Artificial Intelligence – Bubble Debate pt. II

Mr Altman sounds more like the leader of a religion than a company CEO.

• Although the view that OpenAI is seeking US state backing looks overblown, it is clear that the justification of OpenAI’s $1.4tn spend is almost impossible without assumptions that would make Pets.com blush.  
• The net result is that the roll-out is going to take much longer than expected and will very likely be interrupted by a correction that will cause those that are not generating cash to go out of business or be acquired.
• 2 events have riled the doomsayers (see here) into a new frenzy, who are more than ever, predicting that the end is nigh.
  • First, Sam Altman: who both refused to answer the most relevant question he has ever been asked and did so in such a way that made him appear to be more like the leader of a religious cult rather than a company CEO.
  • On a call with an OpenAI investor (see here), the investor asks Mr Altman to justify the $1.4tn spending commitment, and Mr Altman responds by offering to find the investor a buyer for his shares.
  • In plain English, this means “I am tired of trying to justify this, so either believe what I say or there is the door”.
  • These are the kind of responses that Theranos used to give when it was asked how its box could do the things that it said that it could, which is why this has made everyone so nervous.
  • I am not accusing OpenAI of a Theranos-like fraud, but I am concerned that its investment commitments put it way over the front of its skis.
  • If there was an answer to this most relevant of questions, Mr Altman would simply have given it, which means to me that he has no answer.
  • This is because for a company generating $35bn in revenues, this is extremely difficult.
  • If we assume that the average depreciation of this investment is 7 years (most are saying shorter), then the depreciation charge will be $200bn alone every year.
  • If OpenAI makes an operating margin before depreciation of 60%, then it will need revenues of $333bn simply to break even and revenues of $600bn to make a real operating profit of 17%.
  • These are staggering numbers, which is why the path to these kinds of revenues needs to be meticulously supported by detailed analysis rather than being taken on faith.
  • Mr Altman’s unwillingness to do this leads me to think that there is no rational path to these numbers and that he is hoping that if he builds it, they will come.
  • This is precisely what got the internet infrastructure industry into so much trouble 25 years ago.
  • Second, US taxpayer: where OpenAI’s CFO stated that the company would support federal measures to make it easier to finance sizeable investments in AI chips (see here).
  • This was immediately jumped on by the doomsayers as an attempt to make OpenAI too big to fail and set it up for a bailout when it fails to live up to the outlandish expectations that have been set for it.
  • OpenAI has quickly walked back this view, and by and large, I disagree with the doomsayer interpretation.
  • From the commentary, I conclude that what the company meant was that it would like to have a far better credit rating, such that the interest rate it pays on its debt is much lower.
  • I do not think that there is anything more to this than a simple desire to pay a lower interest rate, which any company or borrower would also be in favour of.
• The net result is that this is not a sign of an imminent collapse of the AI bubble or the spending spree that supports it, but it is a sign that economic reality is not being paid attention to.
• Hard numbers and return on investment have given way to optimism and the view that super-intelligent machines are just around the corner.
• I am not questioning whether recent developments in AI will change the way business is conducted and life is lived but I question whether it will do so in the next few years.
• The sudden re-ordering of everything needs to occur in order for OpenAI to make its numbers, and I suspect that there needs to be a correction to ensure that capital is being invested in the right places rather than a mad rush for data centres at any price.
• This will not be as bad as the internet bubble, which had no demand at the time, but those who are not technically solvent (cash positive) will end up going out of business or being acquired.
• It is at this point that Amazon acquires Anthropic, and the likes of OpenAI, Mistral and so on are forcibly acquired or end up in consolidation.
• Everyone else is likely to see a 25% – 50% decline in their valuations, which could quite possibly be a great entry point at which to buy the winners who will probably slowly recover into and surpass previous valuations as real revenues and profits materialise.