The Intent-Driven Enterprise Stack An Executive Patent White Paper on IBM’s Strategic Direction (20

The Intent-Driven Enterprise Stack

An Executive Patent White Paper on IBM's Strategic Direction (2025–2026)

Prepared by anovIP

Executive Brief

Between Q4-2025 and January 2026, IBM published a dense portfolio of patent applications spanning hybrid cloud orchestration, AI governance, secure multi-tenancy, edge intelligence, explainable AI, privacy-preserving computation, and enterprise automation.

From an intellectual-property strategy perspective, these filings are not fragmented innovations. They form a coherent architectural thesis:

The next enterprise computing era will be intent-driven, continuously adaptive, explainable, and cryptographically trustworthy—across cloud, edge, silicon, and AI models.

This anovIP executive white paper distills IBM's recent patent activity into strategic signals, competitive implications, and board-level insights for technology leaders, investors, and policymakers.

1. Strategic Overview: IBM's Patent Philosophy (2025–2026)

IBM's patent direction diverges sharply from consumer-AI or accelerator-centric strategies. Instead, IBM is codifying enterprise invariants:
a) Intent over configuration
b) Trust by design, not compliance afterthoughts
c) AI systems that explain, audit, and self-correct
d) Zero-downtime, always-on operations from core to edge
e) Privacy enforced mathematically, not contractually

Collectively, these patents define what anovIP characterizes as: The Trusted Enterprise Nervous System

2. Intent as the New Control Plane

Patent Highlight

US 20260030202 – Intent-Based Container Image Building

Publication Number: US20260030202A1

Publication Date: January 29, 2026

Applicant: International Business Machines Corporation

Abstract: Methods, computer program products, and systems are presented. The method computer program products, and systems can include, for instance: performing natural language processing to process a text string of a user, wherein the text string specifies characteristics of a container image to be built; processing, with use of natural language processing, instances of text-based data that describe respective ones of a plurality of container images stored within a container image repository; selecting, in dependence on a result of the performing natural language processing, and the processing, a base image from the plurality of container images; and presenting prompting data to the user that prompts building of a new container image, wherein the prompting data references the base image.

First Indepedendent Claim: A computer implemented method comprising:

performing natural language processing to process a text string of a user, wherein the text string specifies characteristics of a container image to be built;

processing, with use of natural language processing, instances of text-based data that describe respective ones of a plurality of container images stored within a container image repository;

selecting, in dependence on a result of the performing natural language processing, and the processing, a base image from the plurality of container images; and

presenting prompting data to the user that prompts building of a new container image, wherein the prompting data references the base image.

What IBM Patented

anovIP Insight

IBM is quietly replacing infrastructure-as-code with infrastructure-as-language. This positions natural language not as a chatbot layer, but as a first-class orchestration primitive.

Executive Implication

Whoever owns the intent layer owns developer productivity, platform gravity, and long-term lock-in.

3. Trust Starts Below Software: Memory-Level Isolation

Patent Highlight

US 20260030176 – Condensed Memory-Management Protection in Multi-Tenant Systems

Publication Number: US20260030176A1

Publication Date: January 29, 2026

Applicant: International Business Machines Corporation

Abstract: An embodiment includes configuring a configuration of a crossbar of a memory management unit comprising a set of regions of N workers and D partitions. The embodiment includes writing by a worker, to D partitions in the set of regions of the crossbar. The embodiment also includes reading by a partition, N regions in the set of regions of the crossbar where each of the N regions is assigned an address that is mapped to a physical memory address and where an access control is achieved by the configuration of a N plus D mapping of workers to partitions of the crossbar of the memory management unit.

First Indepedendent Claim: A computer-implemented method comprising:

configuring a configuration of a crossbar of a memory management unit comprising a set of regions of N workers and D partitions;

writing by a worker, to D partitions in the set of regions of the crossbar; and

reading by a partition, N regions in the set of regions of the crossbar wherein each of the N regions is assigned an address that is mapped to a physical memory address and wherein an access control is achieved by the configuration of a N plus D mapping of workers to partitions of the crossbar of the memory management unit.

