The article explores Arun Dalli’s groundbreaking work in strategic technology decommissioning, an emerging discipline focused on creating value through subtraction rather than expansion. As a Senior Staff Capacity Engineer at ServiceNow and former Meta capacity leader, Dalli reclaims financial and environmental resources by retiring obsolete infrastructure. His approach transforms decommissioning into a data-driven, human-centered process that funds future AI and ML growth. The piece reframes infrastructure retirement as a cornerstone of sustainable innovation.
In the race for AI dominance, tech giants are accumulating vast digital boneyards: underutilized servers and legacy code that quietly drain budgets and power grids. This has given rise to the unseen economy of subtraction, where value is created not by what is built, but by what is strategically dismantled.
In the relentless drive for innovation, a counterintuitive discipline is gaining critical importance: the strategic decommissioning of technology. While the industry often focuses on addition, a significant financial and environmental drain comes from aging infrastructure. This challenge requires a specialized focus on retiring technology in a way that funds future advancements.
Arun Dalli, a Senior Staff Capacity Engineer at ServiceNow with over 12 years of experience in hyperscale infrastructure, focuses on this complex discipline. Having previously managed capacity for Meta/Facebook’s massive global infrastructure, Dalli’s work involves retiring thousands of servers to reclaim millions in operational costs, redirecting those resources toward AI and machine learning initiatives.
The process of retiring vast server fleets involves more than just technical logistics; it requires a deep understanding of the human element. The term ‘digital undertaker’ aptly describes the role of presiding over the end-of-life for technology that was once state-of-the-art. Dalli explains, “Strategic decommissioning isn’t just removing equipment; it’s acknowledging when hardware has served its purpose and orchestrating an exit that honors both its service and the people who depend on it.”
This work is human-centric because every server represents past decisions and workflows. Application teams often become attached to their infrastructure, viewing migration as a significant risk.
This organizational inertia, sometimes described as the legacy paradox, stems from a preference for known flaws over the unknown challenges of modernization. “My role was helping teams see that the greater risk was clinging to aging hardware,” Dalli states.
To manage this, he created a clear lifecycle state framework—Decomhold, Migration Hold, and In Sanitization—to make transitions manageable. This structured approach aligns with change management best practices that empower staff by making improvements visible and tangible, thereby reducing resistance.
Strategic decommissioning creates a direct funding mechanism for innovation by eliminating the hidden costs of maintaining obsolete systems. When Dalli accelerated the retirement of over 15,000 servers at ServiceNow, the effort reclaimed significant financial and human resources. This initiative directly addressed the “innovation tax,” where budgets are diverted from value-creating activities to maintaining outdated technology.
“I can draw a very direct line: the $22M annual savings from decommissioning 15,000+ servers at ServiceNow directly funded our AI/ML infrastructure expansion,” Dalli notes. The savings were not just about hardware; they encompassed power, cooling, licensing, and support contracts, freeing up physical datacenter capacity. This reallocation is critical, as organizations with mature data and AI governance see significant improvements in financial performance.
The real cost of delaying such projects extends beyond operational expenses to the opportunity cost of deferred innovation. Dalli emphasizes, “Decommissioning at the right moment prevents the false economy of ‘sweating assets’ beyond their value.” This perspective is critical as organizations grapple with the fact that some spend up to 80% of their IT budgets on maintaining legacy systems, severely limiting their capacity for growth.
In a corporate culture centered on addition and growth, advocating for subtraction requires overcoming deep-seated organizational resistance. Teams often cling to existing infrastructure due to risk aversion and operational familiarity. Dalli identifies a key challenge as the “‘If It’s Not Broken’ Fallacy,” where teams equate “still running” with “still valuable,” ignoring escalating hidden costs.
This resistance is a form of technical debt, where delaying necessary changes accumulates future costs. According to McKinsey, CIOs estimate that 10% to 20% of their technology budget is diverted to resolving such issues.
To build the case for subtraction, Dalli reframed the conversation around the total cost of ownership, using frameworks like the SQALE method to make the financial impact transparent. “I made the operational burden visible, which shifted the conversation from ‘why remove this?’ to ‘why are we still paying for this?’” he says.
Creating safe migration paths with one-to-one capacity mapping provided teams with the confidence that they were not being abandoned. By framing decommissioning as a strategic reallocation of resources, the process was transformed from a perceived loss into a celebrated enabler of future capabilities.
A primary target for decommissioning is the ‘zombie server’—a machine that consumes power and resources while delivering little to no value. These servers often result from workload migrations or deprecated services where the underlying hardware was never formally retired. Dalli explains, “A zombie server decays gradually into a power-draining liability through predictable patterns.”
