The “Silicon-Steel” Pendulum
For the last five years, the California vs. Texas narrative has been the most boring cliché in tech. It has been framed as a zero-sum migration from taxes and regulation toward a land of freedom and affordable housing. It was a debate about lifestyle and tech culture, when it should have been a debate about the architecture of progress.
But as the AI revolution moves from demo to industrial scaling, the relationship between Silicon Valley and Texas is undergoing a profound structural shift.
We are witnessing the birth of a new tech bipolarity. It isn’t a competition; it’s a pendulum.
On one side, you have the Intelligence Layer: San Francisco and the Peninsula. This is the world’s greatest concentration of “human compute” talent: the researchers, the model architects, and the product visionaries.
On the other side, you have the Physical Layer: Austin and the Texas triangle. This is the epicenter of surveying and steel: the energy, land, manufacturing, and massive-scale infrastructure required to house the millions of NVIDIA H100s that will run the world’s intelligence.
The next decade won’t be defined by which state wins. I believe it will be defined by the pendulum swings between them. As we scale AI, we will oscillate between two hard constraints: the limit of our ideas and the limit of our physical world.
In a world where we are compute and energy-constrained (or NIMBY), we need more Texas. In a world where GPUs sit idle and we lack the next AI application to drive consumption demand for them, we need more Silicon Valley.
The winners of the next decade won’t be the ones who entrench themselves in one camp. They will be the dual-citizens who bridge the gap between the Presidio and the Permian Basin.
The Intelligence Layer: Why SF is the Unrivaled Lab
Every few years, people try to declare Silicon Valley dead. And every few years, a new “God-model” is released, and the gravity of the world’s talent collapses back onto a five-mile radius in San Francisco.
If you are building the next foundational model, or the application layer that will define how 500 million people interact with AI, you have to be in the room. You need to be where the gossip about the next training run is happening in the Hayes Valley coffee shops. You need the density of the Stanford-to-OpenAI-to-Anthropic pipeline.
Silicon Valley is the epicenter of Algorithmic Efficiency. When compute is expensive, the value of the “better idea” skyrockets. SV is the place that finds a way to get 10x the performance out of the same number of parameters. It is the world’s premier lab for software and model-side scaling.
But labs have limits. You can have the most brilliant researchers in the world, but if they have nowhere to plug in their machines, nor the energy to power them, their brilliance remains theoretical.
The Physical Layer: Why Texas is the Essential Factory
For decades, Northern Virginia (Data Center Alley) was the undisputed king of the physical internet. But we are hitting a wall. Power grids in the Mid-Atlantic are straining under the weight of AI’s insatiable hunger. The “Stargate” era of $100 billion supercomputers requires something Virginia, and certainly California, cannot provide at the necessary scale: an absolute abundance of land and a deregulated, high-capacity energy market.
Enter Texas.
Texas is currently on a trajectory to pass Northern Virginia for the highest data center density in the world. Cheap land isn’t the reason, it’s about Velocity of Infrastructure.
In the AI era, “Time to Power” is the most important metric for a scaling startup. In speaking with an executive at a hyper-scaler recently, he remarked that across every single input from site selection through construction, installation and equipment commissioning, the new most important metric: Time to Power.
If you have
in venture funding to build a cluster, and California tells you the grid hookup will take four years, you are dead before you even start. Texas and ERCOT have a culture and growing track record of build first, ask for forgiveness later. It is the only place that can keep pace with the exponential growth of GPU clusters.
Texas is the land of Gigawatts. It is the place where the abstract “cloud” becomes a physical reality of concrete, copper, transformers and liquid cooling units . If Silicon Valley is the brain, Texas is the body. And right now, the brain is growing faster than the body can support.
The Scaling Pendulum: A New Cycle of Innovation
To understand the future of tech, you have to watch the pendulum.
Phase 1: The Intelligence Bottleneck (Need more SF).
In this phase, we have plenty of compute, but the models are dumb. We are looking for the next transformative breakthrough. We need more researchers, more vibes, and more experimental AI Agent applications. During this phase, the energy and power of Texas are underutilized. The GPUs sit idle because we don’t know what to train next.
Phase 2: The Physical Bottleneck (Need more Texas).
Once a breakthrough occurs, say Claude Code (or perhaps Mythos one day), we suddenly know exactly what to do. We just need to do it 100x bigger. Suddenly, the bottleneck isn’t the idea or AI application; it’s the physical world. We need 500,000 more GPUs. We need a dozen SMRs worth of power. We need a million square feet of climate-controlled space. We need fast interconnect and permitting. We need local community partnership.
In Phase 2, the center of gravity shifts to Texas. The most important person in the company isn’t the ML researcher, it’s the person who can secure the power purchase agreement (PPA) and oversee the construction of a GW-scale campus.
For the next decade, tech will oscillate between these two bottlenecks. Every time SF solves a consumption problem, it creates a hardware problem for Texas to solve. Every time Texas builds a massive new cluster, it creates a surplus of compute that requires SF to find a more creative use for it.
Elon, as he so often is, was one of the first to this realization. He realized that you could design the world’s most advanced “software on wheels” in California, but when it came time to build the “machine that builds the machines,” he needed the industrial scale of Giga Texas.
It’s no surprise then that Giga Texas, Terafab, SpaceX Bastrop (Starlink user terminals), Boring Company and more (prediction: Colossus 2.0?) are all Austin and central-Texas based.
The Structural Triumph: Texas as the Counter-Weight
The most interesting part of this shift is the cultural shift.
Ever since I began in tech and venture, every city in the world tried to be “The Silicon Valley of [X].” They tried to compete with California on talent, venture capital, and innovation. Nearly all failed because you can’t easily replicate the 70-year compounding head-start of the Stanford ecosystem and Sand Hill Road.
Texas is the first place to credibly position itself as a counter-weight to Silicon Valley, rather than a competitor. It isn’t trying, and doesn’t have to be, more innovative. It solves the physical bottleneck that AI requires.
This is a massive shift in the world’s power structure. For the first time, everyone from the seed stage founder in a Palo Alto dorm to the GP at a multi-stage VC firm in Menlo Park realizes they need the same thing.
They all need Texas.
The Re-Industrialization of Intelligence
We are moving out of the “incremental software” era, the era of apps, commerce and social media, and entering an era of Industrial Intelligence.
This era looks more like the 19th Century oil boom or the mid-20th Century aerospace race than it looks like the Web2 era. It is capital, energy, and land intensive.
The pendulum will continue to swing, creating massive opportunities for those who can navigate the tension between the two.
The “Silicon-Steel” pendulum is the new frontier.
