The 'Hidden Cost' of the AI Boom is the Electricity Bill—Why the Trump Administration is Pressuring Big Tech for 15-Year Contracts
The generative AI (A.I.) boom is simultaneously upending four areas: power, semiconductors, capital, and regulation. This is a snapshot of that situation. The White House is intervening in rising electricity costs driven by data centers, while simultaneously attracting semiconductor investment through a trade framework with Taiwan. In the private market, AI coding companies are expanding rapidly, and the FTC (Federal Trade Commission) has begun monitoring 'acqui-hires.' While these may look like individual news items, they are actually connected by a single thread.
1. Who pays the 'AI electricity bill'?: The PJM 'emergency auction' concept
The core of this is that the White House and governors of northeastern states are urging PJM, one of the largest grid operators in the U.S., to facilitate an 'emergency power auction' that forces data centers (i.e., Big Tech) to bear the costs of new power generation. Reports suggest that companies could bid on 15-year contracts, potentially supporting approximately $15 billion in new power plant construction.
The phrase 'make the tech companies pay' used in the program was symbolic. What is important here is not the political slogan, but the design philosophy of 'creating funding predictability through long-term contracts to push power generation investment forward.' While PJM procurement has typically been done in short cycles, extending the duration supports 'price stability' and 'construction decision-making.' At the same time, PJM also released proposals to require large-scale consumers to have their own power sources or early curtailment during supply-demand tightness to respond to the rapid increase in AI data center demand, meaning practical solutions on the ground are also moving in parallel.
2. Semiconductors: 'Joint investment' over 'tariffs': The implications of the U.S.-Taiwan deal
Another major pillar is the trade agreement between the U.S. and Taiwan.Reports indicate a framework where, accompanied by Taiwanese investment in the U.S. (semiconductors, energy, AI, etc.), the tariff rate on certain goods from Taiwan would be set at 15%.
The program text mentioned observations such as 'TSMC building additional factories (multiple) in Arizona' and 'a total of $100 billion in additional investment,' but at the very least, the direction of 'investment in the U.S. accumulating with TSMC at the center' can be confirmed across multiple reports.
The point here is that semiconductors are being treated not as a geopolitical issue, but as a supply constraint (bottleneck) problem for AI. The line of AI data center expansion → power shortages → policy intervention, and the line of AI semiconductor production increase → investment framework → supply chain restructuring, are proceeding simultaneously.
3. The dual nature of the 'AI moment': Bottlenecks and capital heat (Replit)
The program discussed memory shortages as a 'bottleneck,' and while the stock market sees related companies benefiting, it also presented the realistic view that data center construction could be delayed by permitting and power procurement. This is precisely the 'AI moment of doubt'—a phase where demand is strong, but supply constraints are erupting in all directions.
Even so, private capital is hot. An example cited was the AI coding company Replit, which is reported to have raised funds (approximately $400 million) at a valuation of about $9 billion.
What is interesting is that, to borrow the program's words, Replit is aiming not just at 'professional developers,' but at enabling non-engineers to 'create prototypes.' If generative AI expands from 'writing' (text) to 'building' (apps), this will certainly become a massive market.
4. Regulation and 'Physical AI': FTC monitoring of acqui-hires, and humanoids
Finally, regulation. It is reported that FTC Chair Andrew Ferguson stated regarding the increase in recent years of acqui-hires (deals close to actual acquisitions through hiring + technology licensing) that he 'will look into whether this is an evasion of pre-merger review.'
The program's comment, 'No clever workarounds needed. The FTC will ensure a fair shake,' can be read as a sign of 'clarifying the scope of the rules' rather than 'intimidating' the parties involved in the transaction.
At the same time, AI is moving from software to 'bodies.' Barclays has mentioned a bullish scenario of up to $200 billion by 2035 in areas including humanoids.
The talk in the program about 'thousands of units operating in factories' and 'unit prices potentially dropping significantly' is a part where exaggeration is easily mixed in, but what is important for investment judgment is not the humanoid itself, but 'where the profit remains in the supply chain,' such as actuators, reducers, batteries, precision machining, sensors, and control software.
