Futurum Research 2026: Key Issues and Predictions

Futurum Research 2026

Key Issues & Predictions

Welcome to
Futurum’s 2026 Key Issues & Predictions Report

From AI Experimentation to Operational Excellence

As we enter the second half of 2026, the honeymoon phase of AI experimentation has officially ended. We’ve moved past the “science projects” and are now facing the cold, hard reality of operationalizing these technologies at scale. At Futurum Research, we’ve been closely tracking this shift. The theme for this year remains clear: execution over hype. Five critical pivots are defining the 2026 agenda:

• The Infrastructure & Supply Chain Wall: Physical deployment barriers will spike the costs of AI factory builds due to acute shortages in transformers, fiber, and NAND flash. This physical friction will temporarily break deflationary trends, creating an intermediate re-rating of token prices higher.
• The Sovereign AI Substance Test: The principle established by the June 2026 US export directive proved that model access is revocable at the source by originating jurisdictions. Regulated enterprise buyers must now mandate native, regional, or customer-controlled architectures to survive procurement scrutiny.
• The Agentic Governance & Identity Crisis: Massive enterprise token volume has matured into a heavy utility load, with over 54% of decision-makers generating 51 trillion or more tokens annually. This shift to autonomous execution breaks traditional IAM systems, requiring specialized authorization governance to prevent identity sprawl, unmonitored actions, and runaway token costs.
• The Value-Based Shortlist Mandate: Non-seat-based pricing has become table stakes, with 70% of enterprise buyers favoring consumption or outcome-linked models. Pure outcome-based models have nearly doubled in preference to 27%, forcing a rapid commercial restructuring across SaaS vendors.
• The Relocation of Buying Authority: As AI removes the production constraint, value is re-pooling around integration, verification, and orchestration. Buying authority is following this value, moving up toward outcome-tied CIOs and out to business-unit leaders who utilize ecosystems like OpenAI and Anthropic as the default corporate fabric.
• The Platform Ecosystem Gravity Shift: Core AI platforms are becoming the gravitational centers of enterprise technology, accelerating adoption through multiple interconnected GTM layers. By embedding directly across compute, GSI channels, and transactional marketplaces, this multi-vector distribution bypasses traditional SaaS rails and forces legacy vendors to integrate natively to remain competitive.

This isn’t just about technical upgrades; it’s a structural rebalancing of cloud strategy, data governance, and talent. This year’s winners won’t just have the smartest models, they’ll have the most resilient, cost-efficient, and reliable architectures to run them. Here’s to shaping what’s next, together.

Intelligent Devices: ROI, Security, Efficiency, and UX Will Continue to Drive Expansion of AI Capability to the Edge Beyond 2026

Prediction:

High-performance, power-efficient AI-capable silicon will continue to enable increasingly sophisticated AI use cases at the edge, accelerating the expansion of AI workloads into edge form factors such as devices, vehicles, IoT, and other physical AI. These edge AI use cases will increasingly complement cloud-based AI functions in an orchestrated, distributed hierarchy of AI workloads.

Physical AI is most often used to refer to robots, but it increasingly encompasses every category of AI-enabled device that is already in use today – our phones, AI-enabled PCs, smart speakers, fitness trackers, Agentic glasses, and AI-enabled cars – and connects us to a digital assistant, an AI agent, or any kind of AI-enabled feature. But to drive demand at scale, this interconnected ecosystem of devices needs to start solving real problems for real people, or demand may soften.”

Enterprise Software & Digital Workflows: Pricing Optionality to Become Tables-Stakes by Q4 2026

Prediction:

By the end of Q4 2026, vendors will need to offer multiple pricing models, including value-linked approaches, to respond to buyer pressure and the need to more closely manage value, cost predictability, and scale. Vendors that fail to provide flexible pricing approaches will find themselves struggling to gain consideration by potential buyers.

“Enterprise buyers have made their pricing preference clear: seat-based licensing is out for core software, but when it comes to AI, they want cost predictability more than they want to chase usage or outcomes. Small and very large enterprises have established a preference for outcome-based AI pricing, while companies in the middle prefer more traditional add-on or consumption pricing. That split creates a real tension for vendors, and those vendors clinging to rigid structures will find themselves quietly dropped from consideration, not because their product is weaker, but because their commercial model doesn’t match how buyers are balancing risk and value today.”

