Elementl and Google Cloud bring Gemini to nuclear siting
Elementl Power says Gemini and Big Query can cut preliminary nuclear site research from months to days or weeks, while controlled design records could carry approvals into standardized projects.

Why it matters
- November 17, 2026: Elementl Power announced a Google Cloud collaboration for Gemini-based nuclear preconstruction tools, beginning with Atlas.
- Atlas screens hundreds of site characteristics across thousands of square miles and multiple reactor designs.
- Elementl reports preliminary desktop research can finish in days or weeks instead of months, a saving that does not cover full licensing or construction.
- May 7, 2025: Google committed early-stage capital to three advanced nuclear projects, each planned for at least 600 megawatts.
- Elementl's goal calls for more than 10 gigawatts online in the United States by 2035.
By Marcus Bennett
On November 17, 2026, Elementl Power announced that Google Cloud will help it develop Gemini-based artificial intelligence tools for its nuclear-project portfolio, focusing on preconstruction (the planning, siting and approval work before construction) and beginning with an application designed to turn months of desktop screening into days or weeks.
Elementl Power said the first Lattice application, Atlas, can screen candidate sites across multiple reactor designs, while the broader Lattice suite will use Gemini for business and Big Query. The approach matters because Elementl wants more certainty about nuclear-project costs and schedules before construction begins.
What will Google Cloud build with Gemini?
Elementl's Lattice, an integrated project-delivery suite (software that coordinates project workflows and records), will use Gemini for business and Big Query to collect data and insights from real-world project development and actual tool use. Elementl said the software will bring that operating experience into integrated, standardized workflows.
Under Elementl's plan, those workflows must improve cost and schedule certainty and establish a repeatable deployment process that can scale across projects. Elementl's stated unit of improvement is a repeatable development process, not a single software task. Lattice will feed project experience into later work, with Atlas handling early site selection and planned capabilities maintaining design and licensing records.
Raiford Smith, Google Cloud's global director of Power & Energy, described a major preconstruction bottleneck as an “intelligence and data orchestration problem.” Elementl used that framing to explain the collaboration's focus: teams need to organize studies, decisions and records before a plant can advance.
Elementl listed four preconstruction inputs:
- Siting requirements
- Technology selection
- Regulatory approvals
- Technical planning
The company said each can prolong development and create schedule uncertainty. Under the partnership, Lattice will coordinate those inputs using Gemini for business and Big Query, the two technologies named in the announcement.
How much faster can Atlas screen sites?
Atlas identifies sites that could host nuclear projects, screening hundreds of site characteristics against geographic, environmental and infrastructure criteria.
The application can compare several candidate locations concurrently across thousands of square miles. Elementl said Atlas also accounts for the requirements of multiple reactor designs and technologies when making those comparisons.
Elementl reports that preliminary desktop research and analysis that would otherwise take months can be completed in days or weeks with Atlas. The company presented the figure as an early-screening gain.
Elementl explicitly limited that claim to preliminary site research. The reported saving does not cover the full licensing or construction schedule. The announcement therefore supports a claim about screening duration, not an equivalent reduction in total project duration.
The practical result is simpler. Teams can prioritize locations for deeper assessment sooner.
Elementl did not list the individual site characteristics, candidate locations, reactor designs or maximum number of simultaneous comparisons. The announcement therefore leaves the early timing claim without a disclosed portfolio-wide performance breakdown.
What does the controlled design record change?\n
Additional Lattice capabilities remain under development. Elementl said they will maintain one controlled design record (a single maintained project file) covering a plant's design, licensing requirements and approved changes.
The record would support reuse of applicable work on projects with standardized designs. Elementl expects that reuse to shorten development timelines and increase licensing certainty for later plants.
Chris Colbert, Elementl's chairman and CEO, said the tools would allow the company to “advance more projects, reduce repetitive work and carry what we learn from one project into the next.” His statement links three operating goals: move more projects, avoid repeated work and transfer lessons between developments.
Under Elementl's design, approved decisions and lessons from one project would become inputs for a later project rather than remain tied to the original effort. The controlled record is the mechanism for that transfer. Atlas identifies candidate sites; the planned record would preserve design and licensing work as projects advance.
Elementl did not provide a release date for the controlled-record capabilities or quantify their expected effect on cost, schedule or licensing. The announcement presents the feature as a planned mechanism for reuse. Its benefits remain expectations.
How does the AI work connect to Google's nuclear deal?
The AI collaboration follows a separate strategic agreement that Elementl and Google announced on May 7, 2025. Under that agreement, Google committed early-stage development capital to three advanced nuclear projects.
Each project was planned for at least 600 megawatts of capacity, with an option for commercial off-take (an arrangement allowing a commercial party to take completed output) once complete. Elementl and Google also planned to work with utility and regulated power partners to identify and advance projects.
At the time, Elementl said Google's support advanced its goal of bringing more than 10 gigawatts online in the United States by 2035. That target gives the software work a numerical context: Elementl's development model depends on moving multiple projects through similar preconstruction stages.
The earlier and later agreements address different constraints, according to Elementl. The 2025 agreement supplies early capital and potential demand. The 2026 AI agreement targets the organization of project information. Together, they connect financing and demand to a development process that Elementl says must become repeatable.
Can Lattice deliver cost and schedule certainty?
Lattice's significance will rest on whether Atlas's faster screening and the planned licensing records produce the cost and schedule certainty Elementl expects across its portfolio. The announcement gives one clear timing claim for Atlas: months of preliminary desktop work can become days or weeks.
It gives no portfolio-wide cost forecast, full-schedule saving or quantified licensing improvement. The controlled-record capabilities also remain under development. Those limits leave the performance claims at different maturity levels, with preliminary site-screening evidence on one side and planned design-record functions on the other.
The first test is whether Atlas can reproduce its reported days-or-weeks result during actual site selection across the reactor designs and technologies in its comparison scope. The second is whether Elementl's teams can use one controlled record to carry approved work into later standardized projects and shorten their development timelines.
That distinction matters because Elementl's more-than-10-gigawatt target depends on a pipeline of projects, according to the company. A speed gain confined to one site would not establish the repeatable deployment process that Elementl says Lattice is intended to create.
For Elementl, the decisive result will be whether Gemini and Big Query can turn an initial screening gain and a planned design record into repeatable improvements in cost and schedule certainty across enough projects to support its 2035 goal. The announcement supplies the planned workflow and an early timing claim, but not portfolio-wide performance data.
Source: unite.ai
More from Marcus Bennett
Show full bio
Senior reporter covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.
718 articles
Related articles
- Google Cloud Opens 'Gemini Factory' to Speed Partner AI Agent Builds
- Google Ships Gemini 3.5 Flash, Promises Pro Model Next Month
- Google Launches Gemini 4 Hydrogen at $1.50 Per Million Input Tokens
- Google ships Gemini 3 Pro with a 1501 Elo LMArena score
- Google Announces Gemini 4 Argon, But No One Outside Can Use It