Artificial Intelligence: Buzz Word, Investment Bubble or Technological Breakthrough?

In the archives of technological advancement, we find ourselves at a familiar juncture - the dawn of a new era propelled by artificial intelligence (AI). As we stand on the precipice of what many herald as the next great leap forward, it's crucial to examine this moment through the lens of historical technological advancements, drawing parallels that may illuminate the path ahead. And answer the question, whether SME should invest in AI or rather wait for the hype to pass?

Similarities to the Internet-Boom

The current AI investment frenzy bears striking similarities to the telecom bubble of the late 1990s and early 2000s. During that period, massive capital was poured into building out the infrastructure for the internet, a technology that promised to revolutionize communication and commerce. The World Wide Web was invented in 1989, and by the year 2000, there were over 400 million users worldwide. Today, we see a similar pattern with AI, as companies race to build the computational infrastructure needed to power the next generation of intelligent systems.

Key parallels include:

• Infrastructure-heavy investment: Just as telecom companies laid miles of fiber optic cables, today's tech giants are investing heavily in data centers and specialized hardware like GPUs.
• Narrow initial beneficiaries: The telecom boom initially benefited equipment makers and infrastructure providers. Companies like Cisco and Nortel saw their stock prices soar during this period. Similarly, the current AI boom has disproportionately benefited semiconductor companies (particularly NVIDIA) and the "Magnificent Seven" tech stocks.
• Promise of transformative impact: Both eras are/were driven by the promise of revolutionary change across industries. The internet ultimately transformed how we communicate, shop, and access information, while AI has the potential to disrupt everything from healthcare to transportation.

As Hilbert (2020) notes, "The world was swamped with internet connections and mobile phones in record time. The result was the resolution of space-time constraints in global communication and the accumulation of vast amounts of stored data, which has more recently been termed 'big data.'" This infrastructure build-out laid the foundation for the current AI revolution.

The Arms Race for Computing Power

The past 18 months have seen a remarkably narrow investment cycle focused on AI infrastructure, characterized by an arms race for GPU capacity. This has led to:

• Unprecedented demand for GPUs, with NVIDIA being the primary beneficiary, making it one of the most valuable companies in the world.
• Major capital expenditure commitments from hyperscalers like Google, Microsoft, and Meta. Google alone is expected to increase Capex by $18 billion in 2024, with the majority going to AI-related investments.
• A stark divergence between infrastructure investment and realized applications / monetization.

This intense focus on building AI infrastructure and acquiring compute power echoes the massive investments in fiber optic networks during the telecom boom. Just as those investments laid the groundwork for the internet age, today's AI infrastructure investments are setting the stage for the next wave of technological innovation.

Recent earnings reports from major tech companies have shown a significant uptick in capital expenditure forecasts for 2024. Google CEO Sundar Pichai and Meta CEO Mark Zuckerberg have both publicly stated that the risk of underinvesting in AI outweighs the risk of overinvesting:

“[…] not investing to be at the front here I think definitely has much more significant downsides.” - Alphabet CEO, Sundar Pichai

DGKI urges SMEs to think about investing in AI application as well as education of their staff early on as well: the return on those investments may only be realized in the future, and the cost may appear high as of today, however, not investing could mean ever diminishing competitiveness going forward. Which would be far more costly, if not fatal.

The Gap between Investment and Realization

Historical technology cycles teach us that there's often a significant lag between initial investment and tangible benefits. This gap, sometimes referred to as the "trough of disillusionment" in Gartner's Hype Cycle, could span several years and multiple budget cycles.

For AI, we're likely looking at late 2025 at the earliest before we see widespread monetization of current investments in application software, with the real inflection likely occurring in 2026/2027. This timeline aligns with historical patterns:

• The internet infrastructure boom of the late 1990s didn't yield widespread commercial benefits until the mid-2000s. Companies like Amazon and Google, founded in the 1990s, didn't become profitable until several years later.
• The mobile revolution, kickstarted by the iPhone in 2007, took several years to mature into the app economy we know today.

When budgeting, every owner and managing director of a SME should ask themselves today, if looking back, they want to be the one responsible not investing in the use of internet or the mobile revolution.

The Scale and Data Dynamics

As AI models grow increasingly complex and computationally demanding, companies with the infrastructure to support large-scale training and deployment will hold a significant edge. The current AI arms race centers around compute power, where hyperscalers like Amazon, Meta, Google, and Microsoft, with their vast infrastructure, have a distinct advantage. These companies can not only train and deploy larger, more sophisticated models, but also offer them at a lower cost per token. This ability to deliver "more for less" will be a key differentiator in attracting customers and driving AI adoption.

However, compute alone isn't sufficient. The quality and quantity of data used to train models are equally crucial. As the adage goes, "garbage in, garbage out." Companies with access to vast amounts of high-quality data will have a distinct advantage in developing more accurate and capable AI models.

For example, Google, with its dominance in search, consumer applications, and Android, sits on a treasure trove of data that can be leveraged to train its AI models. Also other Big Tech companies have access to valuable data. These advantages will be critical in shaping the competitive landscape. Who will win the race, temporarily or permanently, is anybodies guess at this point.

SMEs should therefore stay as flexible as possible with their service providers. That is to be cloud, AI (LLM) and ERP agnostic.

Data Privacy and Security Concerns

The recent data breach at Snowflake underscores the growing concerns around data privacy and security. As AI becomes more deeply integrated into enterprise workflows, organizations will be increasingly reluctant to share sensitive data with third parties. This could lead to increased investment in on-premises AI infrastructure and hybrid solutions that allow companies to leverage the power of cloud-based AI while maintaining control over their data.

German SMEs should ensure to work with AI providers that understand the importance or protecting company data and complying with the GDPR as well as the EU AI Act.

The Path Forward

As we move forward, a key battleground is emerging between open-source and proprietary AI models:

• Meta, under Mark Zuckerberg's vision, is leaning heavily into open source, leveraging the network effects they've mastered in social media and monetizing later via advertising. Meta released its large language model, LLaMA, under a noncommercial license in February 2023.
• Google is betting on its vast infrastructure and data advantage to gain market share and generate quicker returns on its AI investments via its Cloud business.

This dichotomy mirrors earlier tech battles, such as the open Android ecosystem versus Apple's closed iOS environment.

As mentioned previously, SMEs must make it priority to remain flexible and not become dependent on their technology service providers.

Conclusion

As investors and industry observers, it's crucial to maintain perspective:

• While the potential of AI is immense, realizing its full benefits will take time. • The current infrastructure-heavy investment phase will eventually give way to a period of application development and monetization.
• Not all players will survive the transition -- just as many telecom and dotcom companies didn't survive their respective bubbles. Will Perplexity.AI exist 5 years from now or Google?
• The true winners may not be apparent for several years, and they may not be the same companies leading the current infrastructure boom.

DGKI is actively searching for winners and losers within this framework, and will integrate (and disintegrate) technology into their products, and deliver state-of the-art products to SMEs

This Blog article was written by Gemini Advanced and Claude 3.5 Sonnet and edited by DGKI.