internet identity”.
Quantum computing: In July, we collaborated with the University of Chicago, Algorithmiq, and Qedma to achieve three independent quantum advantage demonstrations. The Nighthawk r2 processor reaches 7,500 gates and performs 100,000 circuits per second, improving performance by up to 19 times compared to the previous generation. IBM was selected for the third phase of the US Defense Advanced Research Projects Agency (DARPA) quantum benchmarking program; this year, the quantum business has signed 16 cross-industry partnerships, including AT&T and GM. With the successful connection of the first modular cryogenic system, the continued advancement of the 300,000 square foot Poughkeepsie manufacturing facility, and a $1 billion grant from the Department of Commerce for wafer manufacturing, the company is confident of scaling up.
Full-year outlook: We are reaffirming revenue growth of 4% to 5% in 2026 and raising our free cash flow growth forecast for the full year to over $1 billion.
CFO James Kavanaugh: Financial Data and Outlook
Third-quarter revenue increased 6%, profit margin before sales tax expanded by 70 basis points, and diluted operating earnings per share increased 11%. Free cash flow for the first nine months was approximately $8 billion, up 8% year over year.
Software: Revenue accelerated by 4 points to 9% month-on-month (including more than 4 points of organic growth), driven by high-value annual recurring revenue (ARR), and the base grew 8% to $24.9 billion. Three of the four software sub-sectors achieved double-digit growth and successfully closed about two-thirds of the falling orders in the second quarter. Red Hat grew 11%; OpenShift is already a $2.4 billion ARR business, growing more than 30%; virtualization has signed nearly $800 million since early 2024, and the pipeline is about $750 million. Data increased by 20% (Confluent orders increased by about 20%); TP increased by 10%; HashiCorp increased orders in various regions, and automation was dragged down by a high base.
Infrastructure: Growth accelerated to 5%, distributed infrastructure grew 32%, and storage recorded record growth and continued to expand its share. z17 performed well, making this quarter its strongest mid-tier launch in nearly 20 years. Hosting customers with around 85% of their installed MIPS capacity are maintaining or expanding their capacity, and their cumulative installed capacity has tripled from previous cycles.
Consulting: Revenue increased 2%, strategy and technology grew 4%, and cybersecurity grew by nearly double digits. Generative AI accounted for about 30% of the quarter's consulting revenue and 35% of backlog orders. We are expanding the customer capability ecosystem through a new strategic partnership with OpenAI and deepening our relationship with Google Cloud.
Profit and cash flow: Adjusted EBITDA and profit margins before sales tax expanded by 60 and 70 basis points, respectively; profit margins in the software and infrastructure segment expanded by 210 and 120 basis points, respectively. Free cash flow for the first nine months was approximately $8 billion, up about $600 million year over year, driven by an adjusted EBITDA increase of approximately $1 billion, partially offset by capital expenses, cash taxes, and net interest expenses. Cash of $11.1 billion at the end of the period, $4.8 billion in dividends for the first nine months, and $63.8 billion in debt (including $12.7 billion in financing operations; nearly 80% of the receivables portfolio is investment-grade).
Full year guidance: Maintain revenue growth of 4% to 5% and increase free cash flow to over $1 billion. Software grew by about 7% to 8% throughout the year (benchmark of about 7%). Among them, hybrid cloud double digits, double digits in data, low single digits of automation, and flat TP; low single digits of infrastructure; and low to medium single digits of consulting. The profit margin before sales tax expanded by 100 basis points throughout the year, and the sales tax rate remained in double digits. The free cash flow realization rate was approximately 130% for many years.
II. Q&A session
Evercore ISI analyst Amit Darianani
Question focus: How are customer spending priorities changing? Are infrastructure investments starting to drive incremental demand for software? Does a high double digit Red Hat order indicate a software acceleration in 2027?
Arvind Krishna: Customer investment in AI is still strong, but for AI to generate value, applications must be rewritten or modernized, which directly benefits Red Hat OpenShift, and we're still at the beginning of this cycle. 90% of enterprise data is within the enterprise rather than in the cloud, unlocking data benefits Confluent. Automation will only benefit when applications are scaled up and enterprises need to manage hundreds or thousands of applications. On Red Hat's side, we are confident about double-digit growth in the fourth quarter, but we won't say anything about 2027. Generally speaking, high order growth will translate into subsequent revenue growth.
James Kavanaugh: The overall performance of the software is satisfactory. The 4 points of sequential acceleration are all organically driven. High-value recurring revenue accounted for about 80%, a high single-digit increase, while transactional business (the remaining 20%) rebounded to double digits. Among them, enterprise license agreements (ELAs) grew by more than 20% and net revenue retention rate exceeded 130%. About a quarter of the software acceleration came from falling orders, and three quarters came from net new business. The double-digit increase in TP confirms the strategic demand for host software, and the previous “demand destruction vs. rhythm problem” debate can be put to an end. The z17 has been on the market for 6 quarters. Installed MIPS increased by double digits, and shipment MIPS increased by more than 100%. This is an opportunity for TP to accelerate in the future.
Bank of America analyst Wamsey Mohan
Question focus: What is the percentage of TP in the big drop order? Is 4% to 5% of the year more on the upper end? Can TP crossing cycle performance still be considered to be in the middle single digits?
