Beyond the AI Bubble
Ben Lorica and Evangelos Simoudis discuss signals of an AI bubble, including runaway valuations and circular financing. They explore why many enterprise AI pilots fail and what separates successful adopters—namely, a focus on measurable use cases and organizational transformation—and offer a nuanced perspective on the US-China tech race, cautioning against underestimating China's progress in AI, robotics, and semiconductors.
A Tale of Two Bubbles: Dot-Coms and AI
From an investor’s perspective, the current AI landscape shows clear signs of a bubble, with valuations reaching multiples of revenue not seen since the 1999-2000 dot-com era. Funding proposals are often “euphemistically flimsy,” and investment is heavily concentrated in AI, much like it was in internet companies during the late ’90s.
Two key warning flags stand out:
- Circular Financing: A glaring signal is the trend of companies investing in their own customers. For example, if there is truly overwhelming demand for NVIDIA’s chips, the company shouldn’t need to financially prop up its customers to drive sales.
- Market Concentration: The broader market reflects this concentration, with the “Magnificent 7” driving a disproportionate amount of growth in the S&P 500.
However, there is a critical difference between now and the dot-com bubble. The world has experienced previous AI hype cycles, which should provide a more “educated way of approaching what could happen.” We have precedents for AI’s boom-and-bust cycles, unlike the first internet wave.
Separating AI Winners from Losers in the Enterprise
A significant number of enterprises are initiating AI experiments not from a strategic standpoint, but due to external pressures or internal trends. This adoption is often driven by:
- Board Pressure: Leadership pushing for AI adoption without a deep understanding of its implications.
- Shadow AI: Employees bringing consumer AI tools they use in their private lives into the workplace.
This approach often lacks a clear business rationale. Many organizations fail to articulate specific use cases, define how they will measure performance improvements, or plan their data strategy—distinguishing between internal data assets and necessary external data.
What Successful Adopters Do Differently
The leaders who succeed in AI adoption move beyond mere technological experimentation and focus on holistic transformation. Their success is built on several pillars:
- The Use Case Matrix: Successful organizations develop a “use case matrix” that identifies clear opportunities for improvement across the entire enterprise. Each use case is tied to specific, measurable metrics.
- Cross-Functional Ownership: True adoption requires breaking down organizational silos. It’s not just a technology project; it involves bringing together technology, business units, HR, and finance to drive the initiative forward.
- A Focus on Transformation: AI is not just a tool but an opportunity for fundamental change. This echoes previous technology waves like e-commerce, cloud computing, and Big Data, where the most successful companies were those that undertook organizational, business process, and cultural transformation alongside technology adoption. The primary barrier to success is often organizational, not a lack of technical talent.
Ultimately, every AI initiative should be connected to one of three core business outcomes: productivity improvement, cost reduction, or new revenue generation. Leaders can answer concrete questions like, “How many more clients can we serve?” or “How does this improve the customer experience?” It is about tangible capability expansion, such as enabling a support team to handle 20 times the call volume or provide 24/7 coverage.
A Realistic Look at the US-China Tech Race
There is a common perception in the United States that it is significantly ahead of China in deep tech, but this view may not align with reality. The competition is much closer to parity than many believe, a fact that should drive “continuous activity and investment.”
Key Areas of Competition
- Robotics and Manufacturing: While the US acknowledges China’s lead in battery electric vehicles, there’s a belief that the US is ahead in adaptive and humanoid robotics for manufacturing. This is, at best, a “contested statement.” China is likely ahead numerically in factory robots and is making significant strides in more advanced robotics.
- Foundation Models: The surprising performance of models like DeepSeek demonstrates that China’s capabilities in large-scale AI should not be dismissed.
- Semiconductors: The US maintains a clear lead in cutting-edge semiconductors. However, China is adapting out of necessity, developing methods to train powerful AI models on older, less advanced chips. This may slow their progress at the absolute frontier, but it will not stop them from developing highly capable systems. Furthermore, for most enterprise applications, the “latest and greatest” models are not required.
The focus should not only be on who is at the “tip of the spear” but also on the broader technological ecosystem just behind it, as this has a much larger economic impact. The significant difference in approach—with China’s state-driven industrial policy versus the US’s reliance on the private sector—will continue to shape the competitive landscape.