What is the smartest way to invest in AI today? According to venture leader Nicolas Sauvage, the answer isn’t flashy apps or viral tools—it’s the overlooked infrastructure powering them. Speaking at StrictlyVC San Francisco, Sauvage explained why the most successful AI bets take years to mature and often focus on “boring” but critical technology. His strategy is now gaining attention as AI demand surges globally.
 |
| Credit: Google |
THE FOUR-YEAR AI INVESTMENT STRATEGY EXPLAINED
Sauvage’s core philosophy is simple but unconventional: the best investments only look smart after about four years. That long-term thinking has shaped how TDK Ventures deploys capital across emerging technologies.
Rather than chasing trends, the firm identifies future bottlenecks—problems that don’t seem urgent today but will become critical as industries scale. This requires patience and a willingness to invest before the market fully understands the opportunity. It also means ignoring hype cycles, which often reward short-term thinking over sustainable growth.
This approach has proven effective, particularly in AI infrastructure. While many investors rushed into consumer-facing AI during the generative boom, Sauvage focused on the systems enabling those experiences behind the scenes.
WHY AI INFRASTRUCTURE IS THE REAL GOLDMINE
One of the standout examples of this strategy is Groq, a company specializing in inference chips. Unlike training chips, which teach AI models, inference chips handle the real-time processing when users interact with AI.
Back in 2020, before generative AI became mainstream, Sauvage invested in Groq when it was still relatively unknown. Today, the company is valued at billions, reflecting the explosive demand for AI inference as chatbots, assistants, and autonomous systems scale globally.
The logic is straightforward. Every AI interaction—every query, response, or automated action—requires inference. As AI adoption grows, so does the need for faster, more efficient processing. This creates a compounding demand curve that infrastructure players are uniquely positioned to capture.
Sauvage recognized early that while consumer hardware markets can saturate, AI infrastructure demand continues to expand with every new application.
THE SHIFT FROM GPUs TO INFERENCE AND CPUs
The AI hardware landscape is evolving rapidly. Graphics processing units dominated the early phase of AI development, powering the training of large models. But the focus is now shifting.
Inference chips like those built by Groq are becoming essential as AI moves from development to deployment. They enable faster responses, lower costs, and scalable performance—critical factors for real-world applications.
At the same time, Sauvage predicts a resurgence of central processing units. While not as powerful as GPUs for raw computation, CPUs excel at coordination and decision-making. This makes them ideal for managing AI agents that perform complex, multi-step tasks.
As AI systems become more autonomous—planning, executing, and adapting in real time—the need for orchestration increases. CPUs, with their flexibility, are well-suited to handle this layer of intelligence.
THE RISE OF PHYSICAL AI AND SPECIALIZED ROBOTICS
Beyond chips, Sauvage is closely watching the emergence of physical AI—robots designed to perform specific, practical tasks. Unlike general-purpose robots, these machines focus on doing one thing extremely well.
For example, Agility Robotics is building robots that move goods in warehouses, addressing labor shortages in logistics. Meanwhile, ANYbotics develops machines capable of operating in hazardous environments where humans cannot safely go.
The common thread is specialization. Instead of trying to replicate human versatility, these robots target clearly defined problems. This makes them more reliable, efficient, and commercially viable.
Sauvage believes this focused approach will drive the next wave of robotics adoption, particularly in industries facing workforce challenges or safety risks.
CHINA’S “VIBE MANUFACTURING” ADVANTAGE
Another major trend shaping AI’s future is what Sauvage describes as “vibe manufacturing.” This concept refers to the rapid, AI-assisted iteration of physical products—similar to how software development has accelerated through automation.
In China, manufacturers are already compressing the design-build-test cycle for hardware. This allows them to prototype and refine products at a pace that traditional supply chains struggle to match.
The implication is significant. Countries and companies that can iterate on physical products as quickly as software will gain a major competitive edge. This could reshape global manufacturing dynamics, particularly as AI becomes more deeply integrated into production processes.
Sauvage sees this as both a challenge and an opportunity, pushing investors to rethink where and how innovation happens.
THE UNSOLVED PROBLEM: DEXTERITY IN AI SYSTEMS
Despite rapid progress, one major hurdle remains: dexterity. While AI models are improving at an extraordinary pace, physical systems still lack the fine motor skills required for complex real-world tasks.
This gap limits what robots can do, especially in environments that require precision and adaptability. Solving this problem will be key to unlocking the full potential of physical AI.
Sauvage argues that the breakthrough will come from combining advanced AI models with improved hardware capabilities. When machines can manipulate objects as effectively as humans, entirely new industries could emerge.
WHY “BORING” TECH IS ACTUALLY HIGH-IMPACT
What makes Sauvage’s strategy stand out is his focus on areas that many investors overlook. Batteries, transformers, chips, and robotics may not generate headlines like consumer apps, but they form the backbone of technological progress.
These sectors often require deep expertise, longer timelines, and higher upfront investment. However, they also offer more defensible advantages and less competition compared to crowded consumer markets.
By targeting these “boring” areas, Sauvage positions TDK Ventures to benefit from foundational shifts in technology rather than short-lived trends.
This approach aligns with a broader shift in venture capital, where investors are increasingly prioritizing resilience, scalability, and long-term impact over rapid but fragile growth.
WHAT THIS MEANS FOR THE FUTURE OF AI
The implications of this strategy extend far beyond venture capital. As AI continues to evolve, the balance of power may shift toward those controlling the underlying infrastructure.
Companies building chips, robotics, and manufacturing systems could become the real gatekeepers of the AI economy. Meanwhile, software layers—while still important—may become more commoditized over time.
For businesses and entrepreneurs, this highlights the importance of thinking beyond immediate opportunities. The biggest breakthroughs often come from solving foundational problems that others ignore.
For investors, it’s a reminder that patience and conviction can outperform hype-driven decision-making.
Nicolas Sauvage’s AI investment strategy offers a compelling counter-narrative to the industry’s obsession with rapid wins. By focusing on infrastructure, robotics, and long-term bottlenecks, he’s betting on the systems that make AI possible—not just the applications people see.
As AI demand accelerates, this “boring” approach is proving anything but. It’s shaping the future of technology in ways that are deeper, more sustainable, and ultimately more valuable.
