Clouds, Machines, and Mirage ROI: Lessons from Cloud Computing for Today’s AI Hype

Thank you for stopping by, please note that all opinions expressed on this site are my own personal opinions and not those of any employer or other organization. This is a personal blog expressing personal opinions

Introduction

History often repeats itself in the world of technology. In the 2010s, cloud computing was heralded as the magic solution: “lower costs, instant scalability, innovation without infrastructure.” Now, artificial intelligence (AI) is the new frontier. Executives are swept up in its promise, often driven by the same blend of fear-of-missing-out (FOMO), organizational inertia, and cognitive biases that cloud computing once ignited.

But these dynamics are deeper than market enthusiasm—they’re rooted in human psychology. Leaders are subject to subtle but powerful pressures to “go with the flow”, avoid dissent, and appear visionary. These pressures manifest through biases like the sunk-cost fallacy and recency bias, which shape decision-making in ways that keep organizations locked into hype-driven strategies.

Counterpoint Acknowledgment:
It’s important to acknowledge that AI and cloud are not identical categories. Cloud was primarily a delivery model shift—from capital expenditure to operational expenditure—while AI represents a capability shift, changing what work is possible at all. The analogy between the two holds mainly in the psychological and organizational patterns of adoption, not in the technologies themselves.

The Sunk-Cost Fallacy & Escalation of Commitment

The sunk-cost fallacy describes the tendency to continue investing in a failing course of action simply because resources have already been spent. Arkes and Blumer (1985) famously observed:

“The sunk cost effect is manifested in a greater tendency to continue an endeavor once an investment in money, effort, or time has been made.”
— Arkes & Blumer, The Psychology of Sunk Cost

Psychologists attribute this to negative affect: the discomfort of admitting loss motivates further spending. One study notes:

“The sunk-cost fallacy is driven by negative affect… people are prepared to take the risk of further losses… because thinking about irretrievable investments causes the elicitation of negative affect.”
— Friedel et al., PMC

The sense of personal responsibility magnifies this effect. Individuals who authorized the initial decision often double down—less from rational analysis, and more to protect their reputation. As summarized in escalation-of-commitment literature:

“Participants who were made to feel personally responsible for a failing decision continued committing more resources than those in low-responsibility conditions.”
— Barry M. Staw, Escalation of Commitment Research

This dynamic is especially salient for executives, who face reputational costs if seen as “backtracking” or “admitting error.”

Recency Bias & Organizational Momentum

While sunk-cost fallacy keeps organizations locked in, recency bias propels them forward. Recency bias describes our tendency to overweight the most recent information or stories:

“Recency bias is a cognitive bias that favors recent events over earlier ones.”
— Wikipedia, Recency Bias

In technology adoption, this means a flashy success story—a peer company touting AI breakthroughs, or a vendor case study—looms larger in the decision-maker’s mind than older cautionary tales. As a result, organizations may rush forward, convinced the most recent example reflects an inevitable trend.

How This Frames the Cloud–AI Parallel

• Cloud Era (2010s): Early adopters invested heavily, reluctant to reverse course when costs mounted. Glowing case studies amplified momentum, discouraging skepticism.
• AI Today: Executives feel both pressure to invest (to avoid being left behind) and pressure not to retreat once resources are spent. Each new “AI success” story reinforces urgency—feeding hype cycles even when outcomes remain uncertain.

By grounding the discussion in psychology, it becomes clear that the parallels between cloud and AI aren’t just technological—they’re human. The same biases that once drove cloud overselling now set the stage for AI’s inflated promise.

The Hype Machine

Technological adoption isn’t just a technological narrative; it’s also a psychological story. At the heart of major tech waves—first cloud, now AI—is a sophisticated hype machine where marketing, capitalism, and cognitive bias intertwine to propel forward narratives that often exceed substance.