What IBM Patented

anovIP Insight

This patent reinforces IBM's historical strength: hardware-rooted trust. As AI workloads become multi-tenant and regulator-visible, software isolation alone is insufficient.

Executive Implication

Confidential computing is moving from "premium feature" to regulatory baseline.

4. Silicon Still Matters: Micro-Efficiency as Strategy

Patent Highlight

US 20260030028 – Branch Prediction Correction Based on Nonuse of Relevant Prediction Structure

Publication Number: US20260030028A1

Publication Date: January 29, 2026

Applicant: International Business Machines Corporation

Abstract: A branch prediction unit of the processor powers-up and accesses only a subset of a plurality of prediction structures to obtain a first set of branch prediction information for a conditional branch. During the access, at least one of the plurality of prediction structures remains powered-down. The branch prediction unit thereafter determines whether all of the plurality of prediction structures having branch prediction information relevant to the conditional branch were accessed. Based on a determination that fewer than all of the plurality of prediction structures having branch prediction information relevant to the conditional branch were accessed, the branch prediction unit refrains from outputting a branch prediction based on the first set of branch prediction information, powers-up and accesses a greater number of the plurality of prediction structures to obtain a second set of branch prediction information, and outputs a branch prediction based on the second set of branch prediction information.

First Indepedendent Claim: A method of branch processing in a processor, the method comprising:

a branch prediction unit of the processor powering-up and accessing only a subset of a plurality of prediction structures to obtain a first set of branch prediction information for a conditional branch, wherein at least one of the plurality of prediction structures remains powered-down during the accessing, the branch prediction unit comprising a plurality of arrays storing branch prediction information, each array in the plurality of arrays being a prediction structure, an index pipeline including indices of instruction addresses of conditional branch instructions, a prediction pipeline configured to evaluate branch prediction information from accessed prediction structures, a line input buffer configured to buffer branch related information utilized by the prediction unit to access the prediction structures, a latency accelerator, wherein the latency accelerator is an array of entries configured to store indices of prediction structures and a power mode field configured to be utilized by the branch prediction unit to predict an appropriate power state of the prediction structures during a prediction access, and a regulator circuit configured to control powering down of the prediction structures;

thereafter, the branch prediction unit determining whether all of the plurality of prediction structures having branch prediction information relevant to the conditional branch were accessed; and

based on a determination that fewer than all of the plurality of prediction structures having branch prediction information relevant to the conditional branch were accessed: the branch prediction unit refraining from outputting a branch prediction based on the first set of branch prediction information accessed from the subset of the plurality of prediction structures;

the branch prediction unit powering-up and accessing a greater number of the plurality of prediction structures to obtain a second set of branch prediction information; and

the branch prediction unit outputting a branch prediction based on the second set of branch prediction information; and

the branch prediction unit detecting a prediction structure subset changing (PSSC) event, wherein the PSSC event includes a line split event in which at least one addition conditional branch instruction is encountered in a given instruction cacheline and the at least one additional conditional branch instruction maps to one or more additional subarrays that have not been previously powered-on; and

based on detecting the PSSC event, powering-up and accessing all of the plurality of prediction structures for one or more branch predictions in a limited time window.

What IBM Patented

anovIP Insight

Even in an AI-first world, IBM continues to invest in foundational CPU intelligence—especially relevant for mainframes, financial systems, and sovereign infrastructure.

Executive Implication

Energy efficiency and determinism remain competitive differentiators in regulated enterprise compute.

