AI is making sophisticated intelligence available to more people and businesses. But putting that intelligence to work requires something the digital economy cannot produce on demand: electricity.
As models improve, the constraints increasingly extend into the physical world—power plants, batteries, transmission lines and data centers. A company can access powerful software almost instantly. Building the infrastructure to run it at scale takes years.
That is the investment thesis connecting this week’s multibillion-dollar federal energy announcements, Big Tech’s nuclear agreements and investors’ interest in solar: as intelligence becomes cheaper and more accessible, reliable power becomes more strategically valuable.
The opportunity reaches beyond powering chatbots. This month’s advances in AI-assisted drug research and mathematics offer a glimpse of what wider access to computing could support.
The grid is struggling to keep up
America’s power challenge is already visible in infrastructure planning.
In its 2025 Long-Term Reliability Assessment, the North American Electric Reliability Corporation found that 13 of 23 assessment areas faced resource-adequacy challenges over the following decade. The assessment covers the North American grid and warns that generation and transmission development are lagging projected demand from data centers and other large electricity users.
That does not mean every region is running out of electricity today. It means planned resources are not keeping pace with the requirements identified in those forecasts.
The challenge also extends beyond generating enough electricity over a year. Power must be available where customers need it, when they need it, with the infrastructure to deliver it reliably.
Investment can change that outlook. But money must translate into approved projects, equipment, skilled workers and completed construction.
Washington puts billions behind the buildout
On October 28, the Department of Energy closed a $5.1 billion loan package for Alliant Energy, supporting infrastructure in Iowa and Wisconsin.
The projects include natural gas generation, battery storage and improvements to existing facilities. DOE says the package supports more than 2.6 gigawatts of new or upgraded power capacity.
Alliant projects up to $1.2 billion in cumulative customer savings over the 30-year loans, alongside more than 1,600 construction jobs. Lower borrowing costs are part of the expected benefit.
Those savings are forecasts spanning decades. Access to financing depends on project progress, eligibility and loan terms, while proposed developments must go through regulatory review.
A day earlier, DOE announced a separate conditional commitment of up to $5 billion for Constellation Energy, targeting nuclear upgrades across Illinois, Maryland and Pennsylvania. The department says those improvements would add nearly 1.1 gigawatts of capacity. That financing has not yet reached final close.
Post from @ENERGYView the original post on X↗
Chamath’s solar bet puts speed at the center
Investors are approaching the power opportunity from several directions.
On the October 21 episode of Moonshots with Peter Diamandis, Chamath Palihapitiya described accumulating solar assets before moving into data centers.
His reasoning emphasized how quickly solar can be developed relative to other options. He also described the difficulties involved in bringing infrastructure online, including permitting, financing, supply chains and labor. Solar paired with batteries featured in his discussion of opportunities for more distributed generation.
His thesis is that persistent infrastructure constraints create investment opportunities. That is different from establishing an unavoidable, permanent electricity shortage—or guaranteeing returns for every energy business.
As Blockster’s recent AI infrastructure and markets coverage explored, strong demand is only part of the equation. Companies still need workable financing, competitive costs and the ability to deliver.
Meta, Google and Amazon are securing supply for years
Big Tech’s commitments show how seriously major electricity customers are taking the challenge.
In January, Meta announced agreements with Vistra, TerraPower and Oklo supporting up to 6.6 gigawatts of new and existing nuclear capacity by 2035. Those commitments followed its earlier agreement with Constellation.
The projects span operating plants, expansions and proposed advanced reactors. Some new developments have initial delivery targets in the 2030s, so the headline figure includes both existing supply and capacity still to be built.
Constellation says its recent 20-year agreements with Google and Amazon have encouraged more than $7 billion in planned investment, supporting approximately 1,080 megawatts of additional nuclear capacity. Its Meta agreement supports another 30-megawatt expansion at Clinton in Illinois.
These commitments give a power producer something valuable before construction begins: greater confidence that customers will buy the electricity.
“Our agreements with Google and Amazon demonstrate how long-term customer commitments can incentivize billions of dollars in investment in new, reliable nuclear capacity.”
— Joe Dominguez, Chairman, President and CEO, Constellation
Google’s October agreement supports 890 megawatts of new nuclear capacity and more than $4.3 billion of Constellation investment, with the first upgrade expected in 2028.
The work involves uprates—improvements that allow existing nuclear plants to generate more electricity, including upgrades to turbines, steam generators and control systems.
Constellation’s proposed federal loan could help finance its expansion effort, but still requires due diligence and definitive financing agreements. The conditional commitment does not constitute a final decision to proceed at every proposed facility.
Post from @ConstellationEGView the original post on X↗
What more computing power could make possible
This month’s scientific announcements give the infrastructure discussion a broader purpose.
On October 27, Anthropic reported early drug-development work with six organizations, including Bristol Myers Squibb, Generate Biomedicines and Enveda.
It also disclosed its acquisition of Mithril Medicine, which is developing potential treatments for two rare forms of muscular dystrophy. Anthropic said one program progressed from an idea to promising results in mice in less than 18 months. Human trials could begin as early as late 2028 if development succeeds.
These are early research results, not established cures. Anthropic says most of the work remains preclinical, meaning it has largely not reached testing in humans.
OpenAI’s October 6 mathematics release offers another example. Its public repository currently contains 721 mathematical manuscripts, grouped into related research families. OpenAI cautions that the results are at different stages of verification and some may contain issues. The manuscript count should not be interpreted as 721 independently confirmed solved problems.
These developments suggest how wider access to capable AI could support scientific work. Reliable, affordable computing is part of making those tools available to more researchers and organizations.
Money funds the buildout. Power makes it work.
The economic argument is becoming clearer: as AI makes intelligence more accessible, the ability to put it to work depends increasingly on physical resources.
Money can finance a power plant; it cannot make one appear overnight. Where electricity is constrained, access to usable power can matter more to a project’s immediate prospects than additional capital alone.
That could make energy one of the defining strategic assets of the AI economy. The opportunity extends across generation, storage, transmission and the equipment needed to connect them. Which businesses capture that value will depend on their costs, contracts and execution.
Blocksters, put energy on your investment radar. As intelligence becomes more abundant, the businesses supplying dependable, affordable power could help determine how much of its promise reaches the real economy.
Reporting by Sasha Steenkamp
