The leaders of Nvidia, AMD, and Intel are converging on Taiwan in a high-stakes race to lock down the next five years of AI production capacity, signaling a new front in the chip war is opening up around CPUs.
The near-simultaneous arrival of the three most powerful executives in semiconductors—Nvidia’s Jensen Huang, AMD’s Lisa Su, and Intel’s Pat Gelsinger—in Taipei ahead of the annual Computex conference underscores a critical shift in the AI hardware buildout. While the initial AI boom was a GPU-driven story, the industry is now scrambling to secure capacity for the entire compute stack as demand for central processing units (CPUs) for AI workloads explodes.
“The overall CPU market has had significantly higher demand than any of us predicted a year ago,” AMD CEO Lisa Su told reporters in Taipei on Friday, after meeting with the company's largest customers in China. “I would say the CPU market is tight.”
That tightness is being driven by the rapid adoption of AI inference and agentic AI systems, which rely heavily on CPUs for data scheduling and system control. The numbers support the scramble: Intel executives have disclosed that the typical GPU-to-CPU ratio in AI training clusters has compressed from 8-to-1 to 4-to-1, and could approach parity. AMD, meanwhile, projects the CPU market will see a compound annual growth rate of more than 35 percent over the next five years.
The trips are a direct acknowledgment that the road to AI dominance runs through Taiwan’s semiconductor fabs. At stake is the allocation of future manufacturing capacity, primarily at industry linchpin TSMC, for the next generation of AI infrastructure expected to be deployed between now and 2029. Nvidia, which has so far dominated the market for AI training, is now pushing into the server CPU market, with CEO Jensen Huang pegging the total addressable market for its upcoming Vera CPU at $200 billion.
CPU Demand Surge Opens New Front in AI Chip War
The AI industry is moving beyond simply training large models and into a phase of mass deployment and inference, a shift that has unexpectedly reignited the CPU market. While GPUs remain the core workhorse for training, the execution of these models and the orchestration of "agentic" AI systems that can perform autonomous tasks require vast fleets of powerful CPUs.
“If you think of AI as a 9-inning baseball game, we are probably only in the third inning,” Su said, framing the current buildout as being in its very early stages.
Nvidia’s Huang echoed that sentiment, noting the market has rapidly expanded from training to encompass AI agents and "AI Factories." This new phase requires a different hardware mix, creating an opening for competitors and a new battleground for the three American chip giants. AMD announced it would invest more than $10 billion in Taiwan's AI sector to expand its capacity for advanced chips, including packaging and substrates for entire rack-scale systems.
Intel's Gelsinger Mounts Taiwan Charm Offensive
While Huang and Su are visiting from a position of strength, Intel CEO Pat Gelsinger’s trip is a crucial part of his campaign to restore the company’s manufacturing prowess and credibility. Gelsinger, referred to as Chen Liwu in some local media, is on a mission to demonstrate that a "煥然一新 (completely new) Intel" has arrived.
His agenda includes a series of closed-door meetings with Taiwan’s "electronic five brothers"—the major contract manufacturers like Quanta and Foxconn—as well as PC makers like Asus. The goal is to convince them to use Intel’s own foundry services for future AI servers and PCs, a direct challenge to TSMC’s dominance. Gelsinger has been vocal about progress on the company’s 18A process node and has highlighted that its next-generation 14A node is showing better yields at an equivalent stage of development, with customer commitments expected in the second half of 2026.
For investors, the CEO pilgrimage to Taipei confirms that the AI infrastructure buildout is a multi-year, capital-intensive marathon, not a sprint. The intense focus on securing long-term capacity is unequivocally bullish for TSMC, which remains the critical manufacturing partner for all three rivals, as well as for the broader Taiwanese ecosystem of companies involved in advanced packaging and assembly. AMD's $10 billion investment pledge is a concrete down payment on a much larger wave of capital set to flow into the island's tech sector. The key takeaway is that the AI investment thesis is broadening from a pure-play GPU story to a full-stack compute story, potentially revaluing companies all along the server component supply chain.
This article is for informational purposes only and does not constitute investment advice.
