AI Chip Race Heats Up

by | Nov 17, 2025 | The Doctor's Dose

AI Chip Race Heats Up as Tech Giants Build Their Own Silicon

The AI chip race has reached a new level of intensity as the world’s biggest technology companies rush to design their own processors. For decades, semiconductor innovation was left to specialists like Nvidia and AMD. But the explosion of artificial intelligence applications has changed everything, driving Alphabet, Amazon, Microsoft, and Meta to bring chipmaking in-house.

Apple set the precedent in 2010 with its custom A4 chip for the iPhone 4, demonstrating that controlling silicon could enhance performance and profitability. Now, with Nvidia valued at $4.6 trillion and maintaining near-monopoly margins of 74% in its data center business, every major player wants a piece of the hardware stack. The AI chip race is no longer about innovation—it’s about control.

From Dependence to Diversification

Companies have realized that outsourcing chip design leaves them at the mercy of Nvidia’s pricing and production cycles. Training frontier AI models is becoming prohibitively expensive—costs have risen 2.4 times per year since 2016, and could exceed $1 billion per model by 2027. To stay competitive, tech giants are turning to custom application-specific integrated circuits (ASICs), which can be optimized for their cloud and AI workloads.

Google’s Tensor Processing Units (TPUs), Microsoft’s Azure Maia and Cobalt chips, Amazon’s Trainium processors, and Meta’s in-house silicon strategy all illustrate how the AI chip race is reshaping data center infrastructure. These chips are designed to strike a balance between performance and energy efficiency, thereby reducing dependence on Nvidia while lowering long-term costs.

Winners and Wild Cards

Suppliers like Broadcom, Marvell Technology, and MediaTek stand to gain from this build-it-yourself movement, handling the engineering and manufacturing for cloud giants. Bernstein estimates that the ASIC market could grow at a 55% annual rate to reach $60 billion by 2028. Meanwhile, Nvidia still dominates with projected sales of $375 billion in the same period—proof that it will remain at the center of the AI chip race for years to come.

Outside the U.S., China’s Alibaba and Baidu are developing their own chips to cut reliance on Western technology. Even automakers are joining the fray, betting that custom silicon will accelerate their self-driving ambitions. The AI boom has even inspired Japan’s SoftBank to explore acquisitions, such as Marvell, positioning itself as a global enabler of next-generation chip production.

Trading Insights and Market Impact

  • Nvidia (NVDA): Despite competition, it remains the cornerstone of AI infrastructure. Pullbacks below $185 could offer long entries toward $200–205 ahead of Q4 guidance.
  • Broadcom (AVGO): The biggest near-term beneficiary of ASIC contracts. Support sits near $1,200; a breakout above $1,330 targets $1,400 in the short term.
  • Marvell (MRVL): Speculative buy for traders anticipating M&A activity or rising custom chip demand. Watch $67 as the next pivot resistance.
  • Microsoft (MSFT): Short-term volatility expected as investors digest capex growth for in-house silicon. Look for entries on dips toward $400 support.

For active traders, the near-term setups hinge on capital expenditures (capex) guidance and chip cost efficiency. If these firms can demonstrate that in-house designs lower total AI training costs, we could see a shift toward cloud infrastructure providers that control their own hardware destiny. Conversely, any delays in rollout could spark profit-taking across the AI chip race sector.

Bottom Line

The AI chip race is transforming the global semiconductor ecosystem and reshaping the landscape of technological power. Whether it leads to independence or inefficiency depends on execution. For now, Nvidia remains the benchmark—but every custom chip announcement chips away at that dominance. Traders should keep an eye on hardware efficiency metrics, production timelines, and margin guidance as key catalysts into 2026.