Cloud constraints are leaving many organisations with slower interactive workloads and higher infrastructure costs, according to research summarised by AI Magazine. The findings put a practical question to AI operators: when should compute move closer to the data, rather than stay where it is easiest to host?
Watch Desk analysis
What happened
AI Magazine reports that Forrester and Akamai surveyed more than 500 enterprise infrastructure decision-makers for a report on workload placement and mobility. Seventy-five per cent said they would run compute in more locations if provider, budget and complexity constraints disappeared. Thirty-seven per cent said those constraints were why their infrastructure was centralised, compared with 19 per cent who described centralisation as a deliberate architectural choice.
The reported figures also point to the cost of sticking with the status quo: half of respondents said they had kept workloads in locations they knew were not ideal to avoid the expense and effort of moving them. Only 16 per cent said they consistently met response-time targets for their most critical interactive workload. The survey does not establish that every organisation faces the same bottlenecks, but it gives a useful view of the trade-offs operators say they are making.
Why it matters
For interactive applications, slow responses are not just an architectural footnote. The report says 58 per cent of respondents increased infrastructure spending to compensate for performance problems, while 38 per cent had reduced a feature’s scope and 34 per cent had delayed a product or feature launch. That makes placement decisions a question of what organisations can deliver, not simply where their servers happen to sit.
AI workloads can add pressure when they need to process large amounts of data and respond quickly. The survey points to data location and distance between users and compute as common factors in missed response targets. More distributed infrastructure may help, but it also brings complexity and cost of its own. There is no magic “put it at the edge” button, regrettably.
Our read
The most useful signal is the gap between intentional design and infrastructure shaped by constraints: respondents were nearly twice as likely to cite limits as the reason for centralisation. That does not prove that decentralising workloads is the answer for every business. It does suggest that operators should test where latency and migration costs are actually coming from before treating centralisation as a settled architectural principle.
What to watch
- Whether Forrester and Akamai publish more detail on the survey’s methodology and respondents.
- Whether organisations move workloads closer to users or data, and what that does to performance and total costs.
- Whether future reporting measures actual response-time improvements, rather than plans to make infrastructure more distributed.
Discussion spark: If centralised infrastructure is cheaper to manage but often misses response targets, should organisations prioritise moving workloads closer to users and data, or keep complexity down even when performance suffers?
Sources and evidence
- Forrester & Akamai: Are Cloud Constraints Stalling AI? – AI Magazine (17 November 2026, 14:39 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.