Chip Stocks Slip on AI Jitters; Nvidia Eyes Japan Robotics

Semiconductor stocks extended their recent slide on Friday as investors turned risk-off following a new Chinese AI model release, even as fresh supply-chain signals from TSMC and Nvidia pointed to continued infrastructure buildout.

Key points

• Chip stocks sold off Friday after a Chinese startup released a powerful new AI model, reviving worries about demand sustainability at the high end of the GPU market.
• TSMC raised its capex and revenue forecast after reporting second-quarter results, citing growing AI chip demand.
• Nvidia’s Vera Rubin GPU rollout hit a thermal production snag, pushing initial shipments back by several weeks, though KeyBanc still projects 1.7M to 1.8M Rubin units for 2026 and raised its price target to $330.
• Nvidia CEO Jensen Huang announced partnerships with Japanese robotics firms Fanuc and Yaskawa Electric while visiting Tokyo, framing AI as the key to making robots “smart, easily adaptable and accessible.”
• Wedgewood Partners flagged concerns about Microsoft’s capital expenditure in its Q2 2026 investor letter, adding to a growing body of institutional scrutiny over Big Tech AI spending.

Why did chip stocks sell off again?

The proximate trigger on Friday was a new AI model release from a Chinese startup, which the market read as a fresh sign that capable models can be built with less compute than US hyperscalers are banking on. That narrative has surfaced before (most notably around DeepSeek earlier this year) and tends to hit GPU-exposed names hardest. Asian equities were already under pressure heading into the session, with semiconductor shares facing renewed selling pressure across the region and the Kospi leading declines.

The selloff sits in awkward tension with the fundamental data. TSMC just raised both its capital expenditure guidance and revenue forecast on the back of AI orders, following ASML’s own guidance raise earlier in the week. The gap between what chipmakers are signaling and what the market is pricing on any given day around a new model release has become a recurring feature of this cycle.

What is the Vera Rubin delay actually signaling?

Nvidia’s next-generation Vera Rubin GPU has hit a production snag tied to thermal management, pushing initial shipments back by several weeks. On the surface, any delay in a flagship product is unwelcome. The analyst reaction has been measured, though. KeyBanc maintained its full-year unit projection of 1.7M to 1.8M Rubin units and actually raised its price target to $330, suggesting the delay is viewed as a timing issue rather than a structural problem with yields or demand.

For investors, the more relevant question is whether a multi-week slip affects Nvidia’s fiscal calendar in a material way, or simply compresses the shipment curve into a later quarter. Given the backlog dynamics that have characterized every recent Nvidia product cycle, demand eating into a brief delay seems more likely than demand evaporating because of one.

What is Huang’s Japan push about?

Jensen Huang’s Tokyo visit this week went well beyond executive tourism. He met with CEOs of Kioxia and Tokyo Electron, two companies that sit at critical points in the semiconductor supply chain, memory and chipmaking equipment respectively. He then announced formal robotics partnerships with Fanuc and Yaskawa Electric, two of the world’s leading industrial automation firms.

Japan is not a peripheral market for Nvidia. The country has leading positions in chipmaking equipment, specialty materials, and industrial robotics, all of which feed into the AI infrastructure stack Nvidia sits atop. The robotics angle is also consistent with Nvidia’s longer-term push into physical AI, where its software platforms are being positioned as the operating layer for autonomous machines, not just data center accelerators.

Is Microsoft’s AI spending becoming a liability?

Wedgewood Partners, in its Q2 2026 investor letter, highlighted concerns about Microsoft’s capital expenditure trajectory. The firm’s Composite returned 9.4% net in the quarter versus 15.2% for the S&P 500 and 16.7% for the Russell 1000 Growth, a meaningful gap. The letter singles out MSFT capex as a point of contention, reflecting a broader institutional debate about whether the returns on AI infrastructure investment will justify the scale of spending being committed.

Microsoft is far from alone in facing this scrutiny, but as one of the largest AI capex spenders globally, its spending profile tends to set the tone for how investors think about the entire hyperscaler cohort. Whether elevated capex is a moat-building exercise or a margin headwind depends heavily on monetization timelines, a question that remains genuinely open.