5. Observability Evolves into Autonomic Intelligence

Patent Highlight

US 20260023672 – Dynamic Logging

Publication Number: US20260023672A1

Publication Date: January 22, 2026

Applicant: International Business Machines Corporation

Abstract: Methods, computer program products, and systems are presented. The method computer program products, and systems can include, for instance: generating log messages from one or more data source; evaluating one or more log message produced from the generating, and outputting, in dependence on the evaluating the one or more log message, a log message rating of the one or more log message; evaluating a logging system, and outputting, in dependence on the evaluating the logging system, a logging system rating of the logging system, wherein the logging system includes a logging volume for storing log messages; comparing the log message rating to the logging system rating; and producing an action decision impacting a count of log messages stored in the storage volume in dependence on the comparing of the log message rating, and the logging system rating.

First Indepedendent Claim: A computer implemented method comprising:

generating log messages from one or more data source;

evaluating one or more log message produced from the generating, and outputting, in dependence on the evaluating the one or more log message, a log message rating of the one or more log message;

evaluating a logging system, and outputting, in dependence on the evaluating the logging system, a logging system rating of the logging system, wherein the logging system includes a logging volume for storing log messages;

comparing the log message rating to the logging system rating; and

producing an action decision impacting a count of log messages stored in the storage volume in dependence on the comparing of the log message rating, and the logging system rating.

What IBM Patented

anovIP Insight

IBM reframes observability as a self-optimizing system, not a data exhaust problem.

Executive Implication

AIOps is transitioning from dashboards to closed-loop automation.

6. AI for Regulated Decision Support (Healthcare Case Study)

Patent Highlight

US 20260013800 – Predicting Glucose Values Using Overlapping Time Windows

Publication Number: US20260013800A1

Publication Date: January 15, 2026

Applicant: International Business Machines Corporation

Abstract: A computer-implemented method and system for predicting and displaying glucose values, including receiving CGM data, determining, based on the data, a plurality of first predicted glucose values (33) for a first prediction time window (30), determining, based on the data, that a hypoglycemia event is predicted to occur during a second prediction time window (31) which has a contemporaneous beginning with the first prediction time window (30) but is shorter than the first window (30), and determining a plurality of second predicted glucose values (34) for the second prediction time window (31) and displaying the plurality of second predicted glucose values (34) for the second prediction time window (31) while not displaying predicted glucose values subsequent to the second prediction time window (31).

First Indepedendent Claim: A computer-implemented method for predicting and displaying glucose values, the method being carried out in a system with at least one data processing device (1), the method comprising:

receiving continuous glucose monitoring data indicative of a glucose level in a bodily fluid from a continuous glucose monitoring system worn by a user;

receiving the user's carbohydrate ingestion data and insulin administration data;

determining, based on the continuous glucose monitoring data, the carbohydrate ingestion data and the insulin administration data, a plurality of first predicted glucose values (33) for a first prediction time window (30);

determining, based on the continuous glucose monitoring data and the carbohydrate ingestion data that a hypoglycemia event is predicted to occur during a second prediction time window (31) which has a contemporaneous beginning with the first prediction time window (30) but is shorter than the first prediction time window (30); and

determining a plurality of second predicted glucose values (34) for the second prediction time window (31) and displaying the plurality of second predicted glucose values (34) for the second prediction time window (31) while not displaying predicted glucose values subsequent to the second prediction time window (31).

What IBM Patented

anovIP Insight

IBM's healthcare AI patents emphasize actionable foresight, not black-box prediction—consistent with clinical accountability requirements.

Executive Implication

Decision-support AI, not autonomous AI, will dominate regulated verticals.

7. Orchestration as a Learning System

Patent Highlight

US 20260004181 – Intelligent Orchestration

Publication Number: US20260004181A1

Publication Date: January 1, 2026

Applicant: International Business Machines Corporation

Abstract: Methods, computer program products, and systems are presented. The method computer program products, and systems can include, for instance: training a machine learning model in dependence on traffic between at least a first service and a second service defining an application; querying the machine learning model; generating service characterizing data that characterizes at least one service defining the application, wherein the generating the service characterizing data is in dependence on the querying of the machine learning model; and modifying a performance attribute of the application in dependence on characterizing data of the service characterizing data.

