I seems to me the main risk long-term is that it's not clear where the moat is.

People will pay a premium if there is marginal value in extra intelligence. And Anthropic is able to retain their lead and provide that extra intelligence.

Both of these assumptions seem highly suspect to me. Most businesses don't exclusively hire PHD-level employees for most tasks because a competent slightly less intelligent/knowledgable person can do the job just fine for significantly less.

Then there is the question of if they're able to retain their lead, and again, this seems questionable. Both OpenAI and Google are competing at the frontier, and Chinese labs are still in the race. Things are changing so fast who knows where we'll be in 5-10 years.

My best guess is that longer-term models will be commoditised and it will be how companies offer products and services around their models that gives them their margins. I wouldn't bet on Anthropic's ability to do this. Meta, OpenAI and Google are much better placed to win here imo.

The other place there may be margin is in niche research domains like drug-discovery. Any slight lead in these sectors could give them a lot of pricing power.

I think the reason to invest in this is as a hedge. AI basically competes against everything and arguably for this reason it's risky to not own a basket of AI stocks.