From a data perspective, these machines exhibit distinct signatures, including CPU utilization consistently below 15-20% for extended periods (30+ days) and network traffic near baseline levels. These underutilized assets are a form of dark data, representing collected but unused resources. At hyperscale, thousands of these servers can drain millions of dollars annually while masquerading as productive capacity.
Identifying these assets requires systematic monitoring and automated detection. “I built continuous monitoring through Tableau dashboards tracking multiple signals,” Dalli says, including lifecycle state duration and cost per computational unit. This detective work uses AI and machine learning principles to discover and classify underutilized assets, turning a manual audit into a continuous optimization process.
Safely decommissioning thousands of servers without disrupting live services is impossible without sophisticated tooling and automation. Predictive models forecast future demand, but real-time visibility is essential for dismantling current infrastructure. Dalli notes, “The critical tooling isn’t about forecasting—it’s about seeing what’s consuming your capacity, at what rate, and making informed decisions with governance guardrails.”
Drawing from his experience at Meta with a tool called OverwatchX, Dalli helped implement a similar system at ServiceNow called CloudLens. These platforms provide a multi-dimensional view of consumption across power, cooling, and cost, ensuring that optimization in one area does not create problems elsewhere. Such tools are crucial for the governance of the entire AI lifecycle, from development to retirement.
Automation is key to operational safety. Dalli highlights, “Automated checks confirming servers are truly idle before entering Decomhold—not just low utilization, but zero production traffic.” This level of AI governance transforms decommissioning from a risky manual process into a data-driven operation, preventing billion-dollar decisions from being made on spreadsheets.
The core principle of strategic decommissioning is the unseen economy of subtraction, which operates on a simple rule: every resource consumed by obsolete infrastructure is a resource unavailable for the future. In a culture obsessed with addition, subtraction is often overlooked, yet it is the mechanism that enables sustainable growth. As Dalli puts it, “Subtraction isn’t just cost avoidance—it’s resource reallocation to higher-value activities.”
Every underutilized server imposes a hidden “Operational Tax of Inaction,” which includes power, cooling, licensing, and engineering time. This is compounded by the complexity of legacy systems, which can escalate transformation costs. One analysis found that an average application carries over $1 million in technical debt, and the unseen economy recognizes that removing this waste prevents future costs from escalating.
This principle also has significant environmental implications. “The most sustainable infrastructure is infrastructure you don’t need,” Dalli states. Decommissioning servers eliminate their power draw and eventual e-waste, contributing more to sustainability than efficiency improvements alone.
Decommissioning is not the end of hardware’s value but a transition point. A single project at ServiceNow generated $8M in hardware reuse value by treating the process as a strategic recovery operation. Dalli emphasizes, “At hyperscale, retired infrastructure still has significant market value if you optimize the recovery process strategically.”
The process involves a hierarchy of value recovery, starting with internal repurposing for less critical workloads like testing environments. For hardware destined for the secondary market, speed is essential, as value degrades rapidly.
Creating clear lifecycle transitions accelerates equipment through the disposition pipeline, capturing higher resale values. This must be balanced with secure data sanitization, following standards like NIST 800-88 to avoid the severe regulatory fines associated with data protection failures.
“The difference between $8M recovered versus $2-3M comes from strategic internal retrofitting, fast pipeline execution, maintaining hardware functionality through ERAD, and strategic batching,” Dalli explains. This disciplined approach to asset recovery can compound into millions of dollars.
As AI development accelerates the cycle of hardware obsolescence, the discipline of strategic decommissioning is evolving from a reactive necessity to a proactive, integral part of infrastructure planning. The traditional “deploy and forget until failure” approach is becoming untenable, creating a need for a more holistic technology lifecycle model. Dalli suggests the role will evolve into a “Lifecycle Infrastructure Strategist” or “Sustainable Technology Officer.”
This role will not just manage endings but will orchestrate transitions, balancing performance needs against total lifecycle costs and environmental impact. “The ability to strategically retire infrastructure while maximizing recovery value and minimizing environmental impact will become a differentiator,” Dalli says. As AI costs escalate and hardware cycles compress, this expertise becomes essential.
The work of experts like Dalli highlights a fundamental shift in the tech industry. As companies confront the consequences of unrestrained growth, the discipline of subtraction is no longer a niche operational task but a strategic imperative. Mastering the art of the digital cleanup is becoming essential for any organization aiming for sustainable financial health and long-term innovation.
At Global Business & Economics Journal, we contribute to the future of business and technology with news, innovative research, expert insights, and thought-provoking analysis.