First Indepedendent Claim: A computer implemented method comprising:

training a machine learning model in dependence on traffic between at least a first service and a second service defining an application;

querying the machine learning model;

generating service characterizing data that characterizes at least one service defining the application, wherein the generating the service characterizing data is in dependence on the querying of the machine learning model; and

modifying a performance attribute of the application in dependence on characterizing data of the service characterizing data.

What IBM Patented

anovIP Insight

Applications are treated as living graphs, continuously learning from their own runtime behavior.

Executive Implication

Static orchestration is becoming obsolete in complex hybrid environments.

8. Edge Intelligence Without Downtime

Patent Highlights

US 20250328338 – Hot Upgrade Workflow in an Edge Computing Environment

Publication Number: US20250328338A1

Publication Date: October 23, 2025

Applicant: International Business Machines Corporation

Abstract: An embodiment includes detecting by a Process Monitor of a system, a hot upgrade from a first version to a second version. The embodiment includes responsive to the detecting the hot upgrade, predicting by a Process Analyzer a predicted process instance. The embodiment includes generating by the Process Analyzer an upgrade workflow based on a current process instance, a progress metric of the current process instance and the predicted process instance. The embodiment also includes deploying by a Device Upgrade Manager the hot upgrade to the second version on a device of the system based on the upgrade workflow.

First Indepedendent Claim: A computer-implemented method comprising:

detecting, by a Process Monitor of a system, a hot upgrade from a first version to a second version;

responsive to the detecting the hot upgrade, predicting by a Process Analyzer a predicted process instance;

generating by the Process Analyzer an upgrade workflow based on a current process instance, a progress metric of the current process instance and the predicted process instance; and

deploying by a Device Upgrade Manager the hot upgrade to the second version on a device of the system based on the upgrade workflow.

What IBM Patented

US 20250342090 – Technology-Agnostic Backup

Publication Number: US20250342090A1

Publication Date: November 6, 2025

Applicant: International Business Machines Corporation

Abstract: An embodiment includes a backup trigger by a system. The embodiment includes responsive to detecting the backup trigger, processing by a Backup Executor of the system a manifest comprising of a component for backup. The embodiment also includes orchestrating, by the Backup Executor, the backup between an adaptor proxy of the component in a first location and an adaptor of the component in a second location wherein the orchestrating comprises invoking the adaptor proxy by the Backup Executor wherein the adaptor proxy performs a get from the component in the first location and performs a put to the adaptor of the component in the second location and wherein the Backup Executor is technology agnostic.

First Indepedendent Claim: A computer-implemented method comprising:

detecting a backup trigger by a client system in a first location;

responsive to detecting the backup trigger, processing by a Backup Executor of the client system a manifest comprising of a component on the client system for backup; and

orchestrating, by the Backup Executor, the backup between an adaptor proxy of the component in the client system in the first location and an adaptor of the component in a second location wherein the orchestrating comprises invoking the adaptor proxy in the client system in the first location by the Backup Executor wherein the adaptor proxy performs a get from the component in the first location and performs a put to the adaptor of the component in the second location and wherein the Backup Executor is technology agnostic.

What IBM Patented

anovIP Insight

IBM is engineering for physical irreversibility—systems deployed where downtime or manual intervention is impossible.

Executive Implication

Edge deployments require mainframe-grade reliability.

9. Explainable, Auditable, Governable AI

Patent Highlights

US 20250335798 – Explainer Model Evaluation and Training

Publication Number: US20250335798A1

Publication Date: October 30, 2025

Applicant: International Business Machines Corporation

Abstract: An embodiment includes detecting an explainer model check by a system. The embodiment includes responsive to the detecting the explainer model check, computing a first result by a Data and Model Preparation of the system wherein the first result is based on a first dataset and a second data set generated by the Data and Model Preparation. The embodiment includes generating a second result by an explainer model of a Prediction and Explanation of the system based on the first dataset and the second data set. The embodiment includes computing a difference metric between a first result and a second result by a Judgment Retraining of the system. The embodiment also includes training the explainer model based on the difference metric.

