IBM’s mainframe business has reported what it describes as its best quarter in two decades. The company, which has been repositioning itself for years as a “software-led” firm focused on hybrid cloud and Artificial Intelligence, surprised in its latest earnings report by putting traditional hardware back in the spotlight: its Z division achieved the highest annual revenue related to a mainframe generation in the last 20 years, driven by the launch of the z17 system.
The figure was revealed during the Q4 2025 earnings call and the full-year report, where CEO Arvind Krishna stated that the z17 had experienced “the strongest start” of any mainframe generation in the three quarters since its launch. The message was clear: in a market dominated by discussions of public cloud, hyperscale data centers, and generative models, IBM is asserting that the mainframe remains a competitive—and, for some clients, even the most efficient—piece of infrastructure.
A Upward Quarter: Infrastructure Drives Growth
Financially, IBM closed the quarter with $19.7 billion in revenue, up 9% year-over-year and higher than the previous quarter, which recorded $16.3 billion. Within that total, the infrastructure segment contributed $5.1 billion, an increase of 21%, and the IBM Z business grew 67% year-over-year, according to data shared during the call.
Although IBM reaffirms its strategy as a software and platform company, the quarter’s results clearly indicate that the mainframe renewal cycle is not only ongoing but also serving as a revenue engine at a time when many organizations are reevaluating where their critical workloads should reside.
Why the Interest Returns: Digital Sovereignty, Control, and “Unit Economics”
A significant part of IBM’s argument revolves around digital sovereignty and “on-premise” control. Krishna linked the strong performance of the z17 to a renewed client appreciation of the mainframe’s economic value for specific workloads: the thesis is that, for certain processes, the mainframe can offer the “lowest cost per unit,” a concept that practically points to operational efficiency, workload consolidation, and predictability.
This narrative resonates with a trend gaining traction across regulated sectors: banking, insurance, government, and healthcare, where control requirements, data residency, and auditability weigh as heavily as raw performance. Additionally, the geopolitical and regulatory context has elevated “sovereignty” from a slogan to a contractual requirement in some projects.
AI on the Mainframe: The z17 Facilitates Deployments
The other major theme is Artificial Intelligence. IBM explained that the latest mainframe iteration simplifies AI deployments for clients, highlighting the incorporation of the Spyre accelerator, available on the z17 since October 2025. The company aims to position the mainframe not just as a legacy transactional core but as a platform capable of inference and automation in controlled environments, close to the data, with predictable latencies.
In this context, Krishna projected an ambitious scenario: in the next 3 to 5 years, 50% of enterprise AI usage would be in private cloud or on-premises data centers. Moreover, he suggested that if smaller models are used “on-premise,” then between 80% and 90% of all inference could be performed on private cloud or on-premise infrastructure. If realized, this forecast could reshape the balance of power between public cloud and private deployments, reinforcing IBM’s focus on hardware and software tailored for sensitive workloads.
“GenAI Book of Business” and the Weight of Software
Nevertheless, IBM does not want to be pigeonholed as just a hardware and rack company. The leadership emphasized their strategic positioning as a hybrid cloud and AI platform provider, noting that Software was the largest revenue contributor this quarter, with $9.0 billion.
Additionally, the company highlighted that their “GenAI Book of Business” exceeded $12.5 billion in the quarter. CFO Jim Kavanaugh pointed out that this figure had been in the “low hundreds of millions” in 2023, when IBM first began reporting it. From a corporate narrative perspective, the message is twofold: the mainframe is gaining momentum again, but the company’s long-term, structural growth target continues to focus on software and services related to AI and hybrid cloud.
Rising Memory Demand and a Domino Effect in Servers
Another notable point from the call was a reference to the server market. Krishna described the environment as “volatile,” stating that DRAM prices are about six times higher than a year ago, attributing this to much of the industrial capacity shifting towards HBM (High Bandwidth Memory), a key component for AI servers.
The consequence, according to Krishna, is that AI demand is not only straining the supply chain for accelerators but also increasing the demand for “regular” servers, with CPUs powering these platforms. While acknowledging this could be a “small problem,” he does not expect it to significantly hinder IBM’s hybrid cloud strategy.
Full-Year 2025: $67.53 Billion and Strong Cash Flow
For the full year 2025, IBM reported $67.53 billion in revenue. The infrastructure segment contributed $15.72 billion, above the $14.0 billion from the previous year, up from $62.75 billion overall.
Additional metrics showed IBM’s adjusted EBITDA increased by 17% year-over-year, rising by $2.8 billion. The gross margin for the quarter was 60.6%, and operating margin was 91.8%; for the full year, gross margin stood at 58.2% and operating margin at 59.5%. The company closed the year with $14.7 billion in free cash flow.
Quantum Computing: No New Figures but a Clear Roadmap
In quantum computing, IBM did not provide new business figures in this presentation, consistent with recent practice. Nonetheless, Krishna indicated ongoing “constant” progress and confirmed the deployment in December of the first system based on IBM Quantum Nighthawk with 120 qubits for client use. He reiterated ambitions to achieve “Quantum Advantage” by 2026 and to deliver the first large-scale fault-tolerant quantum computer by 2029.
A Retrospective with a Sense of Era
Beyond the financial headlines, the resurgence of the mainframe offers a deeper insight: in the age of AI, not everything is moving toward “more public cloud.” The combination of sovereignty, control, cost efficiency, and integration needs with critical systems is reinvigorating platforms that once seemed relegated to the past. IBM, with the z17 as its flagship, aims to turn this reality into a competitive advantage by promoting modernization without sacrificing control and promising enterprise AI solutions that don’t require everything to leave the premises.
Frequently Asked Questions
Why is the mainframe becoming relevant again in 2026 for large enterprises?
Because many organizations seek digital sovereignty, on-premise control, and unit cost efficiency for critical workloads, along with integrating AI close to sensitive data.
What does a 67% year-over-year growth for IBM Z indicate?
It signifies that the mainframe-related business grew strongly compared to the same period last year, driven by the adoption cycle of z17 and customer renewals.
What is the “GenAI Book of Business,” and why does IBM emphasize it?
It is an internal metric that quantifies the volume of business tied to generative AI projects; IBM uses it to demonstrate commercial traction beyond hardware sales.
Why are DRAM memory prices rising so much, and what is the role of HBM?
According to IBM, part of the industrial capacity is shifting toward HBM, high-bandwidth memory essential for AI servers, which drives up prices and affects the availability of conventional DRAM.