Brushing off new bubble warnings, Google’s AI comeback and Nvidia’s China threat

Key Concepts

• AI Bubble: Concerns about inflated valuations and expectations in the AI market exceeding actual capabilities.
• Nvidia: Dominant player in AI hardware (GPUs), experiencing high demand and strong earnings, but facing potential threats.
• Google (Alphabet): Shifting from an AI "laggard" to a leader with its Gemini models and custom TPUs, demonstrating strong execution and a full AI stack.
• Gemini 3.0: Google's latest AI model, performing well on leaderboards and powering new products like Nano Banana Pro.
• TPUs (Tensor Processing Units): Google's custom AI chips, offering efficiency and differentiation.
• Nano Banana Pro: Google's image generation model, built on Gemini 3.0, designed for professional design tasks.
• Open Source AI: The global trend towards open-source models, particularly driven by Chinese companies, posing a challenge to closed, proprietary models.
• Huawei: A Chinese tech company, a significant threat to Nvidia's dominance due to its development of domestic AI hardware and open-source models.
• Sovereign AI: The concept of countries developing their own AI capabilities and infrastructure, often driven by geopolitical concerns.
• Compute Cost: The ongoing trend of decreasing compute costs due to innovation, impacting business models reliant on hardware appreciation.
• China Risk: The underappreciated threat posed by China's advancements in AI hardware, software, and its global strategy of promoting open-source models and infrastructure.

AI Bubble Fears vs. Market Momentum

The AI market is experiencing a significant disconnect between growing concerns about a potential bubble and continued strong momentum, particularly driven by Nvidia's impressive earnings. While insiders like Alphabet CEO Sundar Pichai and Google DeepMind CEO Demis Hassabis have voiced concerns about "irrational exuberance" and "classic signs of an AI bubble," citing extremely high valuations for private companies and seed rounds, the capital continues to flow into the sector. Anthropic, Databricks, and XAI are reportedly seeking tens of billions of dollars in new funding.

Despite Nvidia's blowout quarter and raised forecast, its stock saw a temporary dip after earnings, highlighting the market's sensitivity to both positive fundamentals and underlying anxieties. Dan Niles, founder and portfolio manager of Niles Investment Management, notes that while Nvidia's revenue growth (up 56% last quarter, 62% this quarter, and guided to 65% growth) is strong, questions remain about the depreciation life of AI hardware. The assertion that older A100 chips are still fully utilized due to CUDA software is met with some skepticism, as it contrasts with Nvidia's own rapid release cadence of new chips like Blackwell.

A key concern raised is the potential for overstating earnings if depreciation lives for AI hardware are extended beyond their actual useful life. This impacts hyperscalers who borrow heavily to build data centers. The interconnectedness with the private credit market, which has seen issues with firms like Tricolor and First Brands, adds another layer of complexity.

The immense capital commitments of private AI companies, such as OpenAI's $1.4 trillion in capex over eight years against a $20 billion revenue run rate, and Anthropic's $100 billion commitment against a $9 billion run rate, raise questions about their long-term funding viability. This is leading to market differentiation between companies that can fund their AI ambitions with cash flow and those that cannot.

Google's AI Ascendancy and the Gemini 3.0 Advantage

Google (Alphabet) is emerging as a strong contender in the AI space, with its shares reaching all-time highs. This resurgence is attributed to its successful integration of AI into its core products and the development of its full AI stack, including its custom Tensor Processing Units (TPUs). The release of Gemini 3.0, which has topped third-party leaderboards, is a significant driver.

Josh Woodward, VP of Gemini and Google Labs, highlights the capabilities of Gemini 3.0, noting its significant leap in performance benchmarks. The fact that Gemini 3.0 is trained and served on Google's own TPUs provides a crucial differentiation and cost advantage. This strategic investment in custom silicon and infrastructure is seen as a key factor in Google's current success, aligning its technical depth with tangible execution. The re-engagement of co-founder Sergey Brin and the return of key AI talent have also contributed to this shift.

