SaaS: With AI, every which way you lose

The latest spending forecast data from analyst firm Gartner shows that spending on datacentres is set to increase 10% in 2024. This is largely being driven by hyperscalers riding the generative AI wave and kitting out their cloud facilities with the latest AI servers, each of which costs over $32,000.

The hyperscalers have deep enough pockets, the buying power and strategic alliances to secure the supply of graphics processing units they anticipate they will need to run AI workloads. AI powers internal operations and supports business development functions as well as offering a hot feature their customers want to consume. In this highly competitive market, IT buyers will be enticed by better technology such as more advanced LLMs (large language models) or on better price/performance. The hyperscalers have deep enough pockets that they can afford to build out AI infrastructure.

AI in Saas

SaaS (software as a service) providers have a bit less flexibility. They too, operate in a highly competitive market, and need to offer AI functionality in business applications. SaaS firms are having to make big bets in AI: they need to build out AI infrastructure to support functionality that may or may not get adopted en masse by their customer base. It’s a huge risk because unless they operate in a niche where they are the dominant player, customers will have a choice. This means that the SaaS firms are not only competing with each other on functionality and price but now they have to take into account the upfront investment they need to make to build viable AI-enabled functionality. The IT infrastructure required to build and run AI-enabled enterprise software is having a material impact on their revenue.

Every way they look strategically, is likely to end up as a loss. If they charge for AI, they will eventually lose out to those competitors who give away equivalent AI functionality for free, or charge less. If they choose to make it free, they do so knowing that they may lose everything they’ve invested if customers decide not to adopt their AI functionality. And even if some customers do use their AI-enabled enterprise software, they need to ensure that they have provided enough upfront AI capacity to meet demand and have an accurate forecast of how this is likely to grow, in order to ensure they can reserve sufficient instances of GPUs (graphics processing units) and AI-powered servers to keep up with demand.

IT buyers need to recognise the difficulties enterprise software providers are now facing as AI hype intensifies. We need to consider the unique value AI offers our organisation and the cost we are prepared to pay to achieve this.

Data Center
Data Management