
Summary
Uber Chief Technology Officer Praveen Neppalli Naga says corporate sentiment around AI spending is shifting, with the era of tokenmaxxing ending as companies demand clearer evidence of business value from their AI investments.
From Euphoria to Pragmatism: A New Phase in Enterprise AI Investment
After nearly two years of generative AI euphoria, corporate attitudes toward AI investment are undergoing a subtle but significant transformation. Uber Chief Technology Officer Praveen Neppalli Naga recently stated publicly that the era of enterprise "tokenmaxxing" is coming to an end—a declaration that marks the tech industry's shift from blindly pursuing usage volume to rationally assessing return on investment.
Tokenmaxxing refers to the practice of continuously increasing AI spending, measured in tokens consumed by models, under the assumption that more usage automatically translates to higher value creation. This approach was extremely common during the AI boom from 2023 to early 2024, with many companies granting employees open access to AI tools and encouraging teams to use them as much as possible, hoping that high usage rates would drive innovation and productivity gains.
Uber's Costly Lesson: A Warning About Budget Overruns
Uber's own experience provides a vivid illustration of this shift. The company burned through its entire Claude Code budget—originally planned to last until 2026—by April 2024. This eye-catching fact clearly demonstrates how quickly costs can balloon when teams are given free rein to use AI tools without constraints.
Uber President Andrew Macdonald raised sharp questions about this phenomenon earlier this year. He noted that while AI usage statistics were "mind-blowing," the link between higher AI spending and successfully shipping product features "simply isn't there." In other words, massive AI usage has not translated into meaningful product improvements that customers actually use.
This finding prompted Uber's leadership to fundamentally reassess its AI investment strategy. Rather than pursuing maximum usage, the company is adopting a more disciplined approach, continuing to work with major model providers but demanding clearer evidence of value creation.
Industry Trend: From Free-for-All to Budget Discipline
Uber's pivot is far from isolated. Other tech companies like Atlassian have also begun implementing AI budgets for their engineers, signaling that the industry is transitioning from a "free-for-all" honeymoon phase to a more pragmatic "cost control" stage. As the costs of tokenmaxxing become apparent, spreadsheets are replacing laissez-faire attitudes.
Broader research evidence supports these skeptics' concerns. Multiple studies have found that most enterprise AI spending never progresses beyond the pilot stage, producing demos and proofs of concept rather than shippable products. This phenomenon is widespread across industries, exposing a massive gap between AI experimentation and production deployment.
The Real Cost of Tokenmaxxing
The financial implications of unconstrained AI usage are becoming increasingly clear. When Uber exhausted a two-year budget in just four months, it wasn't merely a budgeting error—it represented a fundamental miscalculation about how AI tools would be used in practice.
The problem stems from the disconnect between AI adoption metrics and actual business outcomes. High usage numbers can create an illusion of success, masking the reality that much of this activity may not be generating meaningful value. Teams may be using AI tools extensively for tasks that could be accomplished more efficiently through traditional methods, or for experimental projects that never materialize into production features.
This pattern mirrors challenges seen in other technology adoption cycles. During the early days of cloud computing, many organizations experienced similar cost overruns as teams spun up resources without clear governance or cost accountability. The lesson then, as now, is that powerful new technologies require thoughtful management frameworks, not just open access.
Implications for the Digital Asset Industry
While Uber's discussion focuses primarily on traditional enterprise AI applications, this shift has important implications for the digital asset and blockchain industry. In cryptocurrency and Web3, many projects have experienced similar "maximization" mindsets—whether pursuing the highest transaction volumes, the most smart contract deployments, or the largest token circulation.
Digital asset infrastructure providers need to learn from this trend. For wallet service providers, custody platforms, and trading infrastructure companies, pursuing growth in transaction volume or user numbers while ignoring actual business value and sustainability can lead to severe resource waste.
More importantly, as institutional investors increasingly enter the digital asset space, they will demand stricter cost-effectiveness and return on investment. Service providers that can demonstrate clear value propositions, measurable business outcomes, and reasonable cost structures will gain competitive advantages in this new phase.
The parallel is particularly relevant for institutional-grade digital asset services. Just as enterprises are demanding clearer ROI from AI investments, institutional clients evaluating custody solutions, wallet infrastructure, and trading platforms are increasingly sophisticated in their cost-benefit analyses. Solutions that can demonstrate not just technical capabilities but also clear economic value will be better positioned for long-term success.
The Path Forward: Rational and Value-Oriented AI Investment
The shift at Uber and other tech giants signals that enterprise AI investment is entering a more mature phase. This doesn't mean the end of AI investment, but rather a transition from blind optimism to rational assessment.
Future AI investment will increasingly focus on:
Clear Business Cases: Companies will require AI projects to demonstrate clear value propositions and expected returns before launch. The days of funding AI initiatives based solely on technological promise are giving way to demands for concrete business justification.
Measurable Outcomes: Usage volume will no longer be the sole success metric. Actual business impact, productivity improvements, and product innovation will become key evaluation criteria. Companies will develop more sophisticated frameworks for assessing AI project success that go beyond simple adoption statistics.
Cost Discipline: Organizations will implement stricter budget controls and cost monitoring mechanisms to ensure AI spending aligns with business value. This includes not just setting budgets, but actively tracking spending against outcomes and adjusting allocations based on demonstrated results.
Bridging Pilot to Production: Companies will focus more intently on how to transform successful AI pilots into scalable production applications. The "pilot purgatory" problem—where promising experiments never make it to production—will receive increased attention and resources.
Broader Industry Implications
This transformation has far-reaching implications for the entire tech industry. AI model providers will need to adjust their business models from simple usage-based billing to pricing strategies that emphasize value delivery. This might include outcome-based pricing, tiered service levels based on business impact, or hybrid models that balance usage with value creation.
Enterprise customers, meanwhile, need to establish more robust AI governance frameworks that balance innovation encouragement with cost control. This includes developing clear policies about when and how AI tools should be used, establishing approval processes for new AI initiatives, and creating feedback loops that capture learnings from both successful and unsuccessful projects.
For the digital asset industry, this trend serves as a reminder that sustainable development—whether in traditional AI applications or blockchain technology—must be built on clear value creation and rational resource allocation. Technologies and services that can demonstrate genuine business value will be the ones that succeed in long-term competition.
The end of tokenmaxxing doesn't signal pessimism about AI's potential. Rather, it represents the industry's maturation from experimental enthusiasm to practical implementation. Companies that can navigate this transition—maintaining innovation while demonstrating clear value—will be best positioned to capitalize on AI's transformative potential in the years ahead.
Source: link