Goldman Sachs: GEN AI: Too much spend, too little benefit

Software
,
1

Interesting 32-page analysis by Goldman Sachs.

It seems the general focus is on:

  1. can and how fast will AI replace existing jobs?
  2. where do we get the power for AI.

We’ve discussed this before.

However, I wonder if the reality is more like this:

  1. The interesting applications for AI will be applications we have not thought of yet, not replacing existing jobs.
  2. At the edge is where we really need AI so we are not flooding our data centers with data we don’t need. Also, power is less of a concern at the edge where we can run specialized, efficient models and utilize computing resources that are already running.
2

Some interesting takeaways in that Goldman Sachs newsletter. Money continues to pour into AI investments despite the fact that future ROI seems uncertain. Without a “killer app” for AI returns on investment may never be realized. In spite of all the promises of AI the contribution to annual GDP growth is likely to be less than 1%. It’s likely that AI is in a “bubble” right now, but these kinds of bubbles take a long time to burst. Best investment strategy for the near term is in “picks and shovels” companies like Nvidia that are actually building the hardware infrastructure for AI. Possibilities for investments exist in utility projects, especially in places like Virginia that are facing massive power shortages due to lots of ongoing data center construction. That generative AI is currently massively overhyped is also obvious. There are only a few tasks that readily benefit from AI. Difficult problems exist that AI is not suited to solving. Productivity gains where AI has been deployed are modest at best.

1 Like