Arvind Krishna: People were worried about the decline in demand for hosting last quarter; we always thought it was just a matter of pace. As customers deploy hardware, MIPS, and transaction volume grow, demand for host software will inevitably return. The lower to mid-single-digit decline was previously given to TP due to prudence; the actual delay for the second quarter combined with the normal procurement cycle for the third quarter was realized at the same time. I don't expect TP double digits to continue, but it will return to a more normalized level and enter the fourth quarter or decline somewhat. Behind this are the host's underlying requirements for resiliency, economy, and “8 9" availability architectures.
James Kavanaugh: Software's quarterly revenue was about $7.6 billion, and there was a gap of about $400 million in the second quarter. We closed about two-thirds of these falling orders (over $250 million), accounting for about 25% of the total acceleration, and mainly on TP. We are relieved that we pragmatically anchored software growth of about 7% in the fourth quarter, and annual revenue of about $32 billion was in line with market expectations.
Ben Reitzers, Melius Research Analyst
Question focus: Is there still confidence in the software's long-term “near double digit” growth target? What role do mergers and acquisitions play in this?
James Kavanaugh: We are confident in our strategy and combination. Early signs in the software sector include: the acceleration of high-value recurring revenue, high-single-digit growth throughout the year, which is expected to approach double digits in the fourth quarter; in terms of new innovations, Red Hat accelerated to high double digits, the Lightwell pipeline of $700 million (facing the trillion-dollar open source vulnerability market), virtualization and hybrid cloud containerization opportunities; 2026 was the year of consolidation of Confluent, and 2027 was a year of growth synergy; and on the TP side, the installed MIPS doubled in 5 years and increased 10 times in 10 years, a clear monetization vehicle. We'll give quantitative guidance in January.
Barclays analyst Remo Renshaw
Question focus: How is Confluent's upsell/cross-selling progressing? What are the drivers of gross margin?
Arvind Krishna: Confluent is responsible for moving data in real time so that it remains in place and can still be used by AI; in conjunction with this, Lakehouse (Lakehouse), we are optimistic about the cross-selling opportunities brought by the upcoming Streamhouse. Unlocking data and avoiding additional costs for AI to copy multiple copies of data is a core capability of enterprise AI.
James Kavanaugh: Over the past 4 years, the company's operating leverage increased by about 1,000 basis points, cash growth of about $7 billion, and free cash flow margin increased 700 basis points. Total productivity of 500 basis points was achieved during the quarter, of which about one-third was used to protect profit margins and two-thirds was used for reinvestment. We have unlocked more than $5.5 billion in cumulative productivity, providing a powerful flywheel for investment flexibility and competitiveness.
Jefferies analyst Brent Thiel
Focus of the question: The early promises of some AI projects are not in line with reality. Are companies re-evaluating and returning?
Arvind Krishna: Large enterprises are increasingly scrutinizing the true return on AI spending and are concerned about uncontrolled investment. The productivity benefits of AI are real, but many people underestimate the transformation work required to integrate AI into the workflow; it works without installing an RPA tool. The company began to stop small-scale testing, select a small number of scenarios for large-scale implementation, select trustworthy software partners, and control token costs at the same time. This is fine for OpenShift, Confluent, and Hosting.
Citi analyst Fatima Blaney
Question focus: How do volume and price each contribute to the recovery of transaction processing (TP)? Has pricing power been used?
James Kavanaugh: TP is mission-critical software running on the most durable platform (host). The host carries about 70% of the world's transaction value, 45 of the top 50 banks, and 7 of the top 10 retailers. We operate about 145 million MIPS globally. The six quarters of z17 listing MIPS was 125% higher than the same period of z16. Shipment MIPS increased by more than 100%, and installed MIPS increased by 18%. The former is consumption and capacity that can be immediately monetized, while the latter is future market space. z16 remained flat during the launch year, and then accelerated year by year. z17 currently also makes a prudent assumption that it remains flat, but 100% shipment MIPS and 18% installed MIPS mean considerable monetization opportunities in 2027.
J.P. Morgan analyst Brian Essex
Question focus: How do customers pay attention to “sovereign AI” when evaluating AI transformation, and how can data and intellectual property be protected?
Arvind Krishna: Sovereign AI is becoming an increasingly important topic, with few people paying attention to it in 2024, a few people starting to pay attention to it in 2025, and clearly heating up in 2026. Using large public models will leave traces of use in the underlying logs, and companies want to ensure that data and logs are not obtained by the outside world. This has spawned a dual boom of local AI infrastructure and an open source weighting model, as the latter gives enterprises rights and intellectual property rights to run locally or in private clouds. A major bank CEO made it clear that cutting-edge models can be tested externally, but internal operations must be sovereign. This will benefit infrastructure, semiconductors, and players who help businesses deploy AI in a secure and private environment.
Matt Swanson, Capital Markets Analyst, Royal Bank of Canada
Question focus: How do companies balance “AI costs” and “control”, and how can they design network architectures to balance the two?
Arvind Krishna: At the end of 2024, the gap between the most advanced cutting edge model and the open source model was about 2:1 to 3:1. The gap was too big for no one to accept; by the end of 2025, the gap has narrowed, and will narrow further according to the trend. The real challenge for enterprises is the lack of kernel engineers to optimize GPU clusters, network proximity, and result caching. We estimate that large enterprises can run the open source weight model with 2 to 3 racks. The token cost is only 10% to 25% of the cutting-edge model, and the effect has reached 90%. This is an account any business would calculate: 90% effect, 10% cost. Most tasks such as summarizing emails, reviewing contracts, and responding to customer service can be completed with cheaper solutions, which is the key to driving changes in the industry.