Qualcomm이 Amazon과 대규모 AI 칩 공급 거래를 체결하여 맞춤형 AI 가속기 시장에서 NVIDIA의 의미 있는 대안으로 부상했다.

Chipmaker Qualcomm and Amazon announced a significant artificial intelligence data-center collaboration on Tuesday through Amazon Web Services. Under the deal, Qualcomm issued Amazon warrants to purchase 25 million shares at $161.26 each—an investment of just over $4 billion. According to Qualcomm's Securities and Exchange Commission filing, these warrants vest in tranches tied to the execution of certain commercial arrangements and the purchase of up to $60 billion in Qualcomm chips and other components. The market responded positively, with Qualcomm stock jumping as much as 8.7% during intraday trading.

This announcement represents a watershed moment in Qualcomm's pivot from smartphones to AI infrastructure. In roughly three months, Qualcomm has secured hyperscaler contracts with ByteDance (early June), Meta Platforms (late June), and Amazon. While specific contract values remain undisclosed, these deals are positioned to add billions to Qualcomm's revenue.

Qualcomm's expansion into data centers marks a dramatic shift from its historical dominance in mobile. Over the past decade, the company established itself as the leading force in smartphone System-on-Chip design, holding 23 percent of all global smartphone SoCs shipped through Q2 2026. However, its data-center revenue in 2025 was zero. In fiscal 2027, Qualcomm expects that figure to jump to $5 billion. By 2029, the company is targeting $40 billion in non-handset revenue (up from an earlier forecast of $22 billion), with $15 billion from data centers alone—though the Amazon deal could push this target higher.

Much of this growth will be driven by Qualcomm's new Dragonfly C1000 CPU, designed specifically for data centers with a focus on AI inference. Meta will use this CPU in its data centers once production begins in 2028. The Amazon partnership extends further, with Qualcomm and Amazon collaborating on customized silicon at scale and high-speed optical connectivity technology reaching speeds up to 1.6 terabits per second. This networking capability paired with computing power positions Qualcomm to offer companies a full-stack alternative to Nvidia.

The shift toward agentic AI explains much of this opportunity. While Nvidia maintains near-monopolistic dominance in large language model training through its Blackwell, Hopper, and Rubin GPU clusters, the industry is increasingly moving toward agentic AI systems that require autonomous agents to perform complex reasoning tasks. These systems demand processors that make fast decisions and manage complicated workflows—precisely what high-core CPUs like the Dragonfly C1000 (with 250-plus cores) are built to handle far more efficiently than GPUs or Nvidia's newest Vera CPU (88 cores).

The Dragonfly C1000 works in tandem with Qualcomm's Dragonfly AI300 inference accelerators to create a faster, more power-efficient platform than typical GPUs. Given the immense power requirements of AI and data centers, power efficiency has become a critical competitive advantage. Hyperscalers like Meta and Amazon are increasingly adopting custom architecture solutions that reduce costs and dependence on Nvidia's ecosystem, which remains both costly and power-intensive. With Goldman Sachs projecting global AI spending to exceed $1 trillion in 2026—mostly from Amazon, Meta, Alphabet, and Microsoft—every efficiency gain is material.

While Qualcomm is unlikely to achieve Nvidia's market dominance or reshape the AI industry single-handedly, it need not. Each hyperscaler contract gradually erodes Nvidia's market share and revenue, even if in increments of half a percentage point. AMD has demonstrated this principle: it will probably never close the gap with Nvidia, but it is successfully cutting into the company's market position.

Qualcomm's stock trajectory reflects broader market dynamics. The company peaked at $259.92 in intraday trading on May 29, representing a 62 percent year-over-year gain. By September 9, the stock had settled at $176.40, up 10.6 percent year-over-year. The sharp decline from May reflects Apple's decision to manufacture its own 5G cellular modem chip rather than source from Qualcomm, leading to a 20 percent year-over-year drop in handset revenue during the company's fiscal third quarter of 2026. Investors also likely realized that the...