1. Hype Psychology: Expectation Amplified

Hype culture—the collective enthusiasm around emerging tech—is characterized by exaggerated expectations fueled by scarcity, novelty, and selective storytelling:

“Hype culture … is characterized by an attitude of excessive and positive expectations that consumers attach to products, services or technological advancements which have yet to be released.”
— Laura Waldman, Verizon’s Uppernet Accurately Defines Cloud Computing

This cultural fever accelerates progress: marketers exploit recency bias, where the most salient new success stories overshadow past warnings or failures. Combined with social proof and fear of missing out, hype acts as a gravitational pull—hard to resist and easily amplified.

2. The Hype Cycle: Boom, Bust, and Balance

Gartner's Hype Cycle maps the arc of expectation over time:

1. Innovation Trigger
2. Peak of Inflated Expectations
3. Trough of Disillusionment
4. Slope of Enlightenment
5. Plateau of Productivity

Both cloud and AI have experienced this arc. Cloud’s early hype eventually led to cost revelations and failures in naive “lift-and-shift” projects. AI, meanwhile, now hovers near the top—full of bold promises but uneven delivery.

3. Capitalism’s Role: Marketing as Momentum

Hype aligns tightly with capitalist incentive structures. Products gain value not simply through utility, but through perceived potential—a hallmark of hype marketing:

“Hype allows brands to promote their image above the actual quality of the product.”
— Wikipedia, Hype Marketing

This phenomenon mirrors conspicuous consumption, where purchasing or adopting what’s perceived as cutting edge signals status—often regardless of actual performance.

4. Verizon’s Uppernet Campaign: Hype in Action

A telling early example of cloud hype comes from an ad campaign for Verizon’s “Uppernet”. While this term lacks formal corporate backing, the ad effectively framed cloud infrastructure as robust, reliable, and almost mythic in its solidity:

“Our cloud is not soft and fluffy, our cloud is made of bedrock, concrete and steel. Our cloud is the smartest brains combating the latest security threats…and is scalable as far as the mind can see.”
— Verizon Uppernet Ad Transcript

This messaging welds abstract innovation to emotion—projecting strength and clarity onto something inherently intangible.

What’s instructive is how the Uppernet concept faded quickly—much like many cloud marketing narratives. Despite the dramatic message, the promise failed to hold, illustrating how marketing momentum can decay faster than infrastructure.

5. Bias and the Hype Machine

At its core, hype is powered by human cognitive tendencies:

When evangelists—vendors, media, peers—synchronize on hype, leaders respond not only to technology, but to collective momentum.

Counterpoint Acknowledgment:
Some would contend that hype isn’t distortion at all, but fuel. Without hype, risk-averse organizations may never take bold bets. Even if nine out of ten projects fail, the one success can pay back the portfolio many times over. In this sense, hype serves a catalytic role—accelerating adoption and forcing cultural change. The key is not to eliminate hype, but to channel it responsibly.

Summary

The “Hype Machine” isn’t accidental—it’s a finely tuned mechanism where psychology, marketing, and capitalism converge. Verizon’s Uppernet campaign exemplified this: a bold, concrete branding of the cloud that soared on narrative—but quickly dissipated. Both cloud and AI ride this same wave: elevated expectations, fleeting narratives, and a powerful blend of human optimism and marketing craft.

The Sunken Cost Fallacy in Technology Adoption

Once organizations make large investments—whether in infrastructure, tools, or pilot projects—they often keep going even when the results are underwhelming. This is the sunk-cost fallacy in action: the tendency to continue investing resources because of what has already been spent, rather than because of future returns.

Counterpoint Acknowledgment:
It’s worth noting that high failure rates don’t always signal wasted effort. Innovation funnels in R&D often have attrition rates just as high—most projects fail, but the few that succeed can transform entire industries. In this sense, today’s AI pilots may be less a sign of hype and more a necessary process of experimentation. Still, leaders should remain cautious that persistence is justified by learning value, not simply sunk costs.

For business leaders, abandoning a project may feel like admitting defeat—so continuation becomes a reputational safeguard, even if irrational.

