Base Power raised $1 billion in August 2026 to put more batteries next to American homes. Valar Atomics raised a $1 billion Series B the same month to manufacture nuclear power plants. Hadrian raised roughly $1.37 billion to build more automated factories. Firmus announced a fully subscribed $2 billion equity investment to expand AI infrastructure across Australia and Asia-Pacific.
These are not four versions of the same startup story.
One company wants to manufacture nuclear reactors. Another builds factories filled with automation and robotics for aerospace and defense production. Another is assembling enormous AI computing infrastructure that has to be designed around electricity supply. The fourth is trying to turn thousands of residential batteries into grid infrastructure.
But together, they explain something unusual about the biggest startup funding rounds of 2026: increasingly, the money is not going primarily into software applications.
It is going into the things software suddenly needs.
AI created a software boom. Now it is financing the physical systems underneath it.
For much of the previous startup cycle, software had an obvious advantage over infrastructure.
A successful application could reach millions of users without building a factory. A SaaS company could add customers without buying power plants. Capital-intensive businesses were often considered difficult venture bets precisely because scaling them required expensive physical assets, complicated supply chains and long deployment cycles.
AI has complicated that distinction.
The largest AI models require enormous amounts of computing infrastructure. That infrastructure requires chips, data centres, cooling equipment and, increasingly, access to electricity that is available where and when the machines need it.
Firmus is a particularly direct example of what happens when the AI opportunity moves down the stack.
The Australian AI infrastructure company announced a $10 billion debt financing facility in February 2026, followed by a $2 billion strategic equity investment announced in August. The company says the money will accelerate Project Southgate and its AI Factory expansion across Australia and other Asia-Pacific markets. Reuters reported in February that the company's plans included infrastructure targeting up to 1.6 gigawatts of capacity over three years. (firmus.co)
That is a different capital requirement from funding an AI application.
Firmus does not merely need engineers writing software. Its plans involve power infrastructure, cooling systems, manufacturing capacity, construction and data-centre hardware. The economics increasingly resemble industrial infrastructure because the product itself is industrial infrastructure.
That distinction matters because AI's physical bottlenecks are becoming investment opportunities of their own.
Our earlier piece, Firmus Just Raised $2 Billion for AI Infrastructure, looks at what that financing says about the increasingly blurred line between technology funding and infrastructure finance.
Valar Atomics is raising venture-scale money for something venture capital once avoided
Valar Atomics' August Series B was even more explicit about where this new capital is going.
The company announced a $1 billion Series B led by Sequoia Capital, alongside a $200 million credit facility. Valar said the financing would accelerate manufacturing of standardized nuclear power plants for the growing energy demands of AI, industry and national security. (valaratomics.com)
Nuclear startups have raised significant amounts before, and a funding announcement is not proof that a company has solved nuclear power's engineering, regulatory or economic challenges. Those challenges remain substantial.
But the Valar round demonstrates a shift in what investors are willing to finance.
The question is no longer only whether software can make AI more useful. There is also a much more literal question: where will the electricity come from?
Valar has also been part of the broader push toward smaller reactors. Reuters reported in July that Valar Atomics was among four U.S. companies to reach zero-power criticality milestones through the U.S. Reactor Pilot Program, and that its Ward 250 reactor generated electricity during testing. (reuters.com)
The distinction between a successful demonstration and a commercially scaled reactor fleet is enormous. Investors know that. Yet capital is arriving before every technical and commercial question has been settled because the potential market is increasingly tied to a constraint that already exists: power.
That is the deeper context for Nuclear Reactors Are Suddenly a Hot Startup Category. The funding round is not simply a bet that nuclear technology is interesting. It is a bet that electricity itself has become strategically scarce enough to justify building entirely new industrial companies around its production.
Base Power is betting that the grid can be expanded from people's backyards
Base Power takes the same electricity problem in the opposite direction.
Instead of generating more power through nuclear reactors or building giant centralized infrastructure, the company installs and operates batteries at homes.
