The Enterprise Stack of 2030: Five Bets Worth Making Now
Most enterprise predictions are restatements of last year's roadmap. Here are five structural shifts that will define the stack five years out, and how to position for them today.
Forecasting enterprise tech is mostly a parlour game. The honest version is narrower: which structural shifts are already underway, irreversible, and underpriced by current buyers? Here are five we are willing to put real money behind.
1. Models become infrastructure
By 2030, foundation models will be a utility, priced by the token, consumed like compute. The differentiation moves to the orchestration layer, not the model itself. Buying decisions shift from 'which model' to 'which policy and routing fabric.' The winners will look more like Cloudflare than OpenAI.
2. Operations become declarative
Teams describe desired outcomes, AI agents negotiate the steps. Runbooks become policies. The job title 'operator' becomes 'policy author.' Compliance teams stop writing PDFs and start writing machine-checkable rules that the orchestration layer enforces in real time.
3. Security shifts left of the model
Prompt injection, data poisoning, and model exfiltration become first-class threats. Compliance frameworks already lag, buyers should not. Expect a new category of model-aware security tooling to absorb the budget that used to go to legacy DLP and SIEM.
4. The edge wins the latency war
Inference moves closer to data. Centralised cloud remains, but the operational hot path lives at the edge. For African and Southeast Asian operators, this is a structural tailwind, the edge is where their users already are.
5. Enterprise software gets smaller
Monolithic suites lose to composable, AI-orchestrated micro-tools. The CIO becomes a portfolio manager, not a procurement officer. Vendors who cannot expose a clean API and a policy surface will not survive the next two procurement cycles.
How to position now
- Invest in your policy and orchestration layer before your model layer. Models are interchangeable, policies are not.
- Treat data residency and edge latency as architectural primitives, not afterthoughts.
- Build a vendor exit plan into every contract. Optionality is the new lock-in.
Ready to put these ideas to work? Start a project or run the numbers.