The tech stack feels stable, but four shifts are putting infrastructure under strain
The tech stack is being redesigned in place. Infrastructure is concentrating and diversifying at the same time. Discovery is shifting from human search to agent mediation. Trust models are under regulatory and technological pressure. Interfaces are consolidating around mobile even as spatial and quantum layers advance in parallel.
These shifts are embedding new dependencies across the tech stack. Systems continue to function but become more complex to coordinate and more difficult to change. This is why the stack can feel stable even as constraints are building beneath the surface.
Individually, these shifts might appear manageable. But they’re reshaping the tech stack in ways that might require a more integrated approach to achieve resilience. Four developments appear to be driving this transformation.
1. The hybrid future might be more converged (and more constrained) than it appears
While organizations are planning for multicloud, multimodal, and multiagent environments, the number of potential paths forward is expanding faster than the architectures designed to support them.
This spans everything from the infrastructure layer to the application layer to the power and hardware required to scale.1 As workloads spread across cloud, edge, and emerging models such as self-hosted AI factories, some applications become disaggregated into systems of agents. Data and storage become a first-order constraint, while networking strategy, speed, and control become critical points of coordination.
While modern technology infrastructure might seem highly distributed in that it connects information across the world, it is actually highly centralized, with just a few vendors running most internet services.2 When one layer3 fails, consequences can cascade, raising resilience challenges and concentration risk concerns that are driving many enterprises to implement hybrid infrastructure strategies.
Despite talk of decentralization, the biggest tech players and dominant platforms aren’t giving up their influence. Instead, partnerships among hyperscalers increasingly reflect shared AI workloads, cross-platform interoperability, and embedded services across ecosystems. For organizations, multicloud is shifting from a way to manage risk to an architectural baseline.4 Amazon, for example, recently announced an emerging content marketplace model in an effort to return traffic to origin sites and explore new content models.5
2. As agentic systems scale, the systems that coordinate them will likely become the new control layer
If infrastructure is under strain from below, the discovery layer is being rewritten from above. Modern infrastructures are often built across cloud systems and search capabilities: keyword queries, ranked links, and ad-driven traffic flows.
The most profound structural shift might be the emergence of agentic systems refining application protocols, interoperability, and the digital control plane for a new age. Instead of navigating platforms manually, users are increasingly delegating tasks to AI assistants that search, compare, transact, and coordinate on their behalf. In this model, digital engagement becomes algorithmically mediated.
An “Internet of Agents” envisions autonomous systems discovering, negotiating, and executing across platforms without direct human intervention.6 This shift alters the structure of the information superhighway and the supporting infrastructure as we have come to know it.7
While nearly every technology provider has an AI agent capability, standards will guide engineering principles. Google’s A2A was an early effort to define inter-agent interoperability standards and was subsequently donated to the Linux Foundation to advance open and secure agent-to-agent communication.8 Cisco has proposed an “Internet of Cognition” architecture designed to enable shared context, common goal states, and cross-agent learning—laying the groundwork for more coherent, scalable multiagent collaboration.9
The rise of agentic systems is not just a shift in interaction but a restructuring of data architecture. True multiagent coordination also requires persistent memory, common goal states, and interoperable trust frameworks.10 Efforts such as MIT’s Project NANDA, which focuses on decentralized agent coordination, are exploring decentralized discovery, identity, and trust frameworks so agents can verify capabilities and coordinate without centralized bottlenecks.11 This elevates data from a passive asset to an actively orchestrated layer of the tech stack. As a result, vector databases, retrieval pipelines, and data permissioning systems are becoming as critical as computing infrastructure itself.
While enterprises are focused on the interaction between data and applications, they’re less focused on how application decisions and infrastructure decisions coexist. Agentic orchestration and hybrid infrastructure planning are often happening in silos, even as their dependencies deepen.
Orchestrating across AI agents that operate on shared protocols can unlock new efficiencies, allowing organizations to focus engineering resources instead on next-level challenges such as contextual engineering, process redesign, and human-in-the-loop governance.
Forward-deployed engineering—an operating model where technology engineers are embedded directly in a customer’s environment to build, customize, and deploy solutions in real time—can support this shift. These teams can partner with process owners to define measurable outcomes, pressure-test protocols, and build in governance early so multiagent systems can scale from pilots into enterprise-ready systems.12
3. Interfaces are moving into the real world, so network performance becomes user experience
Enterprise investment is shifting away from fully immersive environments toward lightweight, persistent interfaces embedded in the physical world. Capital is moving from metaverse platforms to spatial computing, smart glasses, and ambient AI—signaling that the next interface won’t replace mobile but extend it into everyday environments.13 This shift puts new pressure on network performance, edge computing, and real-time data synchronization as core infrastructure requirements.
Growth in smart-glasses adoption suggests the real opportunity is persistent, real-world context—placing information directly in a user’s field of vision. Counterpoint Research reports that the smart-glasses category accelerated sharply in the second half of 2025, with shipments up 139% year over year.14 This could mean new capabilities are necessary, from building for augmented reality glasses and omniverse platforms to working with physical spatial data, hologram technology, or even the “invisible” interface of voice for interacting with the world around us.
As a result, the invisible network—Wi‑Fi 6/6E/7, private 5G, and edge connectivity—becomes more important, as it determines speed, reliability, and safety for real-world use cases. As virtual agents extend into physical and spatial AI through sensors, geolocation, or robotics, the interface shift becomes inseparable from network design and data pipelines.
4. Quantum is not a future layer but a breaking point for today’s trust architecture
Post-quantum encryption, quantum networking, and advances in photonic transmission are already forcing organizations to reassess how identity, security, and data integrity are designed across hybrid environments. The issue is architectural exposure. Systems being built today may not be resilient to the cryptographic standards of tomorrow, embedding long-term risk directly into infrastructure decisions.
If agents are going to operate with real autonomy across systems, organizations will likely need new ways to assign identity, define permissions, govern behavior, and ensure the integrity of the transactions those agents carry out. This points to a more adaptive control layer that can monitor and manage systems of agents in real time.
Trust architectures such as distributed ledgers and immutable records could also play an important role in strengthening accountability, verification, and auditability as autonomous activity scales. At the same time, standards are beginning to evolve. The World Wide Web Consortium is advancing decentralized identity and verifiable credential standards,15 including decentralized identifiers.16 Google, is planning to complete its post‑quantum cryptography migration by 2029.17
Yet, quantum and cyber remain an X factor. Reliable solutions are needed, but only 38% of global cyber decision-makers anticipate a transition to post-quantum encryption within the next three to four years, according to Deloitte’s Global Future of Cyber study. The study similarly suggests that adversarial AI is a top concern among cyber leaders, reinforcing the need for new approaches for data security, sovereignty, and encryption. Meanwhile, advancements in photonic integration and fiber-based quantum state transmission are creating testbeds for new forms of networking.18