Reuters reported that Base Power raised $1 billion in August at a $13 billion valuation and said the company had deployed more than 500 megawatt-hours of battery capacity across its fleet. The company also launched a home battery manufactured at its Austin facility. (reuters.com)
The idea is important because the electricity system's problem is not simply generating enough power over the course of a year. The grid must also handle demand when and where it occurs.
A network of batteries can provide backup power to households while also functioning as flexible capacity for the wider electricity system. That turns a device sitting beside a house into something closer to distributed grid infrastructure.
TechCrunch reported that Base Power's new Base Core battery stores 39.2 kilowatt-hours and that the company operates in Texas and Illinois, using different business models depending on the electricity market. (techcrunch.com)
Base Power's $1 billion round followed another $1 billion round in 2025. That is striking not because home batteries are new, but because investors are now willing to value the infrastructure layer around them at startup-scale levels.
Our earlier article, Base Power’s $1 Billion Bet, examines that bet more closely. The company's real product may not ultimately be the battery itself. It may be the ability to coordinate thousands of distributed batteries as useful infrastructure for an increasingly strained grid.
Hadrian's robots are expensive because factories are expensive
Hadrian represents another part of the same movement: the return of manufacturing as a technology investment category.
The company raised approximately $1.37 billion in Series D financing in 2026. Axios reported that the round valued Hadrian at nearly $8 billion and that the company uses AI, automation and robotics in manufacturing for aerospace and defense customers. (axios.com)
The money is not being raised to acquire users more cheaply or improve an advertising funnel.
It is being raised to build manufacturing capacity.
Manufacturing is an awkward business for traditional startup economics because physical capacity takes time and money to build. A software company can sometimes scale faster than its costs. A factory requires land, machinery, automation, skilled workers, supply chains and customers capable of placing large orders.
But that same difficulty can become an advantage when the market needs capacity that does not already exist.
Hadrian has framed its business around building more automated American manufacturing infrastructure. Manufacturing Dive reported that the Series D financing was intended to expand capacity across defense, aerospace and industrial systems. (manufacturingdive.com)
The strategic value is becoming as important as the software layer. AI can improve a factory, but somebody still has to own the machines, operate the building and make the physical components.
That is the premise behind Hadrian Just Raised $1.37 Billion to Rebuild American Manufacturing With Robots.
These companies are different businesses solving the same shortage problem
It would be too neat to say that Valar Atomics, Firmus, Hadrian and Base Power belong to one category.
They do not.
Valar is a nuclear company. Base Power operates in distributed energy storage. Hadrian is an advanced manufacturer. Firmus builds AI infrastructure.
What connects them is the increasing value of solving bottlenecks that cannot be removed with another software update.
AI can generate software faster than previous generations of computing. It cannot instantly create new grid capacity.
A language model can help design components. It cannot manufacture a submarine part without a factory.
Cloud software can move workloads between machines. It cannot create electricity when generation and transmission capacity are unavailable.
Those limitations are attracting capital because they sit underneath multiple industries at once.
AI growth needs computing. Computing needs electricity. Electricity needs generation, storage and grid capacity. Advanced industries need factories capable of producing physical equipment. National security concerns add another source of demand for domestic manufacturing and resilient energy systems.
The result is a funding environment in which the most ambitious startups increasingly look less like app companies and more like industrial projects with software embedded throughout them.
Venture capital is learning to finance things that take longer to build
There is a risk in reading every billion-dollar round as proof that the underlying business model is already working.
It is not.
Nuclear power remains difficult. Data-centre infrastructure depends on enormous capital commitments and power availability. Battery businesses face manufacturing and deployment challenges. Automated factories still need customers and sustained production demand.
Large rounds do not eliminate execution risk. In some cases, they simply finance the next, more expensive stage of it.
Firmus' financing structure itself illustrates how different this new funding environment can be. Its capital stack includes both large equity investments and substantial debt financing. Valar paired its Series B equity round with a credit facility. Hadrian also announced a $360 million revolving credit facility in August to support manufacturing infrastructure and equipment. (firmus.co)
That is increasingly what happens when startups build things measured in megawatts, factories and physical equipment rather than downloads.
The biggest startup opportunity of the AI era may not be another application built on top of the models. It may be everything the models eventually force the physical world to build.
