The Billion-Dollar Chip: Who Can Afford to Design the Future?

U.S. semiconductor policy has focused heavily on restoring manufacturing capacity, but an equally important challenge is emerging upstream: the rapidly escalating cost of designing leading-edge chips. The cost of developing a leading-edge 2-nanometer chip has climbed to approximately $725 million, a 15-fold increase from the $48 million required to develop a 28 nm design in the late 2000s. As fewer firms and markets can support development costs at this scale, specialized chipmakers may be forced to recover those costs from relatively small customer bases, pushing prices higher for the industries that depend on their designs. Rising design costs could therefore make specialized computing increasingly unaffordable across parts of the broader U.S. technology ecosystem.

Greater concentration in semiconductor design is not inherently harmful. Higher development costs reward scale and can allow large firms to sustain the investments required at the technological frontier. The concern arises in specialized markets where those economies of scale are harder to achieve. Chips such as radiation-hardened processors for aerospace and defense or application-specific chips for telecommunications may remain necessary even when their markets are relatively small. Firms developing these chips must recover rising costs from fewer customers, potentially raising the cost of innovation across aerospace, defense, telecommunications, and industrial systems.

The Economics of Frontier Design Are Narrowing the Field

Advanced 2 nm chips can contain more than 50 billion transistors, requiring increasingly sophisticated verification, system integration, and post-silicon validation. As development costs approach three-quarters of a billion dollars, companies must either sell into very large markets or deploy chips at massive internal scale to justify frontier design. That raises the break-even threshold for each new chip project. The result is a narrowing funnel: startups, smaller fabless firms, and specialized chip designers face greater difficulty financing frontier-node projects; established firms become more selective about which designs they pursue; and the most advanced design work gravitates toward firms with the largest balance sheets, customer bases, or internal computing needs.

Hyperscalers illustrate this divide. Companies such as Google and Meta can justify custom AI chips because they deploy them at enormous scale across their own infrastructure. Their success demonstrates the advantages of scale rather than a decline in design costs. Specialized semiconductor firms face different economics: where their chips remain indispensable, they may survive as higher-cost providers that must pass more of their development expenses on to customers.

If frontier chip design becomes concentrated among a handful of vertically integrated firms, the broader innovation ecosystem could face a challenging dynamic. If their chips can meet most downstream needs, smaller design firms may exit the market; if specialized designs remain indispensable, those firms may survive as boutique providers but charge substantially higher prices to recover rising design costs from smaller customer bases. In the latter case, higher chip costs could make innovation prohibitively expensive for startups and specialized industries, narrowing the broader ecosystem that depends on access to advanced semiconductor technologies. 

Preserving Access to Advanced Design

The policy challenge is to preserve access to advanced design capabilities as the economics of the industry change. Scale can help firms manage rising development costs, but that model is harder to sustain in aerospace, defense, telecommunications, and other specialized markets where production volumes are smaller and purpose-built chips remain necessary.

Reducing the underlying cost of design offers a more practical response than trying to preserve a particular industry structure. Investments in precompetitive research, advanced Electronic Design Automation and AI-assisted design tools, shared testing and validation infrastructure, reusable IP, and engineering talent can make specialized projects more economically viable. Such measures would help ensure that rising frontier costs do not constrain innovation in the industries that depend on specialized computing.

These domestic pressures are compounded by an intensifying external dynamic. Faced with increasingly restrictive U.S. export controls, Chinese firms have accelerated investments in proprietary chip design — Huawei's Ascend processors, custom silicon from Alibaba, Baidu, and Tencent, and most recently DeepSeek's reported development of its own inference chip. China has backed these efforts with an estimated $100 billion in government support. Export controls can constrain China's near-term access to advanced American-designed chips, but they simultaneously increase China's incentives to develop domestic alternatives. As those alternatives improve, U.S. firms stand to lose access to the Chinese market — reducing the revenue that helps offset rising design costs. The same economics making frontier design harder to sustain domestically are also narrowing the commercial base that U.S. chipmakers can draw on globally. Addressing the cost of design is therefore not only an industrial policy question but an increasingly urgent one for long-term U.S. technological competitiveness.