Key Points:
- Qualcomm signed Meta as the first customer for its custom-designed Dragonfly C1000 data center CPU, scheduled for production in 2028.
- Microsoft plans to deploy Qualcomm’s High Bandwidth Compute near-memory technology in its Azure data centers starting in 2027.
- The mobile chipmaker confirmed the acquisition of software startup Modular for roughly $3.9 billion to challenge Nvidia’s proprietary CUDA platform.
- Qualcomm established an aggressive revenue target of more than $15 billion from its data center business by fiscal 2029.
The mobile chipmaker has officially taken its biggest step yet into the enterprise market, announcing a highly strategic data center processor agreement with Meta Platforms. Under this multi-generation partnership, Meta will utilize the newly unveiled Dragonfly C1000 server processor inside its massive infrastructure fleet. This deal represents a major validation of the semiconductor designer’s efforts to expand beyond its core smartphone business and challenge Nvidia’s dominance in the highly lucrative artificial intelligence infrastructure sector.
For years, the company has relied heavily on the mature smartphone market, where mobile modems and application processors represent more than two-thirds of its total product revenue. However, with global smartphone shipments peaking and showing cyclical sensitivity, the company has actively sought out faster-growing opportunities in automotive systems, robotics, and cloud computing. The launch of the Dragonfly data center portfolio signals a complete rewrite of the company’s enterprise playbook. By setting a bold revenue goal of more than $15 billion from data center strategies by fiscal 2029, the leadership team is proving that it is fully committed to a multi-year, non-smartphone growth trajectory.
The centerpiece of the collaboration with Meta is the first-generation Dragonfly C1000 CPU, which is slated to enter high-volume production in the second half of 2028. This advanced processor uses a modular multi-chiplet design packed with more than 250 custom-designed Oryon CPU cores, operating at clock frequencies exceeding 5 gigahertz. Built specifically to handle the complex, low-latency requirements of agentic AI orchestration, the processor focuses on delivering exceptional computing performance per core. Crucially, the platform features cutting-edge PCIe Gen 7 connectivity, promising to deliver more than double the power efficiency of existing server CPUs currently running in enterprise data centers.
In addition to securing Meta as a premier customer, the chip designer revealed a deep integration partnership with Microsoft. Microsoft Chief Executive Officer Satya Nadella confirmed that the cloud giant plans to deploy Qualcomm’s new High Bandwidth Compute (HBC) technology across portions of its Azure data centers. This innovative near-memory computing architecture leverages advanced 3D-stacked silicon to move data exponentially faster and break through the industry-wide memory bottleneck. According to the development team, this design delivers up to six times the memory bandwidth per watt compared to traditional setups by separating high-bandwidth memory from the main processing units.
While these announcements represent monumental milestones, the long timelines involved have introduced some near-term caution among stock analysts. Commercial sampling for the first-generation High Bandwidth Compute architecture is scheduled to begin in mid-2027, while the Dragonfly C1000 CPU will not be commercially available to hyperscalers until late 2028. This means that significant data center revenues from these initiatives will not arrive for several years. Following the announcement, the company’s stock price experienced a temporary decline of approximately 5%, as investors carefully weighed the long runway for these products against the immediate costs of separate corporate acquisitions.
To address software challenges and make its hardware immediately attractive to developers, the chip vendor confirmed that it has acquired AI software startup Modular, with media estimates putting the stock-based deal at roughly $3.9 billion. This acquisition is a direct shot at Nvidia’s proprietary CUDA platform, which has long locked developers into a single hardware provider. Modular’s highly sought-after software stack operates as a chip-agnostic computing layer, enabling developers to build complex AI models once and run them seamlessly across diverse compute environments. This software integration represents a critical tool to help customers bypass hardware lock-in and deploy AI workloads with real flexibility.
The primary selling point of the entire Dragonfly ecosystem is its radical focus on energy efficiency. As hyperscalers construct massive, multi-gigawatt data centers to power complex AI models, electrical power has quickly replaced physical space as the primary limiting factor for infrastructure expansion. This is where the chipmaker’s historic expertise in designing low-power, battery-efficient smartphone processors provides a unique competitive edge. By delivering optimized tokens per watt, the company can help hyperscalers drastically reduce their total cost of ownership and manage the extreme electrical loads demanded by next-generation computing clusters.
The momentum extends beyond the announced agreements with Meta and Microsoft. Executives confirmed that the company has secured at least two additional custom silicon design agreements with major global hyperscalers. While the firm did not disclose the identities of these partners, previous industry reports indicated that TikTok owner ByteDance is on track to purchase millions of custom application-specific integrated circuits from the designer. This growing pipeline of bespoke custom design wins proves that large-scale cloud providers are actively seeking alternatives to standard off-the-shelf accelerators to optimize their unique machine learning algorithms.
Ultimately, the developments demonstrate that the semiconductor landscape is rapidly shifting away from a single dominant provider. By combining custom-designed CPUs, near-memory architecture, and a highly developer-friendly open software layer through the Modular acquisition, the company is building a highly credible alternative ecosystem. While the long road to commercial production in 2028 presents an execution risk, the backing of some of the world’s most influential technology giants proves that the strategic transition is already well underway. The battle for the future of agentic AI infrastructure has officially moved beyond mobile devices and into the heart of the global data center.
