Nvidia Computex Preview: What AI Computing "Nukes" Will Jensen Huang Bring?

```

Morgan Stanley predicts that at the upcoming Computex exhibition, NVIDIA CEO Jensen Huang will comprehensively showcase a new generation AI computing architecture centering on the Rubin GPU and Vera CPU, further consolidating the company’s dominance in the AI infrastructure sector.

According to Chasewind Trading Desk, from June 2 to June 5, the Taipei International Computer Show (COMPUTEX 2026) will be held with the theme "AI together". Market attention is rapidly turning to NVIDIA's rack-level system design and advanced packaging updates.

Morgan Stanley’s latest research report predicts that NVIDIA will present AI factory solutions aimed at achieving the lowest per-token cost. This development directly drives a reevaluation of supply chain capacity and brings significant growth expectations for related hardware manufacturers.

Driven by strong demand for AI GPUs and CPUs, TSMC is accelerating the expansion of its CoWoS advanced packaging capacity. Morgan Stanley has significantly raised its expectations for TSMC's CoWoS capacity in 2027 and is fully optimistic about the investment prospects of NVIDIA’s core supply chain.

With the exhibition approaching, investors are closely watching Rubin chip’s power control solutions and the commercial potential of the Vera CPU; these technical details will determine the trends in capital expenditures and competition in the global AI computing market for the coming years.

Rubin Rack Design and Power Solutions

Morgan Stanley analysts Charlie Chan and others pointed out in the report that one of the core highlights of Computex is the Rubin series server rack design. NVIDIA is expected to showcase the combination of NVL, LPU, Vera CPU, BlueField, and NVSwitch racks, aiming to provide the lowest per-token cost for future "AI Factories."

Regarding market concerns about whether Rubin GPUs can achieve the 2.3kW thermal design power (TDP) target, Morgan Stanley revealed that Rubin has reverted to its original chip cooling solution and achieved a 1.8kW TDP. However, with the synergy of server system cooling, the 2.3kW target can still be reached.

In terms of packaging design, for the Rubin Ultra version, NVIDIA will need to adopt a new mask in 2027 to introduce HBM4e memory. To maintain good chip yield and solve warpage problems, Rubin Ultra is expected to maintain a two-die per package design, instead of the previous rumor of four dies.

Vera CPU Unlocks $20 Billion Market Opportunity

In addition to GPUs, NVIDIA’s expansion in the CPU field is also under the spotlight. Morgan Stanley US semiconductor analyst Joe Moore predicts that NVIDIA will reveal more details about the Vera CPU at the show, signifying a market opportunity of up to $20 billion.

Supply chain research shows that TSMC is allocating extra CoWoS-R capacity and 3nm wafers for the Vera CPU. Based on current capacity plans, Morgan Stanley estimates a reasonable shipment assumption for standalone Vera CPUs is 1.5 million units.

The bank expects that final customers for the Vera CPU will include major tech companies and cloud service providers such as Microsoft, Meta, CoreWeave, and Oracle. The launch of this product will further complete NVIDIA’s computing ecosystem and directly compete with traditional CPU manufacturers.

TSMC CoWoS Capacity Forecast Significantly Raised

Strong demand for AI chips is reshaping the landscape of the upstream foundry and packaging market. Morgan Stanley has raised its forecast for TSMC's CoWoS capacity in 2027 from 160,000-170,000 wafers per month to 200,000 wafers per month.

To meet the surging demand, TSMC is converting its Fab 15A 28/22nm capacity to 55nm interposer production, expected to add about 10,000 wafers/month this year. Because of this capacity reallocation, UMC is likely to take on overflow 28/22nm image signal processor (ISP) orders from Sony in 2027, so Morgan Stanley maintains an “overweight” rating for UMC.

For SoIC (System-on-Integrated-Chip) capacity, although Morgan Stanley has slightly adjusted the capacity construction forecasts for 2027 and 2028 to 40,000 wafers/month and 70,000 wafers/month respectively, actual output expectations remain unchanged at 30,000 wafers/month and 60,000 wafers/month. Apple, AMD, and NVIDIA are all key SoIC technology customers.

Based on optimistic expectations for the Rubin and Vera product lines, Morgan Stanley maintains an “overweight” rating for NVIDIA's core supply chain.

~~~~~~~~~~~~~~~~~~~~~~~~

The above wonderful content is from Chasewind Trading Desk.

For a more detailed interpretation, including live coverage and frontline research, please join [Chasewind Trading Desk ▪ Annual Membership]

Risk Disclaimer and Exclusion ClauseThe market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial circumstances or needs of any individual user. Users should consider whether any opinions, views or conclusions in this article are suitable for their particular situation. Investing accordingly is at your own risk. ```