The short version

83% of South African organisations have moved at least some AI workloads off public cloud, a Cloudera survey found, and 47% had moved an AI workload because of higher or more volatile costs.

What happened

Cloudera’s Great AI Re‑Architecture Survey, conducted in June 2026, reported that 83 % of South African organisations have shifted at least part of their AI workloads off public cloud providers.

The background

The State of Enterprise AI 2026 report indicated that 85 % of EMEA decision‑makers have curtailed or restructured AI use cases due to increasing token consumption.

Data‑residency and sovereignty rules influence where AI is run for 82 % of respondents.

What it means for owners and operators

Owners should expect higher total‑cost‑of‑ownership if they keep AI in public clouds; moving to private or hybrid environments can stabilise spend.

Operators will need to redesign applications to run on‑premise or in edge data centres, increasing short‑term engineering effort but giving better control over data and compliance.

Executives must embed governance checks, data source approval, output visibility and error handling into AI project pipelines to avoid costly shutdowns.

Founders building AI agents should separate model logic from data storage so they can switch compute sites without rebuilding the whole system.

The numbers

83 % of South African firms moved AI workloads off public cloud; 47 % did so because of cost volatility.

85 % of EMEA leaders limited AI use due to token costs; 75 % listed governance as the top barrier.

Only 23 % of EMEA respondents say AI generates returns higher than its costs, while 57 % break even.

SA firms moved AI off public cloud
83%
Moved due to cost volatility
47%
EMEA limited AI for token costs
85%
Governance barrier EMEA
75%
Data‑residency influences
82%
EMEA AI returns > cost
23%
Source: as reported by itweb.co.za.

Our take

Analysis by Infercline

Nearly half of South African firms have moved AI work because of cost, so check what each workload costs where it runs.

What to do in your business tomorrow

  1. Audit your AI spend to identify how much token cost is eating into your budget.
  2. Map the data‑governance steps required for each AI use case and decide which workloads can be shifted to a private environment.
  3. Build a pilot that separates model code from data storage so you can move the compute layer without redesigning the whole application.

Related coverage

Sources

  • itweb.co.za: Rising token costs and governance force AI adoption strategy rethink