Who watches the cloud?

The public argument about data centres is usually forced into a false choice: welcome investment or oppose technology.

That misses the practical question.

Who independently measures what happens after approval—when a campus is energised, cooling systems run, generators are tested, water is consumed, contracts are awarded and promised jobs either appear or do not?

Today, those records sit across planning files, operators, utilities, regulators, councils and government announcements. Some data is reported at corporate level. Some is commercially sensitive. Some is published only when a permit or incident makes it visible.

No single public institution produces a comparable, site-by-site account for Melbourne’s west.

The finding

The answer is not another promotional taskforce. It is a small, independent public-interest team with technical authority, protected funding and a duty to publish.

This article proposes that model. It does not claim such a body already exists.

Which experts?

Independence requires more than appointing one auditor. The work crosses engineering, cyber security, planning, finance and public administration. At minimum, the team needs the following disciplines.

Operators should fund the system through a transparent levy. They should not select the staff, control publication or hold a veto over findings. Appointments, conflicts, methods and corrections should be public.

What does it log?

The register should distinguish what is promised, approved, built, energised and actually used. Those are different states.

Corporate reporting under the National Greenhouse and Energy Reporting scheme is important, but public corporate totals are not the same as a current, site-level community ledger. The Clean Energy Regulator sets facility and corporate thresholds and publishes qualifying corporate data. That does not answer every local question.

Tamper-proof is the wrong promise

No serious evidence system should call itself tamper-proof. The defensible standard is tamper-evident: alteration becomes detectable, attributable and recoverable.

Australia’s cyber-security guidance recommends centralised logging, secure transport, cryptographic verification, protected storage, restricted deletion rights, redundancy and trustworthy timestamps. An independent observer should apply those principles to operational and civic reporting.

1. Meters and source systems send signed records directly to a segregated collection service.
2. Every record uses synchronised UTC timestamps and a published data schema.
3. Daily files are hashed; the hash receipts are published outside the operator’s systems.
4. Records are kept in append-only or write-once storage with independent backups.
5. Every view, correction, export and attempted deletion is itself logged.
6. Operators can challenge an interpretation, but the original record and correction history remain visible.

This design protects commercially sensitive and security-critical details by publishing totals and exceptions rather than network diagrams or customer data. Transparency does not require exposing the systems it is meant to secure.

Permanent work that keeps the monoliths honest

A monitoring office creates a different class of full-time employment from construction peaks: engineers, analysts, auditors, incident specialists, data publishers and community-facing staff whose work continues for the life of the campuses.

It also forces employment claims into comparable categories. A crane onsite, a short-term electrical contract, a national vendor and a permanent local technician cannot all be counted as the same job.

The team’s purpose is not to oppose operators. It is to make good operators easier to identify and weak claims harder to repeat. Consistent measurement rewards campuses that use less water, manage power responsibly, resolve noise issues, employ locally and publish without being chased.

Where does every $100 go?

The question is powerful, but it needs a fixed denominator. “Project value”, “economic activity”, “investment”, “revenue” and “benefit” cannot be swapped without changing the answer.

Where site revenue can be independently attributed, publish the split per $100 of revenue. Where a cloud campus has no meaningful standalone revenue figure, publish the split per $100 of operating expenditure and show the limitation.

The ledger must keep unlike flows separate:

• Local community return: wages paid to residents of host municipalities, purchases from locally based businesses, council rates and charges, and enforceable local infrastructure or community-benefit payments.
• Broader public revenue: state and Commonwealth taxes. Valuable, but not automatically returned to the host suburbs.
• Operating payments: electricity, water, telecommunications, imported equipment and national contracts. Economic activity, but not necessarily community return.
• Investor and superannuation returns: financial returns to owners or fund members. These can benefit retirees, including people in the west, but they are not the same as revenue received by the host community.

APRA requires superannuation trustees to act in beneficiaries’ best financial interests. Treasury describes superannuation’s objective as preserving savings to deliver retirement income. That is why pension or superannuation return cannot simply be relabelled as local community revenue.

Victoria already has mechanisms for council rates and infrastructure contributions, including rates based on property value and planning tools that can collect money, land or works. Those mechanisms matter. They still do not produce one consolidated answer for a major data-centre site.

So what is the current number?

From the public sources reviewed, it cannot be calculated credibly.

That is not a reason to invent a number. It is the reason to build the ledger.