Software Lifecycle Engineering: Interoperable Control Planes will Become the Primary Mechanism for Proving AI Outcomes, Prioritizing Open Standards Over Single-Stack Depth

Prediction:

By the end of 2026, the control plane will become the layer where enterprises prove AI outcomes, and for agent-driven work specifically, interoperability through open standards like MCP and A2A, across the models and tools a customer already runs, rather than depth in any single stack, decides which vendors own it. The deployment-to-value gap is already measurable: GitHub Copilot is deployed at 59% of organizations, yet the plurality of adopters have licensed it to no more than 20% of their developers¹.

¹ Source: Futurum ETR AI Product Series, January 2026

“The control plane race in 2026 stops being about who deploys the most agents and becomes about who can prove the agents they deployed are worth what they cost and are safe to expand. The vendors that win own interoperable execution authority across the models a customer already runs, because by year-end the CIO question is not whether agents work, it is whether you can show they moved the needle for the business.”

Observability: Observability-Native Sets the Ceiling on Agent Autonomy

Prediction:

By the end of 2026, the depth of an enterprise’s observability sets the hard ceiling on how much agent autonomy it will grant, and vendors whose platforms cannot capture the full decision cycle of intent, reasoning, constraints, and outcomes fall behind the market as buyers cap them at low-risk use cases. The demand signal is already in procurement: 37.4% of decision-makers rank AI observability a priority in platform selection, and 30.9% rank AI agent observability specifically, placing both ahead of distributed tracing at 23.7%².

² Source: 1H 2026 Software Lifecycle Engineering Decision-Maker Survey, Futurum Research, January 2026

“In 2026, observability stops being how you troubleshoot agents and becomes how much you are willing to let them do. The enterprises that can see an agent’s intent, reasoning, constraints, and outcomes will safely expand autonomy, and the ones that cannot will keep their agents boxed into low-risk work no matter how capable the underlying models become.”

Cybersecurity & Resilience: Agentic AI Identity Sprawl will Break Traditional IAM, Requiring Dedicated Authorization Governance to Prevent Unmonitored Agent Activity

Prediction:

As 2026 ends, enterprises will realize their Identity and Access Management systems are failing because they were built for humans, not goal-directed AI agents. Standard workload identity models designed for service accounts simply do not work for these new entities.

Meanwhile, agent identity sprawl will continue to accelerate due to unmonitored procurement, unmanaged open-source deployments, and agents embedded within SaaS platforms. Organizations will need to urgently build agent authorization governance fit for purpose. Without it, the costly gap between an agent’s authorized capabilities and its actual actions will become a severe liability.

The IAM problem with agents is not that we lack the right credentials standard. It is that we built access control around principals that have accountability, and agents have none. Every control we designed assumes a principal with something to lose. Agents operate outside that contract entirely. The organizations getting ahead of this are mapping what they actually have, categorizing it honestly, and governing proportionally to what each category can do and what it can cost when it goes wrong.”

Data Intelligence, Analytics, & Infrastructure: Data Control Planes will Shift Focus from Passive Insight Generation to Governed, Autonomous Execution of Business Actions

Prediction:

In the second half of 2026, the focus for data and AI leaders will shift from generating trusted insights to executing governed actions. We spent the first half of the year building semantic layers to help models understand business context. Now, organizations face a harder problem: letting agents act on that data in a safe, governed, and performant manner.

“We spent the early part of the year getting models to understand our data. Now we have to figure out if we can safely let them touch it. The market is moving from generating answers to executing actions, which changes the baseline requirements for databases and catalogs. The goal isn’t just building a smarter agent. It’s about letting an agent change a record and being able to prove exactly who authorized it, why it happened, and how to reverse it.”

AI Platforms: Sovereign AI's Substance Test Arrives

Prediction:

By the end of 2026, the principle that AI model access is revocable by originating jurisdictions, demonstrated by the June 2026 US export directive, will become a critical factor in mainstream enterprise procurement. Enterprises can no longer assume “sovereign” labels provide immunity from revocation. This directive proved that, regardless of regional infrastructure or contracts, access remains subject to unilateral control. Consequently, regulated buyers will now scrutinize every AI service for this risk. Vendors unable to provide a credible continuity strategy for their sovereign-label offerings will face exclusion from regulated-sector tenders, as compliance and procurement teams prioritize jurisdictional resilience over simple hosting location.

“The June 2026 export control directive established something the market has not yet absorbed: that model access is revocable by the originating jurisdiction regardless of sovereign infrastructure, regional deployment, or contractual commitment. Most enterprise buyers have treated it as a frontier-model security story with no implications for their own deployments. That reading will not survive the scrutiny of legal, compliance, and procurement functions as AI moves into live, regulated workloads. The question is no longer only where a model is hosted or who operates the infrastructure; it is what happens to a regulated deployment if access is suspended at the source. Vendors with a credible answer to that question are in a different position than those without one. Sovereign AI’s substance test has arrived, and it arrived with a lot more impact than anyone had predicted.”

