Google-Funded AI Upgrade Targets Fusion Modeling

ANU plans to rebuild its SPEC plasma code for AI chips and release it free, a research-stage effort with no date set.

Updated on Nov. 6, 2026 in Nuclear

Large toroidal fusion device with brushed-steel coils and a cylindrical vacuum vessel in a quiet, daylit research hall.
Australian National University will use Google-funded AI to rebuild SPEC, a widely used fusion-plasma modeling code, and release it free to researchers worldwide. AI Illustration. Upload story photo >

Live Poll

Do you trust big tech companies funding scientific research to act in the public's interest?

Google's philanthropic arm has funded the Australian National University to upgrade SPEC, a widely used fusion-plasma modeling code, using AI. ANU announced the project on November 6, 2026, but did not disclose the funding amount, and the upgraded software will be freely available to researchers worldwide. The project is research-stage: the software upgrade is planned work, not a finished tool.

Why it matters

SPEC models the magnetic fields that confine plasma, so faster and more accessible modeling could shorten the design cycle for fusion experiments. The effort reflects a wider push to fold AI into fusion research as private investment in the sector climbs.

In April 2025 the experimental facility produced 8.6 megajoules of fusion energy from 2.08 megajoules of laser energy delivered to the target, a ratio of roughly four to one. That experiment lasted only a fraction of a second, and the facility has not produced electricity for the grid.

The players

Australian National University

A research university whose plasma physics group maintains the SPEC modeling code used by fusion researchers worldwide.

Google

A technology company whose philanthropic arm funded the SPEC upgrade, bringing its AI expertise to fusion modeling.

Professor Matthew Hole

The ANU researcher leading the project to upgrade the SPEC fusion modeling software.

The details

SPEC is software that researchers use to simulate the magnetic fields that hold hot plasma in place, a key step in designing fusion devices. The ANU team will rebuild the code to run more efficiently on graphics processing units, the parallel chips widely used for AI workloads. The team will also develop AI-assisted tools that help researchers convert a scientific question into a simulation, check the calculations, and explain the results.

Timeline

  1. April 2025: the facility produced 8.6 megajoules of fusion energy from 2.08 megajoules of laser energy.

  2. July 2026: the 12-month period ended with a record US$4.48 billion in fusion investment.

  3. November 6, 2026: ANU announced the Google-funded SPEC upgrade project.

  4. 2030s: the industry survey's expected window for the first fusion power plant delivering commercial electricity.

The Tech Race

The project extends SPEC, a plasma modeling code long maintained by ANU, by moving it onto GPUs and layering AI tools on top. It sits within a fusion field where private capital is growing fast, with 56 companies raising a record US$4.48 billion in the year to July 2026.

The upgraded SPEC code is planned to be freely available to researchers worldwide, but no release date has been announced. Fusion remains a research-stage technology, so no practical change will reach everyday users from this project alone.

The takeaway

Watch for the release of the upgraded SPEC code and whether ANU publishes performance benchmarks for its GPU version. Those results would show whether AI-assisted modeling measurably speeds fusion design work.

Further reading

For context on how fusion research is advancing, see the Nuclear section.

Source note: This article includes information reported by Theepochtimes.

Live Poll

Do you trust big tech companies funding scientific research to act in the public's interest?