Hide table of contents

Many analysts have tried to undermine the dangers of advanced AI pointing at an "AI financial bubble", arguing the bubble will eventually burst bringing down AI innovation. But as I briefly discussed in this comment, "AI being in the midst of a financial bubble" and "AI continue developing and severely transforming society" are statements that can be true at the same time.

 

In this post I will discuss:

  • Why asking if we are in an "AI financial bubble" is the wrong question.
  • How is it possible that frontier AI keeps developing even in the scenario of a financial burst.
  • What are the implications for AI Safety and EA cause prioritization.

 

The wrong question: "Are we in an AI financial bubble"?

Questioning the existence of an AI financial bubble is irrelevant by two reasons:
 

  • Because we know the answer: Yes. By definition the AI industry is experiencing a financial bubble; "a market cicle where the price of an asset rises rapidly to levels greatly exceeding its intrinsic or fundamental value". Current AI companies valuations lie on promises of expected value, not real generated value at present.
     
  • Because AI business model remains unknown. Many are aware of the "circular economy" phenomenon AI related companies experience, where each one feeds into one another accounting results. Moreover, scaling compute cost and datacenters expenditures raise the alarms of business viability. [1]

 

What we do not know is when and whether this bubble will burst. 

 

As of early-November 2026 there are signs of distress around AI related companies stocks, plus constraints to reach fundraising targets. Pointing to an upcoming burst of the bubble.

 

Only in the scenario AI companies were able to rapidly become profitable, and show strong signs of sustainable growth, they would be able to reduce the gap between expected and real value feeding the bubble. But that is an unlikely prospect nowadays.

 

With the evidence we currently have, I would assume this financial bubble will burst before it has the chance to shrink. However, that is not the right question to ask either.

 

The right question: "Will a financial bubble burst halt frontier AI development?"

What many analysts imply by criticising the "AI bubble" is that it will lead to a halt in frontier AI development due to a running out of resources. The bubble burst will reveal the "scam" behind AI and will unable further training developments due to unfinanceable costs. However, this conclusion does not necessarily have to be true.

 

My guess goes in an opposite direction: A financial bubble burst may briefly slow down AI development, but will not stop frontier advancement in the medium and long run. Indeed, it might accelerate it.

 

We have historic examples of similar phenomena. The most recent was the dot-com bubble burst, where an estimated $5 trillion was lost between 2000 - 2002 in internet and telecom companies assets. However, in the following years the internet continued profoundly transforming our lives. Furthermore, those losses did not need to be recovered or compensated so new companies supported by this technology could keep growing. I expect something similar to happen with AI as technology, which is a different question regarding the fate of current AI companies.[2]

 

Someone could argue the internet and AI are different infrastructures, with the cost of AI development being significantly greater. This is true and current methods to train LLMs through machine learning require huge amounts of compute and energy. Assuming the training system remains alike[3], how will this be financed?

 

Through an US government bailout and posterior nationalization of AI infrastructure.

 

AI has been gradually embedding into private companies and state facilities, including military systems. This technology has become a matter of national security and a means to exert geopolitical power. Using game theory framework and analysing US incentives, it is only logical to assume the state will step in and finance the development of AI and its infrastructure.

 

November 4 update: The Democrats victory over the Senate and House post-midterms could slightly change this analysis, although not the overall result. Democrats also have the incentive to nationalize AI companies (which is even more aligned with a leftist ideology), although through an "AI regulatory and risk mitigation" discourse, rather than the "geopolitical and national security" one more aligned with current Republicans. Economic reasons could also come into play in how an AI public ownership could be justified to US citizens.

 

How could an AI nationalization be conducted by the US government? Among other possible mechanisms:

 

  • Citizens investment accounts: Bank transfers through schemes resembling current Trump accounts. Money allowances incentivized to be invested on AI related companies, as a means of indirectly financing.
     
  • National sovereign fund: A Norwegian style sovereign fund but focused on AI national companies (which would require actual profit from AI companies to be sustainable).
     
  • AI infrastructure socialization: Direct buyouts of AI companies stocks justified as "national interest". Strategic and critical sectors such as "rare earth" are currently being participated by the state on this scheme. US could approach a "China style" involvement to its national AI infrastructure.

 

If government steps in to finance AI infrastructure, questions such as the short-term viability of AI businesses or IPOs outcomes (e.g. Anthropic) lose relevance to undermine the fate of frontier AI development. Their failure will only accelerate this likely take over of the state on their infrastructure. 

 

Implications for AI Safety and EA cause prioritization

Although a nationalization of AI infrastructure will have clear economic and societal consequences, I will focus on the implications for AIS and EA specifically. [4]

 

1) Frontier AI will continue developing but at multi-stage paces. A financial bubble burst will slow down available capital, leading to employees layouts and datacentres building delays. However, after the "investors panic" is over and the nationalization gradually takes place, the pace will likely accelerate again. 

 

2) AI Safety will need to detach from US affairs. The need for independent and international AIS agencies will be more urgent in an AI nationalization scenario, especially to avoid a conflict of interest. The risk of fragmenting the industry of AIS on national influence spheres will be also a concern.

 

3) AI Safety will continue being one top cause for EA. As frontier AI develops, the risks grow with enhanced technological capabilities (although we might enjoy a temporary slow down during the burst). A secure AI will still be a concern post-bubble burst, and as AI deployment spread over society it will affect other EA current causes.

 

All things considered, EA attention and funding on AI Safety should not be dependent on whether the AI financial bubble will burst and when. If we believe the national interest of US over AI infrastructure is relevant, we should expect a nationalization that will only make our efforts more urgent and critical. 
 

  1. ^

    This is the main difference with risk-capital investments in traditional startups. Angel investors demanded clear business models, grounded future scenarios of revenue and losses, and other requirements to minimize financial risks that AI companies have not been addressing.

  2. ^

    Moreover, a financial bubble burst will not erase all AI scientific progress currently accomplished, which could be taken over by future companies.

  3. ^

    Another possibility could comprise "creative destruction" leading to much cost-efficient methods to train AI models. However, that would imply a stronger temporary halt to frontier developments due to the change of a whole industry standard.

  4. ^

    To not overextend this post and maintain it on topics relevant to EA community.

  5. Show all footnotes

2

0
0

Reactions

0
0

More posts like this

Comments
No comments on this post yet.
Be the first to respond.
Curated and popular this week
Relevant opportunities