Silicon Valley's Ohio power grab exposes a speculative buildout
By Ray Dombrowski · Reporting from Youngstown ·
As John Coogan and Jordi Hays detail gigawatt deals in Ohio, depreciating hardware shows Silicon Valley building speculative capacity faster than durable payrolls can justify.
When news broke that artificial intelligence developers are locking down power grid commitments measured in gigawatts rather than megawatts, the real-world scale of the current tech buildout finally became clear. On the podcast TBPN, hosts John Coogan and Jordi Hays detailed OpenAI's search for utility capacity. The firm is expanding into custom Tensor Processing Units and third-party data center leases. Coogan cited industry analysis showing OpenAI making a multiphase commitment in Ohio backstopped by Nvidia. Meanwhile, Anthropic is securing 1.7 gigawatts globally. Apple launched a touchscreen MacBook Pro where the biggest crowd applause went to a $2 backlit logo. Software startup Instinct saw its private valuation quoted at $30 billion without a standalone consumer application.
The Power Grid Meets the Balance Sheet
From a labor market analyst's perspective, gigawatts of electrical capacity in Ohio are not abstract software metrics. Power line commitments mean concrete footings, substation builds, high-voltage transformers, and physical utility infrastructure. Coogan and Hays noted OpenAI’s move toward Google’s custom chips and massive leasing commitments. The AI race has turned into an aggressive infrastructure land grab. Every major software lab is attempting to lock up available electrical capacity and chip architecture. They are taking Nvidia GPUs and Google TPUs regardless of past corporate rivalries.
Paper Valuations Versus Seven-Year Hardware
Tech investors argue that securing compute capacity early is necessary. It prevents getting locked out of the next generation of artificial intelligence models like Google’s Genie 4 or Gemini 4 Carbon. If software capabilities advance rapidly, spending billions on long-term power contracts and data centers looks like prudent corporate defense.
But the physical numbers tell a starker story about long-term hardware depreciation. On the show, Hays and Coogan highlighted a supercomputer listed on eBay containing Nvidia A100 chips for $60 million. That system originally cost nearly $300 million. Meanwhile, software upstarts like Instinct are being valued at $30 billion in private deals. That valuation grew in less than a year, despite lacking an app in store rankings. When physical hardware depreciates while pre-revenue software ideas command massive valuations, market pricing trades speculation ahead of operating yields.
Paying Premium Prices for Backlit Logos
That same pricing pattern shows up in consumer hardware. Coogan described attending Apple’s event where the company presented touchscreen MacBook Pros. Apple omitted prices from the main screen during the presentation, cushioning rising component costs behind a lease offer. Yet as Coogan observed, the loudest applause went to Apple reintroducing the glowing logo on the back of the laptop. That addition is estimated to cost roughly $2 in parts.
When tech firms ask customers to lease premium laptops while private investors mark up unproven software ventures, the underlying economy eventually demands an audit. Power commitments in Ohio will generate construction payrolls while facilities go up. Once the concrete cures, a data center running depreciating chips is simply a heavy load on the regional grid. Silicon Valley must translate gigawatts and valuation markups into durable payrolls that outlast a seven-year hardware cycle. Otherwise, the bill for this buildout will land on the local utility ratepayer's monthly statement.