A decade ago, Google researchers trained a neural network to recognize cats using 16,000 processors. It might not seem like a milestone today – considering that such capabilities are available on any smartphone – but at the time, what Andrew Ng and his team were able to accomplish showed the potential of these networks in image recognition tasks. Yet, still, the promise of AI in transforming business operations remains largely unfulfilled, even with today’s widespread experimentation with AI tools.
This gap between AI’s theoretical potential and practical implementation reveals a critical challenge facing business leaders. Organizations are simply struggling to move from proof-of-concept to production. While early adopters report productivity gains in specific areas – such as automated documentation, analysis frameworks, and process optimization – the broader transformation promised by AI vendors has proven more elusive.
Deloitte AI Institute revealed a striking pattern in its State of Generative AI in the Enterprise report: more than two-thirds of organizations have managed to move less than 30% of their generative AI initiatives from experimentation to full production deployment. Furthermore, according to Gartner Inc., at least 30% of generative AI projects will be abandoned after proof of concept by the end of 2025, on account of poor data quality, inadequate risk controls, escalating costs, or unclear business value.
Here is the main issue: the financial dynamics of AI present a fundamental challenge that distinguishes it from previous technological revolutions. Take the iPhone’s trajectory, for example. Apple’s initial R&D investment has been repaid many times over, with clear returns reflected in both market dominance and shareholder value. The economics of AI follow a markedly different pattern, with today’s AI leaders – especially the larger players – being confronted with unprecedented capital expenditure requirements that demand sustained or, at least, future revenue streams. This is a challenge considering that enterprise adoption requires significant convincing and many consumer applications are available for free.
While AI technology itself is groundbreaking, its success does not depend on capability alone. It’s therefore important to consider the period between 2007 and 2020 and the critical market forces that created the conditions that enabled AI to finally take off:
- The launch of the iPhone and its App Store ecosystem created unprecedented demand for powerful mobile applications, driving the advancement in both hardware capabilities and software development platforms.
- The post-2008 financial crisis environment reshaped the investment landscape. Government stimulus programs and diminished confidence in traditional financial institutions led to a period of low interest rates. This environment pushed private equity and venture capital firms to seek higher returns, directing capital toward high-risk, high-reward areas, either through debt or through equity, particularly in startups, finance products, and research and development.
- The explosion of mobile applications triggered a fundamental shift in cloud computing economics. As startups and small- and mid-sized enterprises pursued innovative solutions to address underserved market needs, they needed scalable and cost-effective computing resources. This increased competition among cloud providers, driving the rapid expansion of cloud capacity and capabilities. This led to a significant decrease in cloud computing costs, which made it more accessible for AI research and development.
The black swan event of the Covid-19 pandemic served as an unexpected accelerant to these trends, driving unprecedented enterprise demand for digital transformation. This is reflected, for example, in Microsoft reaching trillion-dollar valuation status in the 2019-2020 timeframe, aggressive hiring by hyperscalers, expansion by major cloud providers, and the entry of more companies in the trillion-dollar market cap category.
This convergence of factors explains AI’s current trajectory vis-a-vis previous technological revolutions. The development of AI infrastructure has been primarily supply-driven: advancing technical capabilities and declining costs have made AI more accessible and affordable. Juxtapose that to the iPhone (or more generally, smartphones) revolution, which was fundamentally demand-driven, weaving smartphones into our daily lives. Consumer appetite for smartphones was so strong that it forced the infrastructure to leapfrog (HTML5, 5G, e-sim, WIFI, IoT devices, containerization technologies, etc.)
In the latter, consumers barely think before buying the next phone or app. In the former, the value is still to be tapped. This supply-versus-demand dynamic helps explain today’s adoption patterns of AI. Smartphone adoption was natural and intuitive (pull), while AI adoption remains more deliberate and considered (push) – and this current trend explains market volatility, particularly among leading technology stocks.
The AI market is experiencing significant consolidation, with just three to four large companies providing the majority of services – to the concern of regulators and markets. This is reflected in dramatically shifting valuations: Intel’s market value has declined from $115 billion in 2010 to $93 billion today, while Nvidia has surged from $8.4 billion in 2010 to $3 trillion, with two-thirds of this growth occurring in just the past year.
This concentration of market power has prompted regulatory response, particularly in Europe, where new AI regulations impose stringent controls. Yet while Microsoft generates $75 billion in cloud revenue primarily from enterprises in industrialized nations, the World Bank observes that economies that combine strong service exports, English language proficiency, digital skill availability, and younger demographics – notably India, Brazil, and Indonesia – form the lion share of AI usage today. To some, the addition of features to free tiers of AI services also points to a bubble.
AI’s trajectory, though uncertain, is irreversible.
The AI market presents a complex set of seemingly contradictory indicators. Leading cloud providers are committing unprecedented capital expenditure – with two of the three major hyperscalers announcing investments in the hundreds of billions to support future demand. Yet market sentiment remains cautious, as evidenced by recent tech sector volatility, including Nvidia’s historic $279 billion selloff.
This market uncertainty doesn’t add up when considering accelerating innovation: the number of AI patent filings having surged from 8,000 in 2018 to more than 60,000 in 2022. However, enterprise adoption continues to lag, hampered by multiple challenges including regulatory compliance, misinformation risks, talent scarcity, model selection complexity, ROI uncertainty, and data governance concerns.
Historically, truly disruptive technologies have succeeded because they eliminate friction for users and also form a habit, whether that is a new way to make friends, order groceries, or buy insurance. The smartphone revolution exemplifies this: activities that once took up quite a bit of our time were transformed into effortless interactions requiring just a few taps.
AI has yet to achieve this level of seamless integration. Instead of organic adoption, there are many sides to the story, falling more or less into two camps: either heralding AI’s transformative potential or warning of its risks. While the current reality lies between these extremes, one conclusion seems inescapable: AI’s trajectory, though uncertain, is irreversible. The question facing leaders is not whether AI will fundamentally reshape business operations, but when and how this transformation will occur.
© IE Insights.
