The first three parts of this briefing asked what the speech announced, where it could pay off, and where it could fail. This part asks a different question — the one every builder, founder, board and investor actually cares about: if these standards land roughly as described, who comes out ahead?
The answer is counter-intuitive. The regime is close to sector-neutral. It does not reward "data centres" or "AI startups" as categories. It rewards a way of building — bring your own firmed power, prove where your data came from, keep the capability onshore, make rather than merely host — and it penalises the mirror image of each, wherever it appears. The same sorting line runs through every group below.
This part is analysis and scenario, not reporting. It reasons forward from the mechanisms the speech names; it is not a forecast of what Parliament will pass. Read it as a map of incentives, not a prediction of outcomes.
The Sorting Rule
Strip the speech to its machinery and one axis organises everything. At one end is the warehouse: rent out land and power, draw from the grid, host products made elsewhere, keep as little of the value — and carry as little of the cost — as possible. At the other is the maker: arrive with your own generation, account for your inputs, keep capability and IP in the country, and contribute more than you extract.
The whole point of acting "before the concrete is poured" is to price the warehouse out and pull capital toward the maker. Five mechanisms do that work — and each one is a filter that lets one posture through and holds the other back.
Certainty rewards the ready
A single national framework offers "clarity and speed for approvals" — but only to those who can already meet the conditions. It is a tailwind for the prepared and a wall for the speculative.
The net-generator bar
"Net generators, not net users" is the sharpest filter in the speech. It eliminates the land-and-grid-power model outright and admits only those who bring generation with them.
Priced on the driest continent
Minimise use, pay for the infrastructure. A cost that favours efficient, recycled-water, well-sited builds and penalises the thirsty and the careless.
Provenance becomes an asset
Licence-and-pay "control" turns clean data provenance into a moat — and turns "trained on whatever we scraped" into a liability. It also mints a whole new licensing market.
Make here, don't just host
The push for "more of a stake in where AI is made" tilts everything — investment, procurement, approvals — toward those who build capability onshore rather than renting it from abroad.
Hold these five filters in mind. From here on, every group is just the same five questions asked of a different actor.
Data-Centre & Infrastructure Builders
The marquee case, and the clearest. The energy rule is written almost precisely to sort this group. A hyperscale facility runs 24/7 at high, steady load; requiring it to be a net contributor to the grid is only survivable if generation and firming arrive with the build. That single condition draws the line, and it draws it far to the right.
So the winners here are not "data-centre companies." They are a specific kind of data-centre company — the one that shows up with its own power.
The bar sits far right: the energy rule alone eliminates the pure land-and-power model. Only operators who arrive with their own firmed generation clear it — which is exactly the "maker" outcome the speech is engineering.
Thrives
- Developers arriving with their own firmed renewables & storage
- Projects sited near transmission and recycled / non-potable water
- Builders who pre-commit to net-contribution and full grid-cost share
- Consortia partnering with Australian energy and skills providers
Squeezed
- Pure land-lease / colocation banking on cheap grid power
- Speculative "shell" data parks with no generation plan
- Latecomers hoping to renegotiate after the slab is poured
- Any model that needs its costs passed on to households
The nuance the speech invites — and the whole reason this part exists — is that "build a data centre" is not one bet. The colocation landlord and the generation-integrated developer are in the same industry and on opposite sides of the line.
AI Startups
For founders the line is drawn by two filters at once — copyright and sovereignty — and they happen to reward the same answer. Can you show where your training data came from, and are you building here? A licence-and-pay regime turns provenance into a moat; the "make here" push turns an onshore footprint into a tailwind on investment, procurement and approvals.
Two filters, one answer. Clean provenance and an onshore build put you to the right of the line; a scraped-data wrapper with servers offshore has no story to tell either the copyright regime or the sovereignty push.
Thrives
- Sovereign model, compute and inference builders — "make it here"
- Energy-tech: firming, storage, cooling, water-recycling, VPPs
- Provenance, licensing-market and compliance tooling (the copyright rule creates the market)
- Vertical AI trained on licensed, Australian domain data
Squeezed
- Thin wrappers with no moat and no local presence
- Offshore-only hosting with no sovereign story to tell
- Anything trained on data whose provenance can't be shown
- Capex-light plays that assumed scraping stays free
The copyright rule is usually read as a cost. For the right startup it is a market: someone has to build the rails that license, price and pay for millions of works.
Corporates & Enterprise Adopters
For established firms the sorting filters are sovereignty and jobs. The framework quietly rewards companies that keep capability — models, data, skills — in-house, and it puts a soft penalty on pure cost-out automation through the "good jobs, not replace them" test and approvals that are tied to skills and training. The outsourcer who hands the whole stack to "foreign monopolies" is exactly the dependency the speech is arguing against.
