The AI Hype Cycle: Financial Markets, Innovation Gaps, and the Challenge of Adoption

Are we witnessing the burst of the AI hype bubble? Is AI merely a tool for drafting emails and generating cute cat pictures, or is there more substance beneath the surface?

To analyze whether we are witnessing the end of the AI hype, I offer three distinct perspectives.

The financial view

NVIDIA, the poster child of the AI boom, saw its share price sharply decline by over 30% in less than a month. Sequoia Capital posts: "AI’s $600B Question" and states 'The AI bubble is reaching a tipping point' and asks 'Where is all the revenue?' and Forbes adds “Is The AI Bubble About To Burst?”?

The most frequent comparisons were related to the dotcom bubble and that history is repeating.?

The innovation perspective

OpenAI’s history feature remains unavailable in the EU, and ChatGPT's multimodal capabilities have yet to be widely released. Anthropic and Google are bringing out their updated models but how many of you are using it? And how impressive is it really that text-to-video tools can now create 2-minute-long videos? And even if you think that Sora is impressive - who of you has access to it??

The adoption question

On an organizational level, questions about the concrete business impact are also getting louder. Are AI systems delivering enough value compared to the required investments? Many companies are trying to find and assess the most relevant use cases but seem far from scaling. Recent EU legislation will most likely slow down the deployment of AI for many multinationals even further.

So where do we stand?

Looking at the famous?Gartner hype cycle, many analysts say that we are now moving from the peak of inflated expectations to the trough of disillusionment. (reminder: There are 5 phases:?1. Innovation Trigger 2. Peak of Inflated Expectations 3. Trough of Disillusionment 4. Slope of Enlightenment 5. Plateau of Productivity)

Surely, some predictions were more driven by hopes and excitement and others were - even though directionally right - too bullish on their time horizon. The most famous one is probably Geoffrey Hintons prediction on the disappearance of radiologists by 2021.?

So, are we in a phase now that helps us to look at AI more sober, disillusioned after the big hype??

Let’s look at the three areas above a bit closer:

The financial view

Without providing full financial assessments or even giving financial advice, there are a few key differences between the dotcom bubble and the current AI story.

In the late 1990s a lot of companies lacked viable business models and were not generating sufficient revenue or even profit to justify their valuations. Today a large chunk of the billions of increased valuations belong to the “magnificent seven” (Apple, Microsoft, Amazon, Alphabet, NVIDIA, Tesla, Meta). They clearly generate a lot of revenue and profits. Admittedly not all of it is related to their AI activities but there is much more substance compared to many of the dotcom hopes. To be balanced, it is also worth noting that even Microsoft’s share price dropped significantly from Dec 1999 to Dec 2000 (market cap went down 60%). Those who bought Microsoft at its peak in Dec 1999 needed to wait more than 15 years to balance out their portfolio. Frankly speaking, even for me being still very bullish on AI, this was a quite sobering reminder during my research for this article.?

The innovation perspective

Yes, those extreme tech optimists who predicted AGI (Artificial General Intelligence) in the next 2-3 years (with Elon Musk saying AGI could emerge as soon as 2025) might be a bit ahead of the reality (depending on the definition of AGI).?

The speed of improvements remains mind-blowing, and the perception of a slowdown might be driven more by inflated expectations and maybe even by a deliberate holding back of certain features before the US election (see risk of deepfakes etc).?

At the same time, the progress in many areas are still fascinating:

• Text to video (Sora, Luma AI, Kling AI),?
• OpenAIs demoed multimodal capabilities including live translations and learning use cases or?
• the recent silver medal that an AI tool won at the International Math Olympics by Google DeepMind's AI models, AlphaProof and Alphageometry 2 . These AI systems tackled four out of six complex math problems, scoring 28 out of 42 points, just one point below the gold medal threshold. The achievement highlights significant progress in AI's ability to solve advanced mathematical problems, combining neural networks and symbolic AI to handle tasks that challenge even top human mathematicians.

This shows how AI models are expanding into new areas.?

The adoption question

This brings us back to my favorite topic which I want to spend more thoughts on: organizational adoption.?

The challenge lies not in technological progress but in scaling use cases and experiments. Many large organizations have become slow in truly transforming. Transformations need to go beyond just changing boxes in org charts. Additionally, many companies are still in the midst of their digital transformation, focusing on critical tasks like upgrading to cloud-based infrastructure and enhancing data integration and quality.

Currently, AI use cases often appear like a brain in a jar. Organizations are trying to utilize tools like LLMs in a very isolated way and didn’t find a way yet to embed them into a complete ecosystem.?

“As with other digital transformations, merely automating a process or porting it to a new, more efficient technology doesn’t necessarily deliver enough upside to justify the investment. Real value comes from learning how the new technology changes not just how something gets done but what gets done.” (Bain & Company 2024 )

So where do we stand with the technology adoption for AI? There is a very helpful illustration from McKinsey looking at 5 different stages:

McKinsey sees Gen AI and Applied AI already in phase 4 - I don’t… With 26% of companies (self-reported) being at this stage and 10% fully scaled, it also means that 64% are at "piloting" or lower. I think the graphic really nicely illustrates where the challenge is: moving from piloting to scaling. And that’s also what I get out of the various conversations in my network.

Interestingly, employees are often ahead of their organizations in utilizing AI for boosting their own efficiency and output quality. “Despite all the buzz, most companies have yet to scratch the surface of gen AI’s promise. A recent McKinsey Global Survey reveals that employees are far ahead of their organizations in using gen AI, as companies have been slow to adopt.” (McKinsey 2024 )

Maybe this also explains why there are still many areas where AI use cases are built on “do-it-yourself” rather than off-the-shelf solutions. Besides the lack of established specialized solutions in the market, the AI enthusiasts in companies seem to often drive the progress. For example for HR use cases Bain quotes the 63% of gen AI use cases are “do-it-yourself” applications (Bain Generative AI survey Feb 2024)

Conclusion

In conclusion, adoption and especially the scaling of use cases and re-imagining processes and even solutions are the main factors that currently prevent AI solutions from delivering to their full potential. It’s no surprise that stock markets raise questions given the billion dollar investments that this technology requires. But I surely don’t see an innovation trough and the next “wow”-applications might just be around the corner.?

I firmly believe that the mid-to long-term impact of AI will be profound, with changes that cannot be overstated. They will significantly impact the future of work and with that also the future of our societies. If this change comes in the end a bit slower than some AI enthusiasts predict at the moment, we all should be grateful and use the time to prepare.?