TAR raises $120M because AI data centers cannot wait for the grid

Why it matters

TAR is explicitly asking data-center operators to pay for speed rather than cheap electricity. Its $120M round shows investors financing power delivery as a core part of the AI stack.

Patrice Becker and Leonhard Soenke's TAR said it raised a $120 million Series A to manufacture off-grid power systems for data centers, giving two software founders a roughly $1 billion valuation for moving into turbines, batteries and construction sites.

Bloomberg reported Thursday that Spark Capital led the round, with existing investors Buckley Ventures and Align Fund participating. Becker told Bloomberg that TAR plans to hire engineers and expand manufacturing operations.

The financing follows the $27 million seed round TAR disclosed in June. The two publicly reported rounds amount to $147 million, an unusual amount of capital for a young operator whose first large customer deployment was still on a three-month delivery schedule when Becker discussed it in June.

Becker and Soenke arrived in energy by way of consumer software. They previously built Meeter, a calendar application acquired in 2020, and Throne, a gifting and wishlist platform for online creators. Becker studied computer science at ETH Zurich, where a bachelor's thesis examined interpretable explanations for neural networks. Soenke studied computer science at Imperial College London, according to public profiles of the Throne founders.

As RuntimeWire reported when TAR disclosed its seed round, Becker and Soenke left the creator economy after spending time with AI labs and compute providers. Those conversations pointed them toward the physical constraint underneath the model boom: available electricity near sites where data centers can actually be built.

A $120M bet on deployment

TAR combines solar, wind, batteries and simple-cycle natural-gas turbines into modular power systems intended to run without a utility connection. The gas equipment is designed as backup for periods when renewable generation and stored energy cannot carry the load, according to Becker.

The technical components are established. TAR's bet rests on changing how they are packaged and installed. In a June 15 interview with Forbes, Becker said TAR planned to pre-wire, pre-assemble, pre-test and pre-commission equipment in factories, reducing the amount of work performed at each data-center site.

That approach resembles manufacturing a repeatable product instead of engineering a fresh power plant for every customer. Execution still depends on land, equipment procurement, construction crews, permits and a supply chain that has tightened around transformers and gas turbines. TAR cannot patch those problems with a software release.

The Series A gives TAR capital to build that manufacturing operation before conventional venture milestones such as disclosed revenue or a broad customer list. The roughly $1 billion valuation therefore puts substantial weight on Becker and Soenke's ability to turn a deployment method into defensible infrastructure.

TAR has begun developing Terminal One, a logistics hub in West Texas. TAR also maintains operations in Austin and San Francisco, where Becker has recruited across mechanical, electrical and software engineering, procurement, financing and operations.

The deadline is the product

TAR is selling time. Becker has been direct that TAR power is not cheaper than electricity from the grid. The commercial case is that an AI operator can start running servers while a conventional interconnection remains stuck in a queue.

In June, Becker told Forbes that TAR was developing a 10-megawatt continuous-power pilot on land TAR owns. Becker also said TAR was delivering roughly 20 megawatts of constant capacity to an unnamed large neocloud provider within three months, according to the Forbes report. The Series A announcement lands around that self-imposed deadline, making delivery the next test of the valuation.

TAR also claimed a pipeline exceeding 200 megawatts for 2027 and several gigawatts for 2028. Those figures describe planned capacity rather than operating projects. TAR's ability to manufacture, finance and deploy repeatable systems will determine how much of that pipeline becomes power-producing infrastructure.

The speed argument has a willing market. The International Energy Agency said electricity demand from data centers increased 17% in 2025, while demand from AI-focused facilities grew faster. The IEA expects total data-center electricity consumption to double by 2030 and consumption at AI-focused sites to triple.

Grid connections, transformers, turbines, chips and regulatory approvals are all constraining construction, according to the IEA. That collection of bottlenecks has turned the number of months required to deliver power into a competitive metric for data-center developers.

Investors are funding the power layer

TAR is entering a heavily financed market. In May, VoltaGrid announced a $1 billion investment from Blackstone and Halliburton, including $775 million in primary capital, to expand behind-the-meter generation and acquire manufacturing partner Propell Energy Technology.

Critical Loop raised $26 million in April for modular systems combining batteries, on-site generation and software controls. Exowatt raised another $50 million in November 2025 to manufacture and deploy a solar-thermal storage system near data centers.

Spark's participation is notable because the venture firm also backs Anthropic. That relationship does not establish Anthropic as a TAR customer, investor or technical partner. It does show one prominent AI investor putting additional capital into the energy capacity required to keep model developers growing.

For Becker and Soenke, the round compresses a dramatic founder transition into a few months. Their previous products moved calendar events and creator gifts. TAR must move power equipment, secure land and deliver continuous megawatts. The $120 million gives them the resources to attempt that shift at industrial scale. The first 20-megawatt deployment will show whether a software founder's instinct for repeatable products survives contact with the power business.