First Indepedendent Claim: A computer-implemented method comprising:

detecting an explainer model check by a system;

responsive to the detecting the explainer model check, computing a first result by a Data and Model Preparation of the system wherein the first result is based on a first dataset and a second data set generated by the Data and Model Preparation;

generating a second result by an explainer model of a Prediction and Explanation of the system based on the first dataset and the second data set;

computing a difference metric between a first result and a second result by a Judgment Retraining of the system; and

training the explainer model based on the difference metric.

US 20250335800 – Probabilistic Black-Box Anomaly Attribution

Publication Number: US20250335800A1

Publication Date: October 30, 2025

Applicant: International Business Machines Corporation

Abstract: An embodiment identifies, by a probabilistic black-box anomaly attribution engine, an anomalous sample in test data associated with a black-box model, the black-box model comprising a plurality of variables. The embodiment generates, by the probabilistic black-box anomaly attribution engine, a variable distribution based on the test data using a plurality of outputs generated using a plurality of perturbations. The embodiment generates, by the probabilistic black-box anomaly attribution engine based on the variable distribution, an attribution score representing a responsibility of a variable for the anomalous sample.

First Indepedendent Claim: A computer-implemented method comprising:

identifying, by a probabilistic black-box anomaly attribution engine, an anomalous sample in test data associated with a black-box model, the black-box model comprising a plurality of variables;

generating, by the probabilistic black-box anomaly attribution engine, a variable distribution based on the test data using a plurality of outputs generated using a plurality of perturbations; and

generating, by the probabilistic black-box anomaly attribution engine based on the variable distribution, an attribution score representing a responsibility of a variable for the anomalous sample.

US 20250342181 – Ranking-Augmented Generation for Long Documents

Publication Number: US20250342181A1

Publication Date: November 6, 2025

Applicant: International Business Machines Corporation

Abstract: A computer-implemented method comprising: receiving, as input, a query and a source document intended for a content-grounded question-answering or multi-turn conversation task by a specified large language model (LLM) which has a context window size limit, wherein the source document has a size which exceeds the context window size limit; dividing the source document into a plurality of segments; applying a language model to each of the segments, to assign to each of the segments a relevance score; selecting the k-top segments having the highest the relevance scores; combining the selected k-top segments into a virtual document having a size which complies with the context window size limit; and feeding the virtual document as input to the specified LLM, to generate a response that is grounded in the content of the virtual document.

First Indepedendent Claim: A computer-implemented method comprising:

receiving, as input, a query and a source document intended for a content-grounded question-answering or multi-turn conversation task by a specified large language model (LLM) which has a context window size limit, wherein said source document has a size which exceeds said context window size limit;

dividing said source document into a plurality of segments;

applying a language model to each of said segments, to assign to each of said segments a relevance score;

selecting the k-top segments having the highest said relevance scores;

combining said selected k-top segments into a virtual document having a size which complies with said context window size limit; and

feeding said virtual document as input to the specified LLM, to generate a response that is grounded in the content of said virtual document.

What IBM Patented

anovIP Insight

IBM continues to lead in explainability as infrastructure, not UX polish.

Executive Implication

AI governance is becoming a technical capability, not a policy document.

10. Privacy by Mathematics

Patent Highlight

US 20250307242 – Database Table Joining Under Fully Homomorphic Encryption

Publication Number: US20250307242A1

Publication Date: October 2, 2025

Applicant: International Business Machines Corporation

Abstract: An embodiment appends, into a concatenated table, a second plurality of records in a second table to a first plurality of records in a first table. An embodiment sorts, according to each identification value in the concatenated table, the concatenated table. An embodiment generates, using an equality mask derived from each identification value in the sorted table, an intersection table, the intersection table comprising a record in the first plurality of records with a first identifier value matching a second identifier value in a record in the second plurality of records. An embodiment generates, using a not-in-intersection mask derived from the equality mask, a not-in-intersection table. An embodiment adds contents of the intersection table and contents of the not-in-intersection table together, resulting in a join table.