How It Showed Up in Cloud Computing

During the early days of cloud migration, many companies executed “lift-and-shift” projects—moving legacy monolithic applications into the cloud without redesign. Instead of saving costs, they often increased them. But reversing course was politically difficult:

• IT leaders who had championed the migration feared reputational damage if they admitted failure.
• CFOs rationalized higher bills as “temporary,” expecting long-term savings that never materialized.
• Vendors reinforced persistence, promising future efficiencies to justify sunk costs.

The result: many organizations doubled down on poorly optimized cloud footprints.

How It’s Emerging in AI

Today, the same dynamic plays out in AI adoption:

• Pilot projects: Enterprises spend millions on proof-of-concepts that don’t scale, yet hesitate to shut them down.
• Private LLMs: Some firms build expensive in-house models, only to find maintenance costs unsustainable.
• Licensing contracts: Multi-year AI vendor agreements lock organizations in, making withdrawal politically and financially unappealing.

A recent MIT study highlights the danger: 95% of enterprise AI pilots fail to move beyond incubation, yet organizations continue to pour resources into them. (Windows Central).

Gartner forecasts that 30% of generative AI projects will be abandoned after proof of concept by the end of 2025—suggesting many leaders will eventually face this reckoning. (AI Magazine).

The Human Factor: Escalation and “Face-Saving”

Executives are not just making financial calculations—they’re navigating social and reputational dynamics:

• Face-saving pressures: Admitting a failed investment can threaten a leader’s credibility.
• Groupthink: Teams reinforce the narrative that persistence is rational, especially if dissent is unwelcome.
• Loss aversion: Psychologically, humans feel losses about twice as strongly as equivalent gains, making withdrawal harder.
• Public commitments: Once an initiative has been announced to stakeholders or markets, backing down becomes even more difficult.

Why This Matters for AI Today

The danger is not only wasted money—it’s opportunity cost. Resources tied up in failing AI projects can’t be redirected to high-value initiatives. Worse, organizational credibility suffers if leaders are seen chasing hype without accountability.

By recognizing sunk-cost bias early, leaders can design exit strategies, go/no-go checkpoints, and independent reviews—structural safeguards against escalation of commitment.

The Illusion of Competitive Urgency

If the sunk-cost fallacy explains why organizations struggle to stop failing projects, competitive urgency explains why they start them so quickly in the first place. Both cloud computing in the 2010s and AI today were adopted under a narrative of “adopt now, or be left behind.” This fear of missing out (FOMO) is more than cultural shorthand—it’s a powerful psychological and economic force.

FOMO as Strategy

Psychologists have defined FOMO as:

“A pervasive apprehension that others might be having rewarding experiences from which one is absent.”
— Przybylski et al., Motivational, Emotional, and Behavioral Correlates of Fear of Missing Out (2013)

In consumer behavior, FOMO drives compulsive adoption of trends. In enterprise contexts, it manifests as reactive strategy-making—leaders rush to approve cloud or AI initiatives simply because competitors are perceived to be doing so.

Counterpoint Acknowledgment:
While this article highlights biases that accelerate adoption, the opposite forces exist as well. Status quo bias and risk aversion often make organizations slow to adopt, particularly in regulated or conservative sectors. The reality is uneven: some firms rush in too fast, others hesitate too long. The challenge is finding a middle path between reckless urgency and paralyzing caution.

Cloud Computing: “Migrate or Die”

In the 2010s, cloud marketing played heavily on urgency. Vendors framed migration as the dividing line between modern, agile businesses and obsolete “dinosaurs.” Media reinforced this narrative with case studies of Netflix, Airbnb, and Uber—companies whose digital success stories were held up as proof that everyone needed to “move fast” into the cloud.

The fear wasn’t entirely irrational: early movers did gain efficiencies. But for many organizations, the rush to cloud came without due diligence, leading to ballooning costs and security lapses.