Networking: East-West Traffic Dominates Data Centers

Prediction:

AI usage has changed traffic patterns in the enterprise data center. Traditional user-focused flows to servers (north-south) have given way to server-to-server traffic (east-west). By the end of 2026, east-west traffic will account for 90% of all data center traffic flows.

“Networks evolve because of traffic. Cloud computing did not change how traffic flows from user to server. AI is fundamentally different because of East-West communication. Practitioners need to understand how to deploy new designs to utilize hardware efficiently and why old-school thinking will only lead to pain down the road.”

Semiconductors, Supply Chain, and Emerging Tech: Friction at the Infrastructure Frontier Will Drive Token Prices Higher

Prediction:

In the back half of this year, the AI infrastructure market will collide with a structural paradox. While enterprise token demand has reached a multi-trillion-dollar utility scale, physical system integration barriers will temporarily break the historical trend of deflationary token pricing. The extreme complexity and structural immaturity of next-generation, liquid-cooled rack-scale architectures, combined with rigid enterprise capital constraints, will cause an intermediate re-rating of token prices higher as cutting-edge reasoning models debut before next-gen token factories can fully scale.

“We have officially collided with the physical friction points of the AI buildout. As enterprise demand scales into multi-trillion token agentic loops, the bottleneck has fundamentally shifted from software capabilities to the raw engineering maturity of the data center floor. The resulting token price paradox means raw compute is no longer a deflationary commodity but a tightly rationed premium utility. In this environment, the ultimate competitive weapon is token price relief.”

Hybrid Cloud & Infrastructure: Building New Mega-AI Datacenters will Stall as Supply-chain Issues Bite, Reducing Pressure on RAM and SSD Prices

Prediction:

By the start of 2027, many of the massive AI data center builds announced in late 2025 and early 2026 will be delayed or canceled. Several data center builds have already been blocked by communities and State governments; this trend will continue despite federal efforts to clear the blocks. A harder issue is the supply chain for essential components such as power transformers and optical fibers. High-power electrical transformers are needed between the power grid (or local power generation) and computers; supply is now back-ordered for years. Optical fiber faces another supply challenge, partly due to the war in Ukraine, which has consumed millions of miles of fiber for drones.

“Commodity RAM and SSD prices have risen sharply, and will continue to affect purchasing decisions until other supply chain issues slow the AI wave. Delays and cancellations of massive AI data center projects will relieve pressure on these prices, but not soon enough for many customers.”

Ecosystems, Channels, & Marketplaces:
OpenAI and Anthropic Move into the Gravitational Center of Enterprise Tech

Prediction:

By the end of 2026, enterprise software vendors will face a stark architectural and commercial choice as OpenAI and Anthropic move into gravitational centers of enterprise tech, mandating native platform integration to remain competitive on purchase shortlists. This structural shift is supported by ETR’s March 2026 data, which shows OpenAI’s GPT “o-series” models leading enterprise usage at 57%, while Anthropic’s Claude surged from 21% to 48% within a single year, establishing these two players as the primary engines driving enterprise AI adoption.

“The road to agentic commerce has truly begun, and cloud marketplaces are the critical nexus of this evolution. By providing the governance and interoperability required for autonomous agents to negotiate and transact at scale, marketplaces are no longer just a procurement option; they are the engine driving the next flywheel of enterprise software usage and service consumption.”

Digital Leadership, CIOs & Tech Buyers:
As AI Removes the Production Constraint, Value and Buying Authority Relocate, and Most Vendor Go-to-Market is Aimed at the Wrong Buyer

Prediction:

The defining enterprise-technology shift in 2H 2026 is relocation. Adoption is no longer the story. As agentic systems move into operations, production stops being the scarce step. The binding constraint moves to governing, integrating, and verifying work that systems now generate at scale. Value collapses where production used to be the bottleneck and re-pools around the new one. Buying authority follows the value: up toward CIOs who tie technology to business outcomes, and out toward field CTOs and business-unit leaders who hold budget and execution. Vendors repositioning to sell to the CIO are often aiming at the wrong buyer. Vendors selling production acceleration alone are aiming at the wrong layer.

“Adoption was the easy part. In the second half of 2026, the production bottleneck dissolves, and value moves to governing and integrating what AI now generates at scale. Buying authority moves with it. Having sat in the CIO, CTO, and GM seats, I can tell you the vendors that win the next two quarters are the ones who learn which buyer decides, and which layer they actually sell into, before they rebuild their motion around a title.”