The Hyperscaler Test

A software category can survive competitors, price pressure and technology change. What it frequently does not survive is becoming a checkbox inside something a customer already pays for.

The test is simple to state.

If Microsoft, Google, Amazon, OpenAI, Salesforce, SAP or another platform decided this category should be a feature of their product, what would happen to this company?

This is a scenario, not a forecast. There is no claim here that any platform will do so, and predicting which ones will is not a serious exercise. The claim is narrower: a thesis that cannot survive the scenario is a thesis with an unpriced dependency in it.

Why absorption is different from competition

A competitor has to win the customer. A platform only has to be adequate, because it is already in the account, already procured, already approved by security, and included in a licence the customer is paying for whether they use it or not.

The bar for a bundled feature is therefore not “better”. It is “good enough that the incremental purchase becomes hard to justify”. Products have lost categories to features that were plainly worse, and the mechanism was never a product comparison. It was a budget conversation.

Cheaper software makes this more likely rather than less. The cost for a platform to reach “adequate” in an adjacent category is exactly the cost that AI is compressing.

What raises the probability

• The category is a natural extension of a workflow the platform already hosts
• The data the product needs is already inside the platform
• The product is broadly horizontal rather than specific to an industry
• Buyers are technology functions rather than a specialist line of business
• The category is large enough to matter to the platform, but not large enough to attract regulatory attention
• Adequate is genuinely adequate — precision, compliance and liability requirements are modest

What lowers it

• Deep vertical specificity, where the remaining ten percent of the requirement is where all the difficulty lives
• Regulatory approval, certification or liability that a platform does not want to carry
• Physical installation, service or hardware dependencies
• A buyer who is not the platform’s buyer, and a sales motion the platform does not run
• Data the platform does not hold and cannot acquire without a fight
• A market too small to move a platform’s revenue line, which is a form of protection worth naming plainly

Running it properly

The test is only useful if the answer is specific. “We would be affected” is not an answer. The version worth writing down has four parts.

1. Which platform, and why that one

   Absorption comes from a platform whose customers overlap with the target’s. Naming the specific candidate forces the argument to be concrete, and often reveals that no plausible candidate exists.

2. What “adequate” means here

   Establish the actual functional bar at which a customer would stop paying separately. It is almost always lower than the target’s management believes.

3. Which revenue moves first

   Absorption rarely takes a whole market. It takes the least demanding segment, which is often the most profitable one, and leaves the complex accounts that cost the most to serve.

4. What the response would be

   Verticalisation, moving deeper into the workflow, becoming a partner rather than a competitor, or exiting. Each has a cost and a timeline, and both belong in the model.

A thesis that comes through this exercise intact is meaningfully stronger than one that never ran it. A thesis that does not come through has not been disproved — but the buyer now knows what they are underwriting, which is the entire point of diligence.

The complementary reproduction question is the 80% rebuild test; the assets that tend to survive both are set out in what remains scarce.