In October, researchers at Peking University announced a breakthrough in analog computing, a field that most of the tech world had abandoned decades ago. The news triggered a ripple across both academia and industry because analog systems behave less like today’s neat rows of digital 1s and 0s and more like the natural world: unpredictable, messy, continuous, and astonishingly efficient. At scale, they promise something digital chips have never delivered: staggering computational power that uses only a fraction of the electricity.
The United States has poured hundreds of billions into digital AI infrastructure led by hyperscalers and semiconductor giants. NVIDIA CEO Jensen Huang recently stirred the pot when he said that China could win the AI race because of electricity rather than silicon. He was pointing to a reality that is hard to ignore. AI models are hungrier than ever. Data centers are running out of power. Nations are rationing energy to keep training clusters alive. The world’s digital future depends on a resource that cannot be upgraded with a software patch.
This is where China’s analog surprise becomes more than a scientific curiosity. It becomes a geopolitical spark. A country that already dominates solar panel production, grid-scale battery deployment, and rare-earth supply chains now appears to be investing in a technology that cuts power consumption by orders of magnitude. If the research holds up, the global AI competition will no longer be fought only with faster chips. It will be fought with the physics of efficiency and the politics of energy.
Analog computing is older than silicon itself. Early systems used mechanical gears, fluids, or continuous voltages to solve equations. They were brilliant in theory and disastrous in practice. Heat made them drift. Noise made them unstable. Temperature changes could throw off entire calculations. As digital electronics matured, engineers abandoned analog machines for good.
Yet AI has reopened the door. Neural networks behave much more like analog signals than digital spreadsheets. When an AI model adjusts a weight between two neurons, it is essentially fine-tuning a continuous value, not flipping a switch. Digital chips use enormous amounts of power to simulate these continuous processes with billions of discrete operations. Analog circuits, by contrast, compute them natively. Instead of representing a number with millions of transistors, you can let a tiny current or voltage do the work.
The promise is huge: less energy, more speed, smaller chips. But the nightmare of instability has never gone away. What Peking University claims to have achieved is a new architecture that suppresses noise and drift through layered error correction built directly into the hardware. Their experimental results suggest stable, repeatable analog computations that run as much as 100 times faster and with dramatically lower power usage. If verified, this makes digital AI accelerators look heavy and wasteful, like forcing an electric SUV to compete with a bicycle on battery life.
When Jensen Huang says China will win the AI race because of electricity, he is not exaggerating. China controls more of the world’s renewable power production than any other country. It has built data centers next to hydroelectric stations, solar megafarms, and wind superclusters. The Chinese grid may be sprawling and complex, but it is expanding fast enough to keep feeding AI clusters with comparatively cheap electricity.
The United States is facing a very different landscape. Data center demand is so high that utilities are turning away new sites. Cities are clashing with tech firms over grid strain. Entire states are forecasting power scarcity by the end of the decade. Even if the US builds more chips, it may not have the electricity to use them.
Analog computing flips the script. Instead of building more data centers to feed AI, you build smarter chips that sip power instead of guzzling it. Energy shifts from bottleneck to strategic advantage. This is the geopolitical tension inside the Beijing breakthrough. China might achieve both chip efficiency and the energy supply to run AI at scale.
That is not just an engineering victory. It is an industrial one. The country that controls the global supply chain for low-power AI wins the economic narrative for the next century.
Imagine you are at a luxury resort pool. Two people are racing from end to end. One swims naturally, gliding through the water in smooth strokes. The other climbs out every two seconds to run a few feet, then dives back in again, repeating the cycle inefficiently. Digital AI is the awkward runner-diver. It executes countless discrete actions to mimic a continuous process. Analog AI is the swimmer who never leaves the water.
This analogy is not perfect, but it captures why analog systems can be faster and more efficient. Their operations are inherently fluid. They do not simulate analog behavior. They are analog behavior. Digital chips excel at precision, reliability, and error-free logic. Analog chips excel at processing massive streams of continuous information with almost no overhead. AI has grown so large that overhead is now the industry's highest cost.
Analog computing offers a path where massive AI becomes both sustainable and affordable. If digital AI is the penthouse suite with a breathtaking view and a staggering electric bill, analog AI is the architect who reengineers the entire building so the elevator does not need a power plant of its own.
The Peking University announcement is not a commercial product. It is early research. But innovation often begins in quiet labs before disrupting global markets. China has a long record of taking niche scientific breakthroughs and industrializing them quickly. The US has an equally long record of dominating the high-performance computing space but struggling to maintain manufacturing leadership.
If analog AI becomes the new standard for low-power computing, the nations that adopt it fastest will have a dramatic advantage. Not just in AI speed, but in energy policy, national security, manufacturing strategy, and digital infrastructure.
Inspired by what you read?
Get more stories like this—plus exclusive guides and resident recommendations—delivered to your inbox. Subscribe to our exclusive newsletter
Resident may include affiliate links or sponsored content in our features. These partnerships support our publication and allow us to continue sharing stories and recommendations with our readers.