AI Today: “No AI Strategy = No Future”

The same urgency is now tied to AI. Gartner notes that after the initial hype of 2023–2024, executives are impatient for returns:

“After last year’s hype, executives are impatient to see returns on GenAI investments … organizations are struggling to prove and realize value.”
— Gartner via The Journal

Meanwhile, the MIT NANDA Initiative found 95% of enterprise AI pilots fail to move past incubation—but this hasn’t slowed investment. Instead, competitive pressure accelerates it, creating a paradox: leaders fear missing out even as failure rates are widely publicized.

Recency Bias: Why Urgency Feels Real

Competitive urgency is amplified by recency bias—our tendency to overweight the most recent information or examples. When an executive reads a glowing AI case study in Harvard Business Review or hears a peer boast about deploying generative AI, that story looms larger than older lessons of failed cloud migrations or vendor lock-ins.

As one behavioral analysis put it:

“Recency bias is a cognitive bias that favors recent events over earlier ones.”
— Wikipedia, Recency Bias

This bias explains why each new AI success story can reset executive enthusiasm, even when failure rates remain sky-high.

Capitalism, Markets, and the Speed Imperative

Urgency isn’t just psychological—it’s embedded in capitalism itself. Public companies are pressured by shareholders to demonstrate innovation, and no leader wants to explain why they’re “late” to the next big thing. Marketing amplifies this by presenting adoption not as a choice, but as a survival strategy.

Cloud was sold with slogans about agility; AI is sold with phrases like “future-proof your business.” The subtext is the same: hesitation equals obsolescence.

The Human Cost of FOMO-Driven Adoption

The illusion of urgency often results in:

• Reactive adoption — projects greenlit based on external pressure rather than internal readiness.
• Misaligned priorities — teams pushed to “do AI” or “get to cloud” even without clear ROI.
• Overcommitment — investments made quickly, then rationalized later, feeding into the sunk-cost trap.

Together, FOMO and sunk-cost bias create a cycle: rush in quickly, stay in too long. This pattern defined much of the cloud hype era, and it now risks defining AI adoption as well.

Overselling to Executives

If hype sets the stage and FOMO pushes leaders to act, vendor overselling often delivers the final push. Both cloud and AI have been marketed to executives not simply on technical merit, but through visionary narratives designed to resonate with business ambitions. This approach is effective because it aligns with deep psychological tendencies in leadership decision-making.

The Sales Psychology of Overselling

Sales and marketing don’t just present technology—they present stories. These stories often draw on selective case studies: Netflix on AWS, Airbnb scaling globally, or a Fortune 500 firm cutting costs with AI automation. These narratives are framed to demonstrate inevitability: if these leaders succeeded, others must follow.

Behavioral research explains why this works. Executives are especially vulnerable to social proof—the tendency to model decisions on perceived peers:

“Social proof is a psychological phenomenon where people copy the actions of others in an attempt to undertake behavior in a given situation.”
— Cialdini, Influence: Science and Practice

Counterpoint Acknowledgment:
To be fair, many executives are not naïve victims of sales psychology. They use portfolio management, ROI tracking, and staged investments to filter vendor claims. Frameworks such as Gartner’s AI TRiSM (Trust, Risk, and Security Management) show that leaders are already working to put guardrails in place. Still, even sophisticated governance must contend with subtle biases and market pressure.

In the boardroom, this translates to executives approving cloud or AI projects not solely on technical evaluation, but because “our competitors are doing it.”

Cloud Era: Case Studies as Pressure

Cloud vendors aggressively used customer success stories to push migration. Amazon Web Services highlighted Netflix as the archetype of cloud transformation. The message was clear: if cloud could scale video for millions, it could transform any enterprise workload.

But these examples were not representative. Netflix had a specific architecture and business model that benefited uniquely from cloud elasticity. Many enterprises, attempting lift-and-shift migrations, found costs increased rather than decreased. Still, executives bought in—because the marketing narrative was compelling.

