AI sector: Bubble concerns, deal making, demand, and 2 stocks to watch

Key Concepts

• AI Deal Making: Recent significant transactions in the Artificial Intelligence sector, including investments and partnerships.
• Circular Deals: Investments where companies invest in entities that are also investors in them, creating a closed loop.
• Hyperscalers: Large cloud service providers (e.g., Google, AWS, Microsoft) that drive significant AI spending.
• Capex (Capital Expenditure): Spending by companies on physical assets, such as data centers and AI hardware.
• AI Chips: Specialized processors designed for AI workloads, including GPUs (Graphics Processing Units) and ASICs (Application-Specific Integrated Circuits).
• GPU Dominance: Nvidia's current leading position in the AI chip market.
• Server Rack Level Solutions: Integrated computing systems designed for data centers.
• AI Accelerators: Hardware designed to speed up AI computations.
• Valuation: The assessment of a company's worth, particularly in the context of high growth potential.
• AI Bubble: The concern that AI company valuations have become excessively high and unsustainable.
• Air Pocket: A temporary dip or slowdown in market performance, supported by underlying fundamentals.
• Execution Risk: The possibility that companies may fail to deliver on their projected performance or product rollouts.
• Supply Side Concerns: Potential bottlenecks or limitations in the production and delivery of AI hardware.
• Power Bottlenecks: Limitations in the electrical power supply required to operate data centers.

OpenAI's Strategic Investments and Partnerships

OpenAI has been actively engaged in deal-making within the AI ecosystem. Notably, the company has taken an ownership stake in Thrive Holdings, a firm founded by one of OpenAI's investors. This move is characterized as a "circular deal," where investment flows create interdependencies. Additionally, OpenAI has established a partnership with Accenture, positioning itself as one of Accenture's primary AI partners. These actions highlight OpenAI's strategy to solidify its position and expand its influence through strategic alliances and financial arrangements.

Nvidia's Investment in Synopsis and Broader AI Market Dynamics

Nvidia has also been a significant player in AI investments, announcing a $2 billion investment in Synopsis. This underscores Nvidia's commitment to advancing the AI infrastructure. The discussion with Bloomberg Intelligence senior semiconductor analyst Kujan Shobani delves into the sustainability of AI valuations and the nature of these "circular investments."

Fundamental Outlook and Hyperscaler Capex Projections

Shobani presents a fundamentally positive outlook for the AI market, projecting that hyperscaler capital expenditure (capex) will exceed half a trillion dollars by 2026. He emphasizes that these hyperscalers possess substantial balance sheet cash, mitigating concerns about their ability to fund this spending. The fundamental view indicates no slowdown in demand or spending for AI chips, particularly GPUs and AI ASICs, with visibility extending into the first half of 2027.

Key Inflection Points and Emerging Players in the AI Chip Market

While Nvidia is expected to maintain its dominant market share in AI spending through 2026, the second half of 2026 is identified as a crucial inflection point. This period will see the ramp-up of AMD's first server rack-level solutions in data centers. Concurrently, significant deployments of AI-based accelerators from Google and AWS are anticipated. This marks the first instance of these hyperscalers supplying their AI accelerators to external customers, drawing considerable attention.

Diversification of AI Spending Drivers

The AI market is no longer solely driven by the largest hyperscale cloud service providers (referred to as "Mac 7"). New, emerging hyperscalers, such as OpenAI, are also contributing significantly to demand and spending. Concerns surrounding revolving financing and revenue recycling are primarily linked to this emerging segment of hyperscalers. While their impact is growing, it is not expected to be fully realized within the next year to eighteen months, making it a more forward-looking concern.

Underappreciated Winners: Broadcom and Marvell

Beyond the prominent players like Nvidia and OpenAI, Shobani highlights Broadcom and Marvell as underappreciated winners in the AI space. He points to recent deals, such as Google's and AWS's investments in Anthropic, as indicators of a trend where large hyperscalers, despite having their own AI ASICs, are increasingly supplying these to the external market, thereby competing with Nvidia. Broadcom and Marvell are positioned to benefit significantly from this trend, with their valuations and fundamentals not yet fully reflecting this potential.

Valuation and Execution Risks in the AI Cycle

Regarding current valuations, Shobani notes that while multiples for large AI names are not considered exorbitantly high, they appear to have priced in strong growth for 2026 and a portion of the first half of 2027. The primary concern shifts from the spending side to the supply side. The success of new programs, such as AMD's server rack ramp and ASIC programs, hinges on their execution. Potential hiccups in these rollouts, along with the capacity of data center infrastructure to keep pace, especially concerning power limitations, are key areas of focus.

The "Air Pocket" Analogy for the AI Cycle

Drawing on a Bank of America note, Shobani describes the current AI market sentiment as an "air pocket supported by earnings and fundamentals." He agrees with this assessment, considering the macro environment, interest rate volatility, and the typical performance patterns observed at this time of year. He believes that if execution on the projected 2026 numbers and the first half of 2027 proceeds as anticipated, significant valuation risk is unlikely. However, the expectation of consistently exceeding expectations each quarter means that even minor timing hiccups in large product ramps could lead to valuation volatility.

Top AI Accelerator Investment Recommendations

When asked about specific names within the AI accelerator category, Shobani reiterates Nvidia and Broadcom as top choices. He also mentions Marvell and Aera Labs as potentially interesting names to watch.