First Indepedendent Claim: A computer-implemented method comprising:

appending, into a concatenated table, a second plurality of records in a second table to a first plurality of records in a first table, wherein the first table and the second table each comprise an identification field storing an identification value, wherein the first plurality of records and the second plurality of records in the second table comprise fully homomorphic encrypted data;

sorting, according to each identification value in the concatenated table, the concatenated table, the sorting resulting in a sorted table;

deriving, using each identification value in the sorted table, an equality mask corresponding to the sorted table;

generating, using the equality mask derived from each identification value in the sorted table, an intersection table, the intersection table comprising a record in the first plurality of records with a first identifier value matching a second identifier value in a record in the second plurality of records;

deriving, from the equality mask, a not-in-intersection mask;

generating, using the not-in-intersection mask derived from the equality mask, a not-in-intersection table, the not-in-intersection table comprising a record in the first plurality of records with a third identification value failing to match any identification value in a record in the second plurality of records; and

adding contents of the intersection table and contents of the not-in-intersection table together, the adding resulting in a join table.

What IBM Patented

anovIP Insight

This is a cornerstone patent for post-trust data economies, where even operators cannot see data.

Executive Implication

Future compliance regimes may require computation on encrypted data.

11. Business-Weighted Cyber Risk

Patent Highlight

US 20250298891 – Connected Asset Risk Management

Publication Number: US20250298891A1

Publication Date: September 25, 2025

Applicant: International Business Machines Corporation

Abstract: An embodiment extracts, from vulnerability data describing a vulnerability applicable to a connected asset within a network of connected assets, a set of technical impacts of the vulnerability. An embodiment generates, using a business value context of the connected asset and the set of technical impacts of the vulnerability, a network layer impact score corresponding to the connected asset and the vulnerability. An embodiment generates, using the business value context of the connected asset and the set of technical impacts of the vulnerability, an enterprise layer impact score corresponding to the connected asset and the vulnerability. An embodiment generates, using the network layer impact score and the enterprise layer impact score, a remediation plan for the connected asset, the remediation plan comprising a planned adjustment of the connected asset to ameliorate the vulnerability.

First Indepedendent Claim: A computer-implemented method comprising:

extracting, from vulnerability data describing a vulnerability applicable to a connected asset within a network of connected assets, a set of technical impacts of the vulnerability;

generating, using a business value context of the connected asset and the set of technical impacts of the vulnerability, a network layer impact score corresponding to the connected asset and the vulnerability;

generating, using the business value context of the connected asset and the set of technical impacts of the vulnerability, an enterprise layer impact score corresponding to the connected asset and the vulnerability; and

generating, using the network layer impact score and the enterprise layer impact score, a remediation plan for the connected asset, the remediation plan comprising a planned adjustment of the connected asset to ameliorate the vulnerability.

What IBM Patented

anovIP Insight

Risk is no longer binary or technical—it is contextual, financial, and operational.

Executive Implication

CISOs and boards are converging on a shared risk language.

12. anovIP Strategic Assessment

IBM's Differentiation

Competitive Positioning

IBM is not competing head-on with GPU or consumer-AI players. It is patenting the rules, guardrails, and continuity mechanisms that large institutions cannot operate without.

Conclusion: IBM Is Codifying the Rules of Enterprise AI

From intent-based infrastructure to homomorphic databases, IBM's patent portfolio signals a long-term bet:

The winners of enterprise AI will not be those who build the smartest models—but those who make intelligence safe, explainable, reliable, and governable at scale.

IBM is patenting that future—quietly, methodically, and defensively.

About anovIP

anovIP is a global IP strategy and analytics firm advising technology companies, enterprises, and investors on patent landscapes, competitive intelligence, and innovation risk. anovIP specializes in translating complex patent activity into clear strategic insight for decision-makers.