AI companies are struggling with gross margins, what can they learn from the industrial software giants?
In software, gross margins tend to be where the story begins. High gross margins, and you can invest more in sales, R&D, and G&A, or it can flow to the bottom line. Low gross margins and people question if you are even a software business.
The former is why investors love software, and rightfully so.
This week, TechCrunch published an article questioning if AI companies with low gross margins will ultimately limit their enterprise value. It hit me that, for the first time, some AI companies may be wrestling with a problem industrial and energy transition software has tangled with for a long time.
There are a few lessons here for industrial software has learned that we can apply to this new wave of AI-first software startups.
When the customer needs a lot of help, gross margins will ultimately be the first line on the P&L to take a hit. Why? Services associated with the sale. Data needs to be processed to avoid a garbage-in-garbage-out scenario, and implementation takes longer (e.g. more service hours) as workflows need to be built.
Ongoing customer success will be a part of the business for at least 5 years and maybe more. In industrial software, customers were adopting cloud-based solutions for the first time or applying software to existing problems for the first time altogether. As a result, they called on the vendors regularly to either build new customizations or for help solving the initial problem in a better way.
Both of the above are fine as long as gross and net dollar retention remains high. The best unit economics calculation includes both this cost AND the cost to serve – for industrials (and first-time AI customers), the latter may be higher for an extended period. That’s okay so long as the long-term revenue durability plays out over time and the business’s cost structure can support it.
AI-based companies will likely have lower gross margins for the next several years. Not only will the additional services drag on gross margins, but the high compute cost will too.
The trick will be evaluating the companies appropriately and not buying into the hype when it’s not warranted. That’s something VC hasn’t admittedly been great at for the last 3 years.
Kevin Stevens 