The line sits near the middle — most firms can move either way. Those that retain capability and invest in people clear it; pure headcount-cut automators run into the jobs framing, and pure outsourcers forfeit the sovereignty premium.
Thrives
- Early adopters who retain models, data and skills in-house
- Firms that tie AI investment to training and "good jobs"
- Data owners who can license their corpus into the new market
- Sovereignty-aligned suppliers to government and critical sectors
Squeezed
- Pure outsourcers ceding capability to "foreign monopolies"
- Headcount-cut-only automators, against the jobs framing
- Laggards waiting for the rules to settle before moving
- Anyone reliant on training data they can't account for
Government & Public Sector
Government is both referee and player, and for the player side the filter is political rather than physical. The decisive test is National Cabinet. States that sign the single framework gain coordinated investment and a clarity they can market; hold-outs fragment the framework — and the leverage it promised — back into the "disparate collection of states and councils" the speech warns against. The right column here is less "losers" than the regime's own exposure.
The line is drawn by a signature. States that sign gain the coordinated pipeline and the clarity; hold-outs keep their turf but lose the national leverage — and every undefined term ("net", "large") is a place the framework can still fracture in drafting.
Gains ground
- The new Office of AI and coordinating ministers gain real remit
- States that sign up and court "maker" investment on national terms
- Agencies building an evidence base — real-time labour-market data
- Sovereign procurement that favours domestic capability
Where it's exposed
- States that hold out and fragment the "single framework"
- The Commonwealth's inability to compel state planning powers
- Enforcing copyright "control" against models trained offshore
- Every undefined term that must survive contact with drafting
The federation dependency is examined in full in Part three. Here it matters as a sorting force: it decides which governments end up inside the framework's benefits and which outside them.
Investors & Capital
The energy rule doesn't just sort builders — it reprices the trade for whoever funds them. Requiring generation and firming alongside compute makes the data-centre bet capex-heavy and long-horizon. That is a headwind for fast-flip capital that assumed cheap grid power and quick approvals, and a tailwind for patient infrastructure and energy capital that can underwrite the whole stack and treat regulatory certainty as a durable asset.
The bar sits high and to the right: the net-generator rule turns a quick-return trade into an infrastructure one. Capital that can wait and fund the grid thrives; theses built on cheap power and speedy exits are repriced the day the rule lands.
Thrives
- Long-horizon capital that can fund generation + storage alongside compute
- Investors who price regulatory certainty as a durable asset
- Funds backing sovereign capability, energy-tech and licensing infra
- Those underwriting the grid the "net generator" rule demands
Squeezed
- Fast-flip capital wanting speed without conditions
- Theses built on cheap grid power and lax water access
- Bets on renegotiating terms after construction
- Capex-light "host it overseas" plays losing the sovereignty premium
Creators & Rights-Holders
The copyright pledge is real leverage — but the sorting filter inside it is brutal: can you license and prove provenance at scale? Collecting societies, publishers and platforms that can build or run the licensing rails thrive. The individual creator, celebrated in the speech, runs straight into the transaction-cost wall the speech doesn't solve — millions of works, priced and policed one at a time, against models that may already have ingested them or trained offshore.
The pledge protects those who can act at scale. Organised rights-holders with licensing and provenance infrastructure clear the line; the lone creator has the moral claim but not the machinery to price or enforce it.
Thrives
- Collecting societies and rights-holders who can license at scale
- Platforms building the licensing, payment and provenance rails
- Publishers and catalogues with clear, enforceable ownership
- Creators represented by bodies able to pursue enforcement
Squeezed
- Individual creators facing the transaction-cost wall alone
- Works already ingested by existing or offshore-trained models
- Anyone relying on "plain as day" law before an enforcement mechanism exists
- Rights that can't be proved, priced, or policed across borders
The Master Map
One grid holds the whole argument. Read across a row to see how a segment fares against each filter; read down a column to see who a single mechanism sorts. The values are directional judgements — tailwind to headwind — for the typical player in each group, not scores.
certainty
energy rule
obligation
licensing
"make-here"
Analysis, not scorekeeping. The revealing pattern: the two vaguest mechanisms — the net-generator rule and copyright "control" — are also the two most powerful sorters. The parts of the speech least specified are the parts that most decide who wins.
One Question Per Segment
The map reduces, in the end, to a single test for each group. If you can answer it "yes," the regime is a tailwind. If you can't, no amount of being in the right industry will save you.
It reads like an opportunity map. It is really a filter. It doesn't ask what industry you're in — it asks how you build.
Bring your own power, prove your provenance, keep your capability, make it here — and the regime is a tailwind in any sector. Do the opposite, in any sector, and it is a headwind you priced in too late. That is the whole point of acting "before the concrete is poured": to decide the winners by posture while posture is still a choice.
Read the rest: Part two — the opportunities → · Part three — the caveats → · Overview →