Nvidia’s $6.2 Billion Japan AI Factory and Gulf Export Coup Mask a Rotation-Fueled Pullback

Nvidia lost $197B in market cap temporarily as AI efficiency fears rose, but secured a $6.2B Japan robotics project and UAE export upgrade for advanced chips.

Nvidia may have lost the crown as the world’s most valuable company for a day, but it spent the same week laying groundwork for two major growth corridors. A $6.2 billion state-backed artificial intelligence project in Japan and the unexpected opening of the Gulf market for its most advanced chips point to demand that is anything but fading — even as a rotation out of pure infrastructure plays knocked the stock back by nearly 4% over the week.

On the corporate front, chief executive Jensen Huang jetted to Tokyo to unveil a partnership with the Japanese AI firm Noetra. The pair plan to build what they describe as the world’s first national infrastructure dedicated to physical AI and robotics. Running under the Japanese Ministry of Economy, Trade and Industry’s “FRONTia” programme, the facility will be powered by 27,500 GPUs from the upcoming Rubin generation and is backed by a five-year government commitment worth roughly $6.2 billion. A consortium of 44 Japanese heavyweights — including SoftBank, Sony, Honda and NEC — has signed on to develop localised, multimodal AI models for industrial use.

At the same time, the U.S. Department of Commerce handed the chipmaker a regulatory surprise. On July 10, 2026, the United Arab Emirates was upgraded to the highest trust tier for technology exports, meaning buyers there can now acquire advanced Blackwell processors without requiring individual shipment approvals. Analysts had long viewed export restrictions on cutting-edge silicon as a key risk to Nvidia’s valuation; the easing should accelerate data centre build-out across the Gulf region.

Yet the shares themselves struggled against a broader market current. Nvidia lost roughly $197 billion in market capitalisation in a single session last week as the Chinese-language AI model Kimi K3 rekindled jitters about whether ever-larger compute clusters are truly necessary. Developed by Moonshot AI, K3 is a 2.8 trillion-parameter model built on a mixture-of-experts architecture and claims 2.5 times the scaling efficiency of its predecessor. Although its full weights are not due until the end of July, the mere announcement was enough to send the Philadelphia Semiconductor Index down roughly 10% for the week and more than 20% from its late-June record. Investors rotated capital toward consumer-facing AI applications, briefly vaulting Apple past Nvidia as the world’s largest company by market cap on July 17 — Apple at roughly $4.9 trillion versus Nvidia’s $4.86 trillion.

Huang pushed back against the narrative of waning demand. He denied speculation of a delay to the next-generation Rubin Ultra chips, insisting the systems are in production and will ship this year. He also argued that application-specific integrated circuits do not pose a zero-sum threat to Nvidia’s general-purpose GPUs because of the CUDA ecosystem. The company expects the combined Blackwell and Rubin chip families to generate roughly $1 trillion in revenue across 2026 and 2027. Growth drivers, Huang said, include AI labs, hyperscalers, sovereign AI programmes and industrial applications. The latest quarter underscored the operational momentum: revenue jumped 85% to $81.6 billion, with the data-centre unit surging 92% to $75.2 billion.

The analyst community remains divided. New Street cut its price target to $340 from $343 and downgraded the stock after it had already fallen to $215.33. The consensus, however, still rates Nvidia a “buy” with a median target between $303 and $308.50. A more optimistic scenario, calculated on a projected fiscal 2029 earnings per share of $16.06 and a 25.5-times multiple, yields a target of $409. The stock currently trades at roughly 16 times expected 2028 earnings, well below that bullish assumption.

Insider activity tells a cautious tale. Over the past six months, 70 selling transactions were logged against zero purchases by executives. Major sellers included director Mark Stevens (2.1 million shares), Ajay Puri (800,000) and finance chief Colette Kress (110,000). On the institutional side, 2,736 buyers faced 2,341 sellers; J. Stern trimmed its position by 125 million shares while FMR added 22.8 million.

Technically, the shares remain within a neutral zone. The relative-strength index sits at 48.8, signalling neither overbought nor oversold. The stock closed Friday at €177.46 in Frankfurt, 2.14% lower on the day and 12.37% below its 52-week high of €202.50. It now trades just under the 50-day moving average of €181.83 but still 7.33% above the 200-day line of €165.34, suggesting the longer-term uptrend has not yet broken.

The immediate focus shifts to this week. SIGGRAPH 2026 kicked off in Los Angeles on Monday, with Nvidia’s keynote set to spotlight neural rendering, world models and generative simulation. Investors will be watching closely for how convincingly the company can communicate the transition to the Rubin architecture. Further down the calendar, earnings from Alphabet on July 22 and Intel on July 23 will offer a temperature check on the broader tech landscape before Nvidia reports its own results on August 26.