a16z says data centers lower bills while tech capex hits $1 trillion
By Imani Sutton · Reporting from Atlanta ·
Podcast
The most staggering number from the latest a16z podcast isn’t a software valuation or a user count; it is a receipt. David George, Sarah Wang, Alex Immerman, and Santiago Rodriguez spent the episode, "AI, Infrastructure, and the Next Investment Cycle," laying out a future where just five companies—Alphabet, Amazon, Meta, Microsoft, and Oracle—are projected to spend over $1 trillion annually on capital expenditures by 2027. For context, David George noted that this buildout has already surpassed the railroad expansion as a percentage of U.S. GDP. We are no longer talking about code; we are talking about the physical re-engineering of the planet to support the "age of atoms."
The myth of the shared pole
On the podcast, Alex Immerman argued that this massive buildout of data centers might actually be a gift to the average person. He cited a study suggesting that for every 10% increase in data center capacity, residential electricity rates drop by 40 basis points. His logic is that the power grid is a "shared fixed cost" of poles, wires, and substations, and having a massive, stable customer like a Microsoft data center helps spread those costs across more units of electricity.
This is a classic utility company talking point, and as someone who spends her nights reading Georgia Power rate filings, I can tell you exactly where it breaks. The grid isn’t a static bucket of costs. When a hyperscaler drops a 500-megawatt load into a service territory, it doesn’t just "share" the existing poles; it forces the utility to build new high-voltage transmission lines and peaking gas plants to handle the surge. Under the current regulatory model, companies like Georgia Power earn a guaranteed profit—often around 10%—on every dollar they spend on that new infrastructure. The data center might pay for its own juice, but the "shared" cost of the expanded system gets baked into the rate base for everyone. We aren’t sharing the savings; we are co-signing the loan for their expansion.
The ninety-trillion-dollar pivot
Sarah Wang described this as a "new age of atoms," estimating that global infrastructure investment needs will hit $90 trillion through 2040. This includes the power, water, and roads required to keep the AI dream from overheating. She pointed to Meta’s site in Louisiana as an example of working with communities to lower costs. But we have to ask: who is the infrastructure for?
When these investors talk about "industrial booms," they are talking about companies like SpaceX and Tesla, where the "factory is the product." They are betting on a future where the physical world is as frictionless as a landing page. But as someone who lived through the June when the taps in Vine City ran dry under a heat advisory, I know that "infrastructure" is a zero-sum game. Every gallon of water used to cool a rack of H100s is a gallon diverted from the local water table. Every megawatt reserved for a "long-running agent task" is a megawatt that isn't available to keep a senior center cool during a heatwave. The a16z team sees a "surplus" for consumers; I see a massive transfer of public resources into private compute.
The rent in the machine
Santiago Rodriguez and David George also touched on the "renaissance" of private companies like Stripe and Data Bricks, which are staying private longer and reaching valuations in the trillions. They argued that being private allows these founders to take "bigger swings" with longer paybacks.
But for the rest of us, those swings look like a new kind of rent. Whether it’s Shopify using AI to nudge merchants into "retaining" on their platform or agents taking over the "discovery" process for travel and groceries, the goal is the same: to own the relationship between the person and the product. When Alex Immerman talks about agents making orders "without clicks," he is describing a world where the software makes the decision for you. If the algorithm is choosing your insurance or your groceries, the corporation behind that algorithm—be it Amazon or Instacart—is the one setting the price.
We are being told that this trillion-dollar buildout will make life cheaper and more efficient. But in the South, we know that when the power company builds a new plant, the bill goes up, not down. The "surplus" these investors celebrate is being funded by the very grid and water systems we rely on to survive. If you want to know who is really winning this investment cycle, don't look at the AI's output—look at your next power bill.