Will the AI bubble burst or is GenAI here to stay? The artificial intelligence industry is experiencing unprecedented financial euphoria. Yet, the current situation is very confusing. AI investments are reaching dizzying heights. Let’s mention OpenAI’s $40 billion funding round at $300 billion valuation and Mistral AI’s €1.7 billion funding round. Yet, some commentators are very critical of the situation. For instance, Ed Zitron predicts that the AI bubble will burst in Q4 2025. All this is fueling intense, rather than rational, debate. I wanted to confront these concerns with the expertise of Bernhard Schaffrik, Principal Analyst at Forrester Research. His analysis is insightful and nuanced. In his mind, there will be some sort of correction, but at the same time, GenAI is too popular to disappear.

When Will the AI Bubble Burst?

<figure class="wp-block-image size-large"><figcaption class="wp-element-caption">Is the AI bubble about to burst or is GenAI here to stay? Forrester’s Schaffrik predicts corrections but says GenAI is too popular to go — photo by Forrester.com</figcaption></figure>

Forrester’s Bernhard Schaffrik is recognized as one of the most insightful experts in artificial intelligence. He provides a nuanced analysis that transcends simple financial considerations. His perspective on the AI bubble burst scenario offers first-hand insights for understanding where this transformative technology is truly heading.

The AI Bubble: Financial Reality, Technological Continuity

The question of a potential AI bubble burst cannot receive a univocal answer. As Bernhard Schaffrik rightfully points out, it all depends on one’s perspective. This duality of vision probably constitutes one of the keys to understanding the current situation and the likelihood of an AI bubble burst.

<figure class="wp-block-image size-large"><figcaption class="wp-element-caption">Like us, Schaffrik righfully points out that the main issue with AI is societal and philosophical — image generated with Adobe Firefly </figcaption></figure>

“It’s almost impossible to get a one-sentence response from an analyst. Allow me two sentences. Number one is, of course, it always depends on the role or the profile you’re asking. If we are talking about financial investors, then yes, there are strong signals of this being a bubble because there is so much money being pumped into it—more than $120 billion US dollars in capital expenditure on AI infrastructure alone, just by the Magnificent Seven tech providers. So that bubble could burst,” explains Forrester’s expert.

This assessment gains particular relevance when considering Google’s $9 billion AI data center investment in Oklahoma for advanced AI training infrastructure.

This financial perspective, however, tells only part of the story. Technological adoption follows a different logic from financial markets, as Schaffrik confirmed during our exchange about the AI bubble burst potential.

“But now, if you put yourself in the shoes of enterprise decision makers, tech decision makers, also AI users, there are many who would say, ‘I don’t care if that bubble bursts, the technology is there, and it won’t go away.’

“Regardless of the amounts all the financial transactions surrounding the AI industry, people are actually using this technology. And they like what they are seeing. It might not be the disruptive, transformative value some are surmising. It’s probably more incremental than that, but the adoption of that technology is undeniable.”

The Revenue Challenge: A $25 Billion Gap to Bridge

Fortune’s analysis reveals a concerning gap between current investments and generated revenues. To justify current investments, AI companies would need to generate $40 billion in annual revenue, while they currently produce only $15 to $20 billion.

<figure class="wp-block-image size-large"><figcaption class="wp-element-caption">Schaffrik doesn’t believe in an AI bubble burst right now — image made with Adobe Firefly</figcaption></figure>

I was wondering whether this $20-25 billion gap could represent a systemic risk that could trigger an AI bubble burst.

Schaffrik remains relatively optimistic on this point: “There is still enough money in that market to back these revenue gaps at least for a while. And what I’m also seeing is that especially when it comes to the largest enterpriseson the planet, they are convinced to continue using that software. And if it comes at a premium which is decent, arguably, maybe a couple percentage points higher than what they are paying today for the software, then this seems to be acceptable.”

This acceptance of additional costs by large enterprises stems from the incremental value they perceive, even if it hasn’t yet reached the promised transformation level that might prevent an AI bubble burst scenario.

LLM Regression: A Warning Signal?

A particularly troubling element in the current ecosystem is the recent NewsGuard study revealing that major LLM systems are no longer progressing but regressing, generating more hallucinations and errors. This observation raises fundamental questions about current technology maturity and its impact on AI bubble burst predictions.

“I’m not saying that LLMs and generative AI are progressing in a linear fashion nor that this technology will be disruptive in any way, despite the promise. As we have seen with emerging technologies for decades and even centuries, it takes breakthrough technological revolutions rather than evolutions to fulfill such promises,” analyzes Schaffrik.

This vision of the current limitations of AI doesn’t diminish Bernhard’s long-term optimism: “But I’m also convinced that these breakthroughs will happen, not within the next seven, eight, nine, 12 months, but maybe in the long term. Something else will be coming up.”

Energy Efficiency: The Achilles’ Heel of AI

One of Schaffrik’s most compelling criticisms concerns the energy efficiency of current systems. His comparison between the human brain and data centers is striking and relevant to understanding whether we’re facing an AI bubble burst.

“If we look at the amount of energy our brains are requiring to create a certain inference, and how much an LLM would require to achieve the same result with electricity, this cannot be the way forward.”

This energy inefficiency constitutes a major barrier to scalability and will require significant technological breakthroughs to overcome, potentially influencing AI bubble burst timing.

Pilot Failures: Business as Usual or Red Flag?

The 95% failure rate of corporate AI pilots revealed by MIT research might seem alarming and suggestive of an impending AI bubble burst. Yet Schaffrik places this figure in its historical context: “It’s quite normal. As an analyst covering innovation management, what I have been observing over time is that about 10% of all innovation-related minimum viable products, proof of concepts, pilots, will turn into a product.”

The problem would rather lie in unrealistic expectations: “Everybody rushed at it because one believed that since it’s accessible through natural language, it should  be easier to deploy, to implement, and there are no drawbacks and negatives. That might explain that the failure rate is slightly higher than with technologies we saw in the past.”

This assessment aligns with Gartner’s prediction that 30% of GenAI projects will be abandoned after proof of concept by 2025 due to poor ROI and unclear business value.

AGI: The Ne