Morgan Stanley expects that at the upcoming Computex exhibition,$NVIDIA (NVDA.US)$CEO Jensen Huang will comprehensively showcase NVIDIA’s next-generation AI computing architecture centered on the Rubin GPU and Vera CPU, further solidifying its dominance in AI infrastructure.
From June 2 to 5, COMPUTEX TAIPEI 2026 will be held under the theme “AI Together,” with market attention rapidly shifting toward NVIDIA's rack-scale system designs and advancements in advanced packaging.
In its latest research report, Morgan Stanley forecasts that NVIDIA will unveil an AI factory solution designed to achieve the lowest cost per token—an initiative that is directly prompting a reassessment of supply chain capacity and generating significant growth expectations for related hardware manufacturers.
Driven by robust demand for AI GPUs and CPUs,$Taiwan Semiconductor (TSM.US)$Taiwan Semiconductor is accelerating the expansion of its CoWoS advanced packaging capacity. Morgan Stanley has significantly raised its forecast for Taiwan Semiconductor’s CoWoS capacity in 2027 and maintains a bullish outlook on investment opportunities across NVIDIA’s core supply chain.
As the exhibition approaches, investors are closely monitoring Rubin chip power consumption management solutions and the commercialization potential of the Vera CPU—technical details that will shape global AI computing capital expenditure trends and competitive dynamics over the coming years.
Rubin Rack Design and Power Consumption Solutions
In their report, Morgan Stanley analysts including Charlie Chan noted that a key highlight of COMPUTEX will be the server rack design for the Rubin series. NVIDIA is expected to demonstrate a combination of NVL, LPU, Vera CPU, BlueField, and NVSwitch racks, aiming to deliver the lowest cost per token for future 'AI factories.'
Addressing market concerns over whether the Rubin GPU can meet its 2.3kW thermal design power (TDP) target, Morgan Stanley revealed that Rubin has reverted to its original chip cooling solution and achieved a TDP of 1.8kW. However, the 2.3kW target remains attainable through coordinated thermal management at the server system level.
Regarding packaging design, NVIDIA will need to adopt a new photomask in 2027 for the Rubin Ultra variant to incorporate HBM4e memory. To maintain favorable chip yields and address warpage issues, Rubin Ultra is expected to retain a two-die-per-package configuration rather than the previously rumored four-die design.
Vera CPU Unlocks a $20 Billion Market Opportunity
Beyond GPUs, NVIDIA's expansion into CPUs has also drawn significant attention. Joe Moore, Morgan Stanley’s U.S. semiconductor analyst, expects NVIDIA to disclose more details about its Vera CPU at the upcoming event, representing a market opportunity worth up to $20 billion.
Supply chain research indicates that Taiwan Semiconductor is allocating additional CoWoS-R capacity and 3-nanometer wafers for the Vera CPU. Based on current capacity planning, Morgan Stanley estimates a reasonable shipment assumption of 1.5 million standalone Vera CPUs.
The firm expects the end customers for the Vera CPU to include$Microsoft (MSFT.US)$、 $Meta Platforms (META.US)$ 、$CoreWeave (CRWV.US)$and$Oracle (ORCL.US)$major technology companies and cloud service providers. The launch of this product will further complete NVIDIA’s closed-loop computing ecosystem, posing direct competitive pressure on traditional CPU manufacturers.
Taiwan Semiconductor’s CoWoS capacity forecast significantly revised upward
Strong demand for AI chips is reshaping the landscape of upstream foundry and packaging markets. In its report, Morgan Stanley substantially raised its forecast for Taiwan Semiconductor’s CoWoS capacity in 2027 from 160,000–170,000 wafers per month to 200,000 wafers per month.
To meet surging order volumes, Taiwan Semiconductor is reallocating 28/22-nanometer production space at its Fab 15A facility to produce 55-nanometer interposers, adding approximately 10,000 wafers per month in new capacity this year. Due to this capacity crowding-out effect,$United Microelectronics (UMC.US)$is expected to secure overflow orders for 28/22-nanometer image signal processors (ISPs) from$Sony Group (6758.JP)$in 2027, prompting Morgan Stanley to maintain its “overweight” rating on United Microelectronics Corporation (UMC).
Regarding System-on-Integrated-Chips (SoIC) capacity, although Morgan Stanley slightly adjusted its capacity build-out expectations for 2027 and 2028 to 40,000 and 70,000 wafers per month respectively, its actual output forecasts remain unchanged at 30,000 and 60,000 wafers per month. Major clients of SoIC technology include Apple, AMD, and NVIDIA.
Based on optimistic expectations for the Rubin and Vera product lines, Morgan Stanley maintains an “overweight” rating on NVIDIA’s core supply chain.
Looking to pick stocks or diagnose stock performance? Want to know the opportunities and risks in your portfolio? For all investment-related questions,just ask Futubull AI!
Editor /rice