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.

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.

The progression from early neural networks to today's massive models is fundamentally driven by the exponential increase in available computational power, from the initial move to GPUs to today's million-fold increases in training capacity on a single model.

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.

The narrative of energy being a hard cap on AI's growth is largely overstated. AI labs treat energy as a solvable cost problem, not an insurmountable barrier. They willingly pay significant premiums for faster, non-traditional power solutions because these extra costs are negligible compared to the massive expense of GPUs.

The plateauing performance-per-watt of GPUs suggests that simply scaling current matrix multiplication-heavy architectures is unsustainable. This hardware limitation may necessitate research into new computational primitives and neural network designs built for large-scale distributed systems, not single devices.

The International Energy Agency projects global data center electricity use will reach 945 TWH by 2030. This staggering figure is almost twice the current annual consumption of an industrialized nation like Germany, highlighting an unprecedented energy demand from a single tech sector and making energy the primary bottleneck for AI growth.

Model architecture decisions directly impact inference performance. AI company Zyphra pre-selects target hardware and then chooses model parameters—such as a hidden dimension with many powers of two—to align with how GPUs split up workloads, maximizing efficiency from day one.

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).