I broadly agree with your thesis in the long run, as someone who works in tech and uses AI daily. There are many components to a successful software company and vibe coding impacts only a few of them. However, once the narrative takes hold it may be a while before it turns. Would think it better to average in over time rather than take a position now.
His little case study isn't equivalent to an ERP system, but in my view there are parallels.
In the old days, software selection exercises involved long lists of functionality to compare between vendors. In 2027 if a given vendor doesn't offer a bit of functionality you want, you'll just tell the sales guy and he'll have it done for you by tomorrow (I exaggerate, but not much).
The key implication is that software becomes a commodity with almost unlimited potential vendors and I suggest it may be difficult to maintain the current high margins and high multiples in that environment.
That's not to say that software becomes worthless, but should the providers be valued on the same multiple that they were in recent years? In my view, no.
Brilliant read. I’m of a similar view. In my world of work, we rely heavily on external tools, but they have their limitations and costs. By feeding various AI the requirements, logic, and tools we have available internally, we can close process gaps, fully customise, control and enhance our processes and sever expensive external software products.
That said the potential of AI in its current form has been over sold.
It should be used as an assistant to the right user, and can be incredibly useful, but the likes of what happened to Duolingo is a good case study.
Working nicely as a bounce trade. I meet with a variety of software companies and the general view is that AI represents a productivity tool but isn’t going to displace real deep vertical expertise any time soon. One company said it is great for producing product/release documentation. Another, that it allows segments of code to be produced in first draft as an input into code creation. Given the tendency of AI to hallucinate, and the huge liability of getting commercial software wrong (eg if it is running a telco billing system collecting £m) whatever a company does to cap its contractual liabilities (not sure how insurance works if you’ve written your software via AI), I don’t imagine that mission critical software is going to be displaced any time soon.
Might it somehow skew the economics of software businesses as they operate now, with efficiency savings being passed on to purchasers? Not sure here either. Might increase gross margins a bit (or quite a bit), but no-one buys software on the basis of what the gross margin is, and most of the creation cost is sunk in past R&D costs. If a purchaser has other options, they can assess one software product against others which do somewhat the same thing, including price and other T&Cs, including “maintenance and support” and so forth. AT the moment, I’d suppose that for software companies with real competitive advantage, including deep vertical expertise, it may increase profitability a bit in the short term, but perhaps lower it a bit long term. Substantially uncertain though I’d say, in either direction.
My largest holding (when I include my SIPP) Sold some in June, alas not enough. No intentions of selling any in the near future. Need more evidence that AI will destroy saas.
6 directors bought yesterday and they started a buy back with a 200,000 share purchase. Recent sales have been at 59% and 80% premium (I think). The downside is the Visma (20% of assets ) ipo is off (I assume). I was waiting for the ipo before selling a decent amount....such is life.
Excellent breakdown on the capital vs labor dynamic here. The point about AI essentially being K substituting for L gets overlooked too much when evreyone panics about legacy software. I've seen first hand how companies are way more scared of system migration than they are excited about shiny new AI tools, plus that 5% success rate from MIT is pretty sobering.
Lots to agree with (as usual!) but I think the issue here is that software companies will face more competition, so they become less profitable and have lower growth prospects. Pretty much any functionality is now cheap to replicate, so barriers to entry are drastically reduced. You are right that migration is a challenge, but it's not insurmountable and will itself become a lot cheaper with AI. From a software procurement perspective, would customers be less inclined to sign very costly licences when much cheaper alternatives might be just around the corner?
I don't know the answers here and I'm not sure anyone does, but I can see lots of risks to the value of software companies which have previously been valued on very high multiples.
So HGT have issued a well timed trading statement today, with a year end NAV of £5.62. Share price as I write of £4.26 (up 5%ish) . Interesting. I am also watching CSU Constellation Software.
CSU is a “bottom feeder”/buyer of last resort in the software space, buying on 6x EBITDA, no fancy revs multiples. HAd a strong upward trend until mid 2025, when the long time CEO retired on medical grounds. Prior to that, the upward trend in sp had closely tracked the broker tps. Is the new team doing as good a job as the former CEO (who I think was a founder). Share price says not. Valuation at 3x revs, and 12x EBITDA looks ok, but is the new team as good as the old one.
I broadly agree with your thesis in the long run, as someone who works in tech and uses AI daily. There are many components to a successful software company and vibe coding impacts only a few of them. However, once the narrative takes hold it may be a while before it turns. Would think it better to average in over time rather than take a position now.
Further to my previous comment, I think it's worth reading this for an alternative view to this article:
https://www.nexteconomy.co/p/the-digital-gutenberg-moment?triedRedirect=true
His little case study isn't equivalent to an ERP system, but in my view there are parallels.