Google's product strategy emphasizes accessibility and user engagement. The viral success of Nano Banana, an image generation model, has been followed by Nano Banana Pro, built on Gemini 3.0. This new version offers enhanced capabilities for professional design, including infographics, slide decks, and maintaining character consistency across multiple images. The model's ability to visualize previously non-visual concepts is a key draw for users.

Woodward describes the current period at Google as "unparalleled" due to the pace of innovation and the new use cases enabled by AI models. The company is focused on meeting users where they are, whether through integration into search or by creating fun and whimsical product experiences. The concept of "success disasters" is used to describe situations where products take off unexpectedly, leading to compute limitations, but also highlighting high demand.

Google's approach to monetization is currently focused on building beloved products, with Gemini envisioned as a personal, proactive, and powerful assistant. While subscription models are in place for the Gemini app, the company is open to exploring other avenues like advertising in the future. The focus remains on scale and user acquisition, balanced with the efficiency gains from their custom TPUs. Gemini 3.0, while trained on TPUs, can also be served on GPUs through Google Cloud, offering flexibility.

The Underappreciated Threat of China and Open Source AI

Aaron Jinn, co-founder and CEO of Hydrahost, known as the "GPU whisperer," identifies the "attitude of the West" towards AI and the rise of Chinese AI capabilities as the most significant underappreciated threats to Nvidia's dominance. He argues that investors and traders in the West often fail to grasp the threat posed by companies like Huawei because they don't fully understand Nvidia's integrated hardware-software advantage, which is akin to Apple's ecosystem.

Jinn points out that China is not only adept at engineering but is also strategically building its own AI capacity and frameworks, particularly through open-source models. This approach contrasts with the West's tendency towards closed, proprietary models. The global embrace of open-source AI, largely driven by Chinese models, is seen as the dominant trajectory, with models becoming akin to commodity software running on GPUs or XPUs.

The ban on Nvidia chips in China and the push towards domestic accelerators from Huawei and others signify a move towards hardware control without the risk of future US restrictions. This creates a significant challenge for Nvidia, as China is a major market. Jinn believes that the West's restrictive policies, while intended to maintain a technological edge, inadvertently encourage China to accelerate its own advancements.

Furthermore, China's strategy is not solely about IP theft but about achieving internal stability and sovereignty. Their focus on internal adoption and security, rather than external projection like the US, drives their approach to AI development. Huawei's open-source models are seen as a means to sell infrastructure to governments, establishing a footprint similar to their Belt and Road Initiative.

The trend towards open-source models means that while frontier labs in the US are protected by VC funding and access to GPUs, the global adoption is shifting towards low-cost, accessible models. This creates a mismatch between domestic AI development and actual global adoption. Jinn suggests that the US should leverage its current dominant market share to accelerate the adoption of American infrastructure and technology in countries where China is also vying for influence.

The OpenAI release of OSS, a purportedly open-source model, is viewed by Jinn as a "half-hearted attempt" with limited adoption. He emphasizes that the world is embracing open source, not closed models, and that this trend, coupled with China's strategic hardware development and global outreach, represents a significant, underappreciated risk to the Western AI trade. The US government's hyper-focus on specific users or countries, while overlooking China's role as a dominant trading partner for most of the world, is also seen as a strategic misstep.

Conclusion and Key Takeaways

The AI landscape is characterized by a dual narrative: immense excitement and rapid innovation, juxtaposed with growing concerns about market sustainability and geopolitical shifts. While companies like Google are demonstrating impressive execution and technological prowess with their integrated AI stacks and custom hardware, the broader market faces the challenge of inflated valuations and the potential for a bubble.

The rise of open-source AI, largely propelled by Chinese companies like Huawei, presents a significant strategic threat to Western dominance, particularly to Nvidia. This trend, combined with China's focus on sovereign AI and global infrastructure development, suggests a fundamental shift in the AI value chain. Investors and policymakers are urged to recognize the underappreciated risks associated with China's advancements and the global embrace of open-source models. The future of the AI trade will likely be defined by the ability of companies and nations to balance innovation with sustainable business models and navigate an increasingly complex geopolitical environment.