Wall Street Analysts Say Chinese AI Rival Poses Less Threat to U.S. Chip Giants Than Markets Feared

When Chinese artificial intelligence startup DeepSeek released its R1 model in late January, it triggered one of the sharpest single-day selloffs in U.S. technology stocks in recent memory. Nvidia alone shed roughly $600 billion in market capitalization in a single session, a record loss for any company in stock market history. Broadcom, AMD, and a constellation of AI infrastructure suppliers fell in sympathy. The implicit fear was straightforward: if a Chinese lab could match American frontier AI performance at a fraction of the cost, the insatiable demand for expensive semiconductors and data center buildouts might evaporate.

That fear, according to a growing chorus of Wall Street analysts, is likely overblown. As reported by the Wall Street Journal, strategists at several major investment banks argue that DeepSeek’s emergence does not fundamentally alter the long-term capital expenditure cycle underpinning U.S. AI hardware demand. If anything, cheaper and more efficient models may accelerate the proliferation of AI applications, which would in turn require more compute, not less.

rows of server racks inside a large-scale data center facility, illuminated by cool blue overhead lighting with cable management visible along the floor

The Efficiency Paradox and What It Means for Chip Demand

The analytical crux of the bull case rests on what economists have long called the Jevons paradox, the counterintuitive principle that greater resource efficiency tends to increase total consumption rather than reduce it. Applied to AI, the argument holds that if inference becomes dramatically cheaper, enterprises and developers will deploy AI in far more applications, ultimately driving aggregate demand for chips higher over a multi-year horizon. Morgan Stanley and JPMorgan analysts have each noted in recent research that historical technology cycles support this reading, pointing to how declining memory prices in the 1990s expanded computing adoption rather than contracting the semiconductor market.

For Nvidia specifically, its H100 and Blackwell-series graphics processing units remain the dominant hardware for training large language models at scale, a segment DeepSeek has not disrupted in any material way. Training workloads, which require sustained parallel processing across thousands of GPUs simultaneously, are structurally different from inference efficiency gains. Broadcom, meanwhile, has benefited from hyperscaler demand for custom application-specific integrated circuits, a market that continues to expand as Google, Meta, and Amazon invest aggressively in proprietary silicon. Neither company’s core revenue driver has been directly challenged by DeepSeek’s software-level breakthroughs.

Investor Positioning and the Recovery in AI Equities

Markets have partially reflected the more tempered assessment. Nvidia shares recovered a meaningful portion of their post-DeepSeek losses within two weeks of the initial selloff, stabilizing above the $120 range as institutional buyers returned on dips. Options market data showed a notable decline in put volume relative to calls on Nvidia after the first week of February, suggesting hedging activity was unwinding as the panic faded. Broadcom similarly retraced toward its pre-shock levels, supported by its most recent quarterly earnings which showed continued strength in AI-related networking revenue.

Analysts caution, however, that the DeepSeek episode has introduced a new variable into the AI investment calculus: valuation risk tied to competitive disruption from unexpected quarters. Even if demand fundamentals remain intact, the speed and severity of January’s selloff demonstrated that AI-linked equities carry concentration risk that investors had perhaps underappreciated. Our earlier coverage of the DeepSeek market shock captured the initial damage in detail, while the broader question of stretched valuations across the technology sector has been a persistent concern, as examined in our analysis of Wall Street valuations.

exterior facade of a semiconductor fabrication facility at dusk, with industrial ventilation systems and clean-room signage visible on the building surface

Structural Tailwinds Remain Largely Intact

The broader infrastructure buildout shows few signs of deceleration. Microsoft has committed to spending $80 billion on AI data centers in fiscal 2025 alone, while Alphabet and Amazon have each signaled capital expenditure increases in their most recent earnings calls. These commitments represent binding financial obligations that will take years to play out, providing a durable demand floor for GPU clusters, high-bandwidth memory, and the networking equipment that connects them. Analysts note that even if DeepSeek-style efficiency improvements compress the cost per AI query, the scale of planned deployment more than offsets any reduction in hardware intensity per workload.

The geopolitical dimension adds a layer of complexity that purely financial models may not fully capture. U.S. export controls on advanced semiconductors to China remain in place and were tightened further in late 2024, limiting DeepSeek’s ability to access the most powerful Nvidia chips for future training runs. That structural constraint, rather than any particular technical breakthrough, may ultimately prove the binding limitation on Chinese AI ambitions at the frontier level. For now, the consensus view among professional investors appears to be that the U.S. AI infrastructure trade, though more volatile than once assumed, remains fundamentally intact.