Commodity Hardware SaaS RIP
Last week's Salesforce earnings call sent Wall Street and many analysts into a fit about the company's growth prospects, and whether it can stay relevant in an increasingly AI-dominant world. But to point to Salesforce misses the larger point, commodity hardware SaaS is dead. By that I mean software is no longer eating the world, and new software companies are no longer creating tools that run on cheap hardware to automate sales, customer service, or finance, but rather create agents that run on expensive hardware, or make software that optimizes the hardware that runs these agents and other AI tasks.
No Startup wants To Identify as "SaaS" Anymore
There are few better indicators of what software companies of the future will look like than the most recent batch of funded Y Combinator companies. Of 169 companies YC backed in that batch, exactly one considers itself a "SaaS" company. About 80% consider themselves "AI". But this goes beyond branding. SaaS economics were built on financial metrics like customer lifetime value (LTV), customer acquisition cost (CAC), and net dollar retention (NDR) with an emphasis on revenue growth.
In the 2010s, SaaS companies built software in an era of peak Moore's Law and ever declining costs per GB for storage and cost per Gbps for network, so it made sense to focus on high sales costs, not ever declining unit costs for infrastructure. This also fueled the growth of public cloud, as many SaaS providers realized they could run their whole operation without much thought about underlying hardware.
Post-SaaS Financial Management Needs to Center on Infrastructure Costs, not Sales and Marketing
The FinOps profession itself really only developed at the end of the commodity hardware SaaS run about five years ago, as companies started to see more commas in their AWS bills. But even then, most SaaS CFOs were far more familiar with sales efficiency metrics than infrastructure efficiency ones. But this approach is dead.
Unlike seat-based 2010s SaaS products, agents are increasingly priced based on the outcomes they produce, not how many employees have access to them. And the primary resource used to create an outcome is not a salesperson, but the memory, power, and GPUs supporting it. Therefore, aligning outcomes to resource consumption is essential to managing margins for post-SaaS software companies.
While embraced by startups like Decagon and 11x , established software companies like Zendesk and Salesforce are employing outcome-based pricing. Whether per conversation, per resolution, per lead generated, outcomes are replacing seats as the means for billing for software.
Cost per Outcome is More Important Now than LTV or CAC
This is completely transforming the concept of unit economics. No longer a "nice to know" about what it costs to serve a customer or deliver a product feature, it is now the number one margin driver for outcome-based software. The side FinOps practice setup to manage the AWS bill needs to become front and center in managing infrastructure margins. But this places a responsibility on the FinOps team to move beyond anomaly detection, reserving compute capacity, and storage tiering tactics of the 2010s that should be automated now, and focus on the infrastructure margins needed to sustain an outcome-based pricing strategy.
Salesforce is not the only company that needs to adjust its thinking, the entire software industry needs to re-think its financial management practices, especially around unit costing.
Ultimately, do you care more if Salesforce is ready for the post-SaaS future or if you are?