Counterpoint Acknowledgment:
Case studies like Netflix on AWS or ChatGPT in enterprise use cases are not inherently misleading—they are valid demonstrations of what is possible. The danger lies in treating them as universally applicable. The challenge for executives is not to dismiss these stories, but to contextualize them: to ask “Is our environment comparable?” and “What assumptions make this case work?”

AI Era: The New Playbook

AI marketing follows the same pattern. Vendors highlight glamorous applications:

• Chatbots that “never sleep” in customer service.
• Fraud detection algorithms touted as “instant and infallible.”
• Generative design tools that promise to “cut R&D cycles in half.”

Yet failure rates tell another story. As MIT reported, 95% of AI pilots never move beyond incubation (Windows Central). Gartner projects that 30% of generative AI projects will be abandoned by 2025 (AI Magazine).

The gap between polished case studies and real-world outcomes is wide—but the presentation of inevitability creates a compelling sales hook.

Why Executives Buy the Story

Executives face unique pressures and biases:

• Information asymmetry: Many leaders don’t have the technical depth to evaluate claims directly.
• Vision signaling: Approving AI or cloud projects signals to boards and shareholders that they are forward-looking.
• Authority bias: Vendor representatives often present themselves as trusted experts, and executives are inclined to defer.
• Loss aversion: Saying “no” risks falling behind peers—so the safer option is to say “yes.”

As Cialdini notes in persuasion research, urgency combined with social proof is one of the most effective influence tactics. Vendors wield both liberally.

The Cycle of Overselling

Overselling contributes to a repeating cycle:

1. Visionary narrative presented to executives.
2. Case studies used to demonstrate inevitability.
3. Approval under pressure (fear of missing out, desire to appear innovative).
4. Projects underdeliver, but sunk-cost bias keeps them alive.

The outcome is not just wasted investment, but strategic distraction. Organizations may pursue vendor-defined futures instead of aligning technology adoption with their own business needs.

Summary

Cloud vendors oversold agility and cost savings with selective examples like Netflix. AI vendors now oversell automation and intelligence with equally narrow case studies. Executives, subject to social proof, authority bias, and reputational pressure, often buy into these narratives. The result is a cycle of overcommitment: grand promises, underwhelming outcomes, and difficulty backing out.

Lessons from Cloud for Today’s AI

If cloud adoption taught us anything, it’s that hype eventually collides with reality. After the early rush, organizations learned that migration wasn’t a magic bullet—it required re-architecting applications, managing costs, and setting clear business objectives. In short, cloud computing settled into a more pragmatic phase where it became useful, but not utopian.

AI today stands at a similar inflection point. The promises are vast, but the risks of overinvestment, poor alignment, and wasted effort are equally real. By looking back at the cloud hype cycle, leaders can extract practical lessons for navigating AI adoption responsibly. It should be noted that Cloud did eventually deliver massive value once adoption matured. The lesson isn’t cynicism—it’s patience and discipline.

Counterpoint Acknowledgment:
Critics might argue that judging AI on today’s pilot failures is premature. Cloud, too, looked inefficient at first—costs spiked before practices like workload tagging and optimization matured. Over time, the long-term ROI became clear. The same may prove true for AI once data pipelines, governance, and talent mature. Acknowledging this, leaders should balance patience for maturation with discipline in managing expectations.

1. Clarify Objectives

Cloud migrations went wrong when they were pursued as an end in themselves. Companies moved applications simply because “everything must be in the cloud,” without asking whether migration would add business value.

The same trap now threatens AI. A survey of enterprise pilots found that 95% fail to deliver measurable returns (Windows Central). Many of these projects lacked clear objectives beyond “we need an AI strategy.”

Lesson: Define specific goals upfront. For example, “reduce customer service handle time by 15%” is actionable; “implement AI in customer service” is not.

2. Quantify ROI

In the cloud era, vague expectations of “cost savings” often turned into higher monthly bills. Organizations only began to succeed once they implemented cost monitoring, workload tagging, and optimization practices.

