Chipmaker Unveils 2nm Process, Reigniting the Foundry Race

A leading contract chipmaker unveiled its next-generation 2-nanometre manufacturing process on Friday, promising substantial gains in power efficiency and reigniting the high-stakes race among the world's most advanced foundries.

The company said the new process delivers roughly 25% better power efficiency, or alternatively about 15% higher performance at the same power, compared with the current 3-nanometre generation. Volume production is scheduled to begin next year, with the first chips expected to appear in premium smartphones and data-centre accelerators.

The announcement matters far beyond the specifications. Each new generation, or "node," of chip manufacturing requires tens of billions of dollars in investment and years of painstaking development, and only a shrinking handful of companies can still afford to compete at the leading edge. Success or failure at each node reshuffles the competitive order for years and determines which device makers gain access to the fastest, most efficient silicon.

The efficiency gains are particularly significant for the data-centre market, where the explosive growth of artificial-intelligence workloads has made power consumption a critical constraint. "At this point, performance per watt is the whole game," said Elena Voss, a semiconductor analyst at Meridian Strategic. "The customers building enormous AI clusters care about their electricity bills as much as their compute, and a 25% efficiency gain translates directly into money saved at scale."

The company's rivals are not standing still. At least two competitors have their own advanced nodes in development, and the gap between the leaders has narrowed in recent years as governments have poured subsidies into domestic manufacturing. Industry watchers expect a competing announcement within months, setting up an expensive battle for the most demanding — and most lucrative — customers.

The economics are brutal. The cost of building a leading-edge fab has roughly doubled with each recent generation, and the research and development required to make each node work has climbed just as steeply. That relentless cost escalation is why the number of companies able to compete at the frontier has dwindled from more than a dozen two decades ago to a mere handful today. "Every node is a bet-the-company decision now," Voss said. "The winners take an enormous share of the profits. The losers do not get many second chances."