Some unpolished thoughts on where SaaS is actually going. Less a thesis, more a set of observations I keep returning to in board conversations.
1. Big-cap SaaS is being repriced in public
One of the largest names in the category recently lost roughly $50B in market cap on a single growth deceleration. That is not noise. It is the market saying: the multiples we paid for steady-state SaaS growth required a future that no longer looks guaranteed. When that big a tanker turns, every smaller boat moves.
2. The PE write-down nobody wants to talk about
A $3.5B private-equity acquisition in the developer-skills space was recently written down to zero. Not down 30%. Not impaired. Zero. That is one of the most violent signals you can get about a category, a thesis, and a vintage. It tells me three things:
- Some of the LBO models written in 2020–2022 are now mathematically broken.
- "Recurring revenue" is not the same as "durable revenue" when AI is restructuring the underlying job to be done.
- Operating partners — real ones, with execution scars — are about to become the most valuable people in the PE stack.
"Recurring" was a financial label. "Durable" is an operating verdict. AI is collapsing the gap between them.
3. The IPO window is frozen — and that has consequences
The exit market has been closed long enough that it is shaping behaviour. Founders are tired. Boards are tired. Secondaries are doing more work than they were designed for. When the window finally opens, the first cohort through will not be the loudest — it will be the ones with clean unit economics, defensible data, and AI inside the product, not bolted on.
4. Hyper-growth privates vs. struggling publics
The most interesting split right now: a handful of private hyper-growth names continue to compound at rates the public comp set cannot touch, while several once-darling public names are being punished for missing by a single point. The market is rewarding category leadership and punishing category participation. There is no middle.
5. AI is not a feature line — it is rewriting the budget
The most underappreciated CFO conversation of this cycle: AI is not showing up as a new line in the IT budget. It is showing up as a re-allocation of the existing one. Seats are being deferred. Renewals are being rescoped. Token spend is moving from R&D experiments into production cost of goods. Companies that still pitch "AI on top of your current stack" are going to be priced as add-ons. Companies that re-architect the workflow will be priced as platforms.
6. The arms race is real, but the moat has moved
The competition between the frontier labs is genuine and consequential. But for operators, the moat is no longer "which model do you use." It is:
- How fast can you switch models without breaking your product?
- What proprietary data and feedback loops are you accumulating that no model alone can replicate?
- How disciplined is your eval and routing layer?
If the answer to any of these is "we'll figure it out," you are building on rented land.
What I am telling boards
- Assume the IPO window opens in 12–18 months, not 36. Be ready.
- Re-underwrite every SaaS investment with an explicit "AI substitution" stress test.
- Stop debating "which AI vendor." Start designing the substrate — data, evals, routing, governance — that survives any vendor.
- Operating talent is the scarcest input. Hire it before you need it.
None of this is a prediction. It is a posture. The cycle will do what cycles do. The question is whether your operating model is built to compound through it, or to defend against it.