AI requires similar discipline. Gartner warns that 30% of generative AI projects will be abandoned by 2025, largely because they fail to demonstrate value (AI Magazine).

Lesson: Define metrics of success before investing. ROI can be expressed in time saved, error rates reduced, or revenue uplift—not just “innovation potential.”

3. Plan an Exit Strategy

A hallmark of the sunk-cost fallacy is that leaders hesitate to shut down projects once resources have been committed. This dynamic plagued early cloud migrations and is already visible in AI, where enterprises persist with failing pilots because executives fear reputational damage from admitting failure.

Lesson: Establish go/no-go checkpoints and predefine exit criteria. Treat pilots as experiments that can be discontinued without stigma if results don’t justify scaling.

4. Adopt Iteratively

The most successful cloud adopters didn’t migrate everything at once—they started small, tested workloads, and scaled gradually. Similarly, AI adoption should be iterative: begin with narrow use cases, measure results, and expand only when benefits are proven.

Princeton computer scientist Arvind Narayanan, co-author of AI Snake Oil, makes this caution explicit:

“AI snake oil … is AI that doesn’t work as advertised and probably can’t work as advertised.”
— Narayanan & Kapoor, AI Snake Oil

The antidote to snake oil is disciplined, evidence-driven adoption.

5. Align Technology with Business, Not Hype

Ultimately, the greatest lesson from cloud is that technology adoption must serve strategy—not the other way around. Cloud became valuable when companies stopped chasing hype and started aligning migration with business priorities like scalability, agility, and customer experience.

AI will follow the same trajectory—but only if leaders resist the pull of hype and ground decisions in evidence.

Summary

Cloud adoption eventually stabilized when organizations moved past hype and focused on clear objectives, measurable ROI, exit strategies, and iterative scaling. AI is at the same turning point today. Leaders who learn from cloud’s growing pains can avoid repeating the same cycle of overinvestment and disappointment—and instead realize AI’s true potential.

Conclusion

History shows us that new technologies rarely fail because of their underlying potential—they stumble when human psychology, marketing narratives, and competitive pressure collide. Cloud computing promised limitless scalability and lower costs, yet many organizations discovered higher bills and hidden complexity. Artificial intelligence now carries the same weight of expectation, and early evidence suggests it may be following a parallel path. This analogy is imperfect, but the psychological and organizational dynamics rhyme even if the technologies differ.

The underlying forces are strikingly similar:

• The Hype Machine: Vendors and media amplify success stories, while failures are quietly buried.
• The Sunk-Cost Trap: Once projects are funded, leaders find it difficult to stop—even when outcomes disappoint.
• The Illusion of Urgency: FOMO convinces executives that hesitation means obsolescence.
• Overselling to Executives: Case studies and visionary pitches create an aura of inevitability.

These aren’t simply business mistakes—they’re reflections of well-documented biases: sunk-cost fallacy, recency bias, social proof, and authority bias. Together, they form the psychological scaffolding that sustains hype.

But cloud also taught us an important lesson: hype eventually gives way to reality. After the initial rush and the trough of disillusionment, cloud adoption matured into a rational toolset. Organizations that succeeded did so by aligning cloud adoption with clear objectives, measurable ROI, exit strategies, and iterative scaling. AI can—and must—follow the same trajectory.

The challenge for today’s leaders is to resist the gravitational pull of hype. That means asking hard questions before approving AI projects, demanding evidence of value, and normalizing exit strategies for pilots that don’t deliver. It also means remembering that adopting AI is not a strategy in itself—it’s a tool that must be aligned with broader goals. The risk is history repeating itself—investing heavily in technologies that underdeliver. The opportunity is to resist that cycle and build evidence-driven innovation.

In the end, the organizations that thrive in the AI era will not be those that moved first or spent most, but those that moved wisely. By learning from the cloud’s hype cycle, today’s leaders can avoid chasing mirages of ROI and instead build durable, evidence-driven innovation.