In the old days, software selection exercises involved long lists of functionality to compare between vendors. In 2027 if a given vendor doesn't offer a bit of functionality you want, you'll just tell the sales guy and he'll have it done for you by tomorrow (I exaggerate, but not much).
The key implication is that software becomes a commodity with almost unlimited potential vendors and I suggest it may be difficult to maintain the current high margins and high multiples in that environment.
That's not to say that software becomes worthless, but should the providers be valued on the same multiple that they were in recent years? In my view, no.
Brilliant read. I’m of a similar view. In my world of work, we rely heavily on external tools, but they have their limitations and costs. By feeding various AI the requirements, logic, and tools we have available internally, we can close process gaps, fully customise, control and enhance our processes and sever expensive external software products.
That said the potential of AI in its current form has been over sold.
It should be used as an assistant to the right user, and can be incredibly useful, but the likes of what happened to Duolingo is a good case study.
A great read. Thank you.
Working nicely as a bounce trade. I meet with a variety of software companies and the general view is that AI represents a productivity tool but isn’t going to displace real deep vertical expertise any time soon. One company said it is great for producing product/release documentation. Another, that it allows segments of code to be produced in first draft as an input into code creation. Given the tendency of AI to hallucinate, and the huge liability of getting commercial software wrong (eg if it is running a telco billing system collecting £m) whatever a company does to cap its contractual liabilities (not sure how insurance works if you’ve written your software via AI), I don’t imagine that mission critical software is going to be displaced any time soon.
Indeed.
Hallucination, contractual liability and mission-critical systems. Are we really saying those three aspects in the same sentence?!
Recent Court judgments are that you implement AI in your business at your own risk.
E.g.
https://aibusiness.com/nlp/air-canada-held-responsible-for-chatbot-s-hallucinations-
The liability of AI actually reminds me of the 1992 Hoover US round trip flights give-a-way promo.
Let alone the reputational damage as the recent case of Deloitte and the Australian Government illustrated:
https://futurism.com/future-society/deloitte-government-ai-hallucinations
OB
Might it somehow skew the economics of software businesses as they operate now, with efficiency savings being passed on to purchasers? Not sure here either. Might increase gross margins a bit (or quite a bit), but no-one buys software on the basis of what the gross margin is, and most of the creation cost is sunk in past R&D costs. If a purchaser has other options, they can assess one software product against others which do somewhat the same thing, including price and other T&Cs, including “maintenance and support” and so forth. AT the moment, I’d suppose that for software companies with real competitive advantage, including deep vertical expertise, it may increase profitability a bit in the short term, but perhaps lower it a bit long term. Substantially uncertain though I’d say, in either direction.
My largest holding (when I include my SIPP) Sold some in June, alas not enough. No intentions of selling any in the near future. Need more evidence that AI will destroy saas.
6 directors bought yesterday and they started a buy back with a 200,000 share purchase. Recent sales have been at 59% and 80% premium (I think). The downside is the Visma (20% of assets ) ipo is off (I assume). I was waiting for the ipo before selling a decent amount....such is life.
Excellent breakdown on the capital vs labor dynamic here. The point about AI essentially being K substituting for L gets overlooked too much when evreyone panics about legacy software. I've seen first hand how companies are way more scared of system migration than they are excited about shiny new AI tools, plus that 5% success rate from MIT is pretty sobering.
Lots to agree with (as usual!) but I think the issue here is that software companies will face more competition, so they become less profitable and have lower growth prospects. Pretty much any functionality is now cheap to replicate, so barriers to entry are drastically reduced. You are right that migration is a challenge, but it's not insurmountable and will itself become a lot cheaper with AI. From a software procurement perspective, would customers be less inclined to sign very costly licences when much cheaper alternatives might be just around the corner?
I don't know the answers here and I'm not sure anyone does, but I can see lots of risks to the value of software companies which have previously been valued on very high multiples.
So HGT have issued a well timed trading statement today, with a year end NAV of £5.62. Share price as I write of £4.26 (up 5%ish) . Interesting. I am also watching CSU Constellation Software.
CSU is a “bottom feeder”/buyer of last resort in the software space, buying on 6x EBITDA, no fancy revs multiples. HAd a strong upward trend until mid 2025, when the long time CEO retired on medical grounds. Prior to that, the upward trend in sp had closely tracked the broker tps. Is the new team doing as good a job as the former CEO (who I think was a founder). Share price says not. Valuation at 3x revs, and 12x EBITDA looks ok, but is the new team as good as the old one.