Artificial Intelligence – The business

Pressure is mounting to generate revenue.

• The early signs of pressure are appearing in the AI industry which looks like a precursor to price erosion which RFM has predicted will be the pin that bursts the AI bubble.
• Roughly $50bn has been ploughed into an industry that has generated around $3bn in revenues, but as of today, most participants remain unconcerned with how proper returns on these investments will be made.
• I suspect that the real return on investment in AI will be made in terms of better productivity and lower costs rather than revenue, but this is not what the market appears to be pricing in.
• Instead, the market thinks that the industry is well on the way to creating superintelligent machines that can take over most economic tasks from humans creating a new multi-trillion-dollar industry.
• It is here where I see the disconnect between expectations and reality.
• RFM Research has concluded that generative AI offers breakthrough improvements in the use of natural language as a man-machine interface and in the ability to categorize, store and retrieve unstructured data.
• RFM finds that there is no concrete evidence that generative AI can reason and plenty of evidence indicating that it cannot.
• Furthermore, very little progress on reducing hallucinations has been made and controls that are being put in place are crude and greatly hamper the functionality of the algorithms.
• This is roughly equivalent to making a service so secure that no one can get access to it thereby rendering it useless.
• When it comes to market-facing generative AI services, red flags are popping up regularly.
• For example, OpenAI will now let you use ChatGPT without an account which looks like a move to shore up usage which by most objective data, has been flat for a while and may now be flagging.
• Perplexity AI is now going to start selling advertisements despite previously promising not to in a sign that other monetisation methods are failing to deliver the goods.
• RFM research has also concluded that generative AI services can be successfully financed through advertising but only when the amount of revenue needed is less than $3 per user per month.
• The problem here is that most start-ups appear to be planning on charging $20 per user per month which looks unrealistic when one takes into account the growing number of services (which are pretty similar) and the thousands of free models available from open source.
• Furthermore, a survey by PagerDuty reveals that the vast majority of Fortune 1000 companies have paused their generative AI projects due to security risks while Elastic sees companies considering increasing their investments but remaining concerned about privacy and security (see here).
• I think that the criticism that no one knows what to do with AI is unfair as there are several very good use cases for companies but the industry’s inability to control it is causing a rethink.
• In the early days of 4G, no one was very sure what users would do with all the capacity they now had but the result was massive growth in smartphone usage as developers invented thousands of useful apps and services.
• Hence, I can see a period of consolidation coming just like in 2001 while everyone figures out what to do with this new technology and how to deal with the problems that it throws up.
• This will mean that the price for a generative AI service is not $20/user/month but more like $1.7/user/month and there are signs that this correction in terms of pricing may be close.
• This will upend the business plans of most start-ups forcing them to go back to their VCs for more money who will in turn reduce the valuations they are willing to pay.
• It is only if we are on the cusp of super-intelligent machines that the current hype and frenzy be maintained but there is no evidence to support this contention other than hopes and dreams.
• If I was forced to invest in this area, it would be Nvidia which has revenues and profits now or Qualcomm which is in a very good position to benefit as generative AI starts to be implemented at the edge following the period of consolidation.
• I already own Qualcomm which remains reasonably valued and has further upside from its position in automotive as well as the potential to take share from Intel and AMD in laptop processors.