A quiet but consequential transformation is underway in China’s automotive and autonomous vehicle industries. Across the country’s leading car manufacturers, robotaxi operators, and electric vehicle upstarts, a systematic effort to reduce dependence on Nvidia semiconductors is gaining momentum, and Wall Street is beginning to take note. According to reporting by CNBC, Goldman Sachs now believes that “the pivot to domestic chips will accelerate over 2026–28,” a forecast that reflects not just corporate strategy but the structural realities of an industry navigating export controls, supply chain nationalism, and the pressure to compete on cost.
The shift is not happening in isolation. It is the product of years of strategic investment, regulatory pressure, and hard lessons learned from supply disruptions, and it is now reaching an inflection point that will define the competitive landscape for autonomous mobility in China for the rest of the decade.
A Multi-Supplier Strategy Takes Shape
One of the clearest illustrations of this shift is Zelos Technology’s multi-supplier robotaxi chip strategy, which deliberately sources semiconductors from a range of suppliers rather than relying on any single vendor. For a company operating a fleet of 20,000 autonomous delivery vans and planning a major Hong Kong IPO, chip sourcing is not a peripheral concern, it is a core business risk. By building redundancy into its hardware architecture from the ground up, Zelostech is effectively insulating itself from the kind of supply shocks that have repeatedly rattled Chinese tech companies dependent on American components.
This multi-vendor philosophy represents a maturation of China’s approach to semiconductor risk. In earlier years, the response to export controls was often reactive, scrambling to find alternatives after restrictions had already been imposed. Now, companies are designing around vendor lock-in before it becomes a crisis. The lesson has been internalized: concentration on any single chip supplier, particularly one subject to Washington’s policy whims, is an unacceptable strategic liability.
BYD, Nio, and Xpeng Go In-House
If Zelos’ approach represents the diversification strategy, then BYD, Nio, and Xpeng represent the more aggressive play: full vertical integration through in-house semiconductor development. All three companies are now developing their own driver-assist chips, a move that would have seemed ambitious to the point of implausibility just five years ago.
BYD’s Xuanji A3 is perhaps the most striking example. As China’s first 4nm autonomous driving chip, it signals that BYD is no longer content to be a hardware assembler that sources brains from abroad. The company has committed to owning the full stack, from battery chemistry to silicon — and its chip ambitions are a direct extension of that philosophy. Paired with BYD’s pledge to cover accident costs when its Gods Eye driver-assist system is active, the company is making a high-conviction bet that domestic chip performance can meet the safety thresholds required for mass-market autonomous driving.
Nio and Xpeng are pursuing similar trajectories, each investing in proprietary silicon to reduce both cost and dependency. For these companies, in-house chips are not just about avoiding Nvidia, they are about achieving the kind of hardware-software integration that gives Tesla its performance edge in the United States. Owning the chip means owning the optimization loop, and in autonomous driving, that loop is where competitive advantage is built and sustained.
The Goldman Sachs Signal
Goldman Sachs’ forecast that domestic chip adoption will accelerate through 2026 to 2028 carries significant weight, both as a market signal and as a validation of the strategic bets these companies are making. For investors, it suggests that the window for Nvidia to recapture lost ground in China’s automotive sector is narrowing. For the Chinese companies themselves, it provides a degree of confidence that the domestic ecosystem, including foundries, EDA tools, chip designers, will mature quickly enough to support their ambitions.
The forecast also implies that the transition will not be painless. Domestic chips still lag Nvidia’s most advanced offerings in certain workloads. But the calculus has shifted: a chip that delivers 80 or 90 percent of the performance at a fraction of the geopolitical risk is increasingly seen as the better choice, particularly for automakers whose production volumes give them the leverage to demand customization and competitive pricing from domestic suppliers.
This dynamic is already visible in China’s broader AI infrastructure buildout. Alibaba’s launch of a 10,000-card AI computing cluster powered by its Zhenwu AI chips demonstrates that domestic alternatives are scaling fast, and the automotive sector is watching closely. Meanwhile, analysts have consistently noted that Chinese companies are finding ways to deliver datacenter-grade AI performance without Nvidia, a question that directly translates to the edge-computing demands of autonomous vehicles.
Geopolitical Pressure as Industrial Policy
The chip independence push in China’s automotive sector cannot be fully understood without acknowledging the role of U.S. export controls. Successive rounds of restrictions have targeted the most advanced Nvidia chips, effectively forcing Chinese companies to either pay steep premiums for smuggled components, as documented in a $2.67 billion federal crackdown on AI chip smuggling networks, or accelerate investment in domestic alternatives. For most legitimate companies, the latter is the only viable long-term path.
In this sense, Washington’s export control regime has functioned as an inadvertent subsidy to China’s domestic semiconductor industry. Every restriction that makes Nvidia hardware harder to procure raises the relative attractiveness of homegrown alternatives and increases the urgency of investment in domestic chip design and manufacturing capability. The ongoing tensions visible at events like Hannover Messe, where Chinese firms arrived with AI robots and advanced hardware to signal technological parity, underscore how deeply this industrial competition has become embedded in corporate strategy across sectors.
A Structural Shift, Not a Temporary Workaround
What distinguishes the current moment from previous cycles of China’s chip substitution efforts is the combination of scale, sophistication, and genuine commercial urgency. BYD is not developing driver-assist chips as a research project — it is doing so because it sells more electric vehicles than any company on earth and needs to control its cost structure and supply chain at scale. Zelostech is not diversifying its chip suppliers as a contingency plan — it is doing so because fleet economics demand it.
The Beijing Auto Show’s framing of intelligent driving as the real competitive frontier captures the stakes precisely. In a market where every major automaker is racing to ship Level 2 and Level 3 autonomy as standard features, semiconductor performance and availability are not secondary considerations, they are the primary determinants of product roadmaps. Companies that own their chips own their destiny.
Goldman Sachs’ 2026–28 acceleration forecast is not a prediction of a smooth transition. Supply chain maturation, chip design talent constraints, and the persistent performance gap with frontier American hardware will all create friction. But the direction of travel is clear, and it is not reversing. China’s automotive and autonomous vehicle industries are building the capability to compete without Nvidia, and increasingly, they are demonstrating they can.