The performance gains from Nvidia's Hopper to Blackwell GPUs come from increased size and power, not efficiency. This signals a potential scaling limit, creating an opportunity for radically new hardware primitives and neural network architectures beyond today's matrix-multiplication-centric models.

From a first-principles perspective, space is the ideal location for data centers. It offers free, constant solar power (6x more irradiance) and free cooling via radiators facing deep space. This eliminates the two biggest terrestrial constraints and costs, making it a profound long-term shift for AI infrastructure.

Digital computing, the standard for 80 years, is too power-hungry for scalable AI. Unconventional AI's Naveen Rao is betting on analog computing, which uses physics to perform calculations, as a more energy-efficient substrate for the unique demands of intelligent, stochastic workloads.

While solar panels are inexpensive, the total system cost to achieve 100% reliable, 24/7 coverage is massive. These "hidden costs"—enormous battery storage, transmission build-outs, and grid complexity—make the final price of a full solution comparable to nuclear. This is why hyperscalers are actively pursuing nuclear for their data centers.

The massive energy consumption of AI has made tech giants the most powerful force advocating for new power sources. Their commercial pressure is finally overcoming decades of regulatory inertia around nuclear energy, driving rapid development and deployment of new reactor technologies to meet their insatiable demand.

When power (watts) is the primary constraint for data centers, the total cost of compute becomes secondary. The crucial metric is performance-per-watt. This gives a massive pricing advantage to the most efficient chipmakers, as customers will pay anything for hardware that maximizes output from their limited power budget.

A "software-only singularity," where AI recursively improves itself, is unlikely. Progress is fundamentally tied to large-scale, costly physical experiments (i.e., compute). The massive spending on experimental compute over pure researcher salaries indicates that physical experimentation, not just algorithms, remains the primary driver of breakthroughs.

For years, access to compute was the primary bottleneck in AI development. Now, as public web data is largely exhausted, the limiting factor is access to high-quality, proprietary data from enterprises and human experts. This shifts the focus from building massive infrastructure to forming data partnerships and expertise.

Beyond the well-known semiconductor race, the AI competition is shifting to energy. China's massive, cheaper electricity production is a significant, often overlooked strategic advantage. This redefines the AI landscape, suggesting that superiority in atoms (energy) may become as crucial as superiority in bytes (algorithms and chips).