Amkor Lifts Arizona Packaging Investment to About $12 Billion as AI Demand Grows

Nvidia CEO Jensen Huang said the company could sell twice as many chips next year compared to this year, speaking during a speech in Scotland last Thursday. Nvidia just started shipping its next-generation data center system, Vera Rubin, which includes the Rubin GPU, the Vera central processing unit, and upgraded networking components, and can deliver up to 30 times more performance than the equivalent Blackwell Ultra system while cutting inference token costs by 97%. Huang said every frontier AI company planned to adopt Vera Rubin at launch, unlike Blackwell Ultra. Nvidia is on track to deliver around $411 billion in total revenue during its current 2027 fiscal year, based on Wall Street's average estimate, and management recently told shareholders to expect 70% revenue growth in fiscal 2028, implying almost $700 billion next year. Nvidia stock trades at a price-to-earnings ratio of 28.1, less than half its 10-year average of 61.4, though risks include roughly $175 billion of fiscal 2028 sales potentially funded by Nvidia's own balance sheet, about 25% of that year's revenue according to CFO Colette Kress.

Advanced Micro Devices reported Q2 FY2026 revenue of $11.54 billion, up 50.11% year over year, with non-GAAP EPS of $1.66 beating the $1.61 consensus, as Data Center revenue doubled 107% year over year to $6.72 billion and now represents 58% of total revenue. Management guided Q3 to roughly $13 billion, up about 41% year over year, and expects data center segment revenue to more than double year over year in 2027 on the back of Helios rack-scale systems, MI450 GPUs, and Venice EPYC CPUs. 24/7 Wall St. set a 12-month price target of $510.66 on the stock, which trades at $559.82 after a 254.5% one-year rally, implying roughly 8.78% downside, with a hold recommendation at 90% confidence. The firm's bull case above $616 rests on anchor customers including Anthropic, Meta, Microsoft Azure, and OpenAI, while its bear scenario points to $399.25 on risks including HBM memory constraints, U.S. export controls, and NVIDIA's CUDA moat. NVIDIA trades at a P/E of 45 with 75% non-GAAP gross margin and Data Center revenue of $89.02 billion, up 117% year over year, while Intel's DCAI segment grew 59% to $6.26 billion in Q2 2026.

Goldman Sachs expects U.S. hyperscalers to deploy $1.4T in capital in 2027, an estimate that sits above Wall Street consensus as artificial intelligence shifts from experimentation to implementation. The bank expects the AI infrastructure spending cycle to remain elevated through 2027, noting that computing demand continues to outpace supply and that much of the infrastructure under development has already been contracted. Eric Sheridan, a business unit leader at Goldman Sachs Research, said the move toward implementation was a key takeaway from the bank's Communacopia + Technology Conference in San Francisco earlier this month, where most companies presented distinct examples of moving from AI experiments to deploying AI in internal operations and workflows and externally with customers. Goldman Sachs also sees consumer AI agents moving beyond chat toward tasks such as shopping, travel bookings, and calendar management, with monetization expected through advertising and subscriptions while more complex tasks drive higher token usage. The bank said AI growth still faces constraints around memory chips, power, and land, and that lower token prices will be key to driving mass adoption.