Data centres are now a battleground. A cancer result also demands attention.

Picture a planner in New South Wales. On her desk is a queue, not a load: seventeen data-centre projects that want a plug to the grid, most of them not built yet. She is not inviting a political fight. She is inviting jobs, rates, and a seat in the race, the same invitation America already ran. This week showed what comes after.

A county invited the buildings for the tax base and the jobs, and Loudoun got rich on that invitation. Loudoun is now asking at what cost, because the grid could not keep up. Texas remains halted, as discussed last week. The buildings that exist keep drawing, and the ones that want a plug wait on the grid. Household bills are how an invitation becomes politics. Then the mid-terms arrive and Governors who championed the same projects are campaigning against them.

Hyperscalers tried to skip the queue. Bring-your-own-power means putting a plant behind the fence so you do not have to ask the grid. Easier said than done, plus in a market with less energy regulation than Australia. Regardless this week turbine and engine failures widely published at xAI’s Colossus site in Abilene. The workaround gets you past the interconnect. It does not get you past physics.

That is the American cycle: invitation, grid, household bill, ballot, workaround. The rush starts as competition. It becomes a bill. Then it is a debate people vote on. This week the Australian states tried to write away the vote in advance.

AEMO’s June-quarter report, published in late July, is the picture on that planner’s desk: seventeen data-centre projects with 9.0 gigawatts of maximum connection capacity. Twelve of those, 7.6 gigawatts, are still at application. Five, 1.4 gigawatts, have reached implementation. The load already drawing in New South Wales averaged 451 megawatts; Victoria averaged 204. Connection capacity is not the power already being used.

Australia spent the same seven days writing the firebreak: take the local veto off the table before the bill arrives. New South Wales published data-centre guidelines. Fast-track the campuses if operators pay for their own water and power, promise no extra cost on household bills, and take 40 per cent of their electricity from wind (I’d argue a mathematician would struggle with these sums). An Office of AI now sits in the NSW Cabinet Office. The same week, the state made it harder for objectors and councils to block a project. Capital Brief reported that councils are aghast. The pipeline being protected is $50 billion.

Queensland looks like the opposite fight. The state government is arguing with Canberra over mandatory renewables. The policy allows gas, the option America is already trying. Queensland also took planning off councils. People can still object. They do it through consultation, not a veto. New South Wales Labor and Queensland LNP disagree on the fuel. Both are cutting the local right to say no in different ways.

The Investment is huge, it is a massive opportunity for the Australian economy. The vote is also important, especially read against US Governor Abbots latest statement saying the AI industry and data-centres have themselves to blame for failing to get community support. Political accountability will be framed and this debate will heat up.

The tumour has already been cut out. What remains is the risk it comes back. Merck and Moderna sequenced that tumour and built an mRNA shot for the mutations that belong to that person alone: up to 34 neoantigens, the fragments the immune system can see, encoded on a strand. They gave the shot with Keytruda, after surgery, to people with high-risk melanoma, and compared it with Keytruda alone. 1,137 patients were assigned at random. Two received the combination for every one who received Keytruda alone. Neither the patient nor the doctor knew which group the patient was in.

The companies said the combination had met both of the trial’s main tests: whether the cancer came back, and whether it spread. Phase 3 is the large late-stage trial meant to show a treatment works, not only that it looks promising. This was the first positive Phase 3 for a therapy built from a patient’s own mutations, and the first for an mRNA cancer treatment. The companies called the improvement clinically meaningful against Keytruda alone after surgery.

The shot is personal because the mutations are that person’s, so choosing which fragments to put on the strand is a modelling problem. Adnan Khattak, a trial investigator in Perth, told Nature that Moderna uses artificial-intelligence tools to identify which neoantigens are most likely to trigger a strong immune response, and that researchers then narrow the set. Neither Moderna’s release nor Merck’s uses the word AI. Personalisation is the mechanism, and the ranking is one step inside it. Whatever ran that step, it did not need a campus the size of Colossus (Empathy here in that AI Leaders have repeatedly asked for a pause, I’ve written about the AI overhang, the reality is as long as their is a race West vs. East and the threat of weapons nothing slows down). That is research and modelling, not proof a frontier model designed the vaccine.

The companies will take the data to a medical meeting and to regulators. Stephen Hoge, Moderna’s president, said that thousands of people whose high-risk melanoma have already been cut out could benefit as soon as next year, if regulators approve. That is a dated hope, not a label. The market believed the result anyway: Moderna jumped as much as 160 per cent on the news.