Qualcomm (NASDAQ:QCOM) is positioning its newly announced High Bandwidth Compute, or HPC, technology as a way to address memory-bandwidth constraints in artificial intelligence data centers, while also expanding its server CPU, custom silicon and software efforts.
During a fireside chat, Durga Malladi, Qualcomm’s executive vice president of technology planning and edge data center, said the company’s HPC development program had been underway for years before its public introduction. He said Qualcomm identified a growing mismatch around 2021 between rapidly increasing compute capacity in data-center racks and more modest growth in memory bandwidth.
High Bandwidth Compute Roadmap
Malladi described HPC as a compute-and-memory co-design approach rather than a standalone memory technology. Qualcomm does not manufacture the DRAM stacks used in the architecture, he said, but provides the compute die and system design while working with memory vendors and foundry partners including TSMC.
Qualcomm has said its first-generation HPC product is expected to be commercial and shipping in 2027, with a second generation planned for 2028. Malladi said silicon for the first generation has returned to Qualcomm’s lab and is progressing well, though the company did not provide a specific date for further validation disclosures.
The executive reiterated Qualcomm’s previously stated performance targets, including a claimed sixfold improvement in tokens per second per watt relative to high-bandwidth memory, or HBM. He also said Qualcomm’s architecture is designed to provide memory bandwidth in the “hundreds of terabytes per second,” compared with approximately 21 to 22 terabytes per second for HBM today, according to his remarks.
He said the company sees an advantage versus SRAM in total cost of ownership because SRAM’s speed comes with a larger area requirement, potentially requiring more racks for equivalent workloads. Malladi added that Qualcomm has received interest from memory suppliers and hyperscalers, including discussions on supply and potential product configurations.
- HPC can be offered as a standalone component alongside a customer’s existing CPU and AI accelerator, according to Malladi.
- Customers may also adopt Qualcomm’s CPU, AI accelerator and HPC technology together.
- Qualcomm expects early adopters to use the first generation, while potentially broader volume demand could develop with the 2028 second generation.
Malladi said HPC is targeted at decode workloads, while prefill remains predominantly compute-limited. He said customers may use their existing compute hardware for prefill and incorporate Qualcomm’s technology for memory-bandwidth-intensive decoding.
Server CPU and RISC-V Plans
Qualcomm’s Dragonfly C1000 server CPU is scheduled for commercial availability in the second half of 2028, Malladi said. Meta has been announced as the first customer for the product.
Malladi said Qualcomm designed the C1000 to compete with projected data-center CPU offerings available in 2028 rather than products currently on the market. He said the company believes its performance-per-watt claims will be compelling, though he acknowledged that additional customers want to see silicon validation before committing.
He also discussed Qualcomm’s parallel investment in Arm-based and RISC-V processor roadmaps. Malladi characterized RISC-V as an increasingly commercial technology rather than an academic initiative, citing stronger engagement from hyperscalers in the U.S. and China. Qualcomm acquired Antenna Microsystems last year, he said, and has integrated that team into its RISC-V efforts.
Rather than replacing Arm, Malladi said Qualcomm expects RISC-V to be a parallel option based on customer requirements. “It is not A or B, it is A and B,” he said.
Modular Software and Cloud-to-Edge Strategy
Malladi also highlighted Qualcomm’s acquisition of Modular, whose Mojo programming language and MAX compiler stack are intended to support AI workloads across a range of third-party hardware platforms. Qualcomm has said the Modular stack can provide performance on par with or up to 50% better than native stacks in certain cases, though Malladi said the company wants independent third parties to validate performance.
He said Mojo is being made open source, while MAX is largely available under an Apache 2.0 license, with some components licensed separately. Qualcomm recently demonstrated Modular software running on its own platforms, including its AI 100 data-center hardware and Snapdragon X Elite laptop platform, Malladi said.
The company plans to release additional performance data over the next one to two quarters comparing existing rack software stacks with workloads running on Mojo and MAX, he said.
Qualcomm is also working with Hugging Face to simplify deployment of AI models onto Qualcomm platforms. Malladi said the partnership is intended to enable more automated model onboarding for developers, including through Hugging Face’s conversational tools, and to expose developers to the Modular software stack.
Finally, Malladi said custom silicon remains an important part of Qualcomm’s data-center strategy following its acquisition of Alphawave. He said Qualcomm retained Alphawave’s custom silicon business and gained networking capabilities such as SerDes, while continuing engagements with Alphawave’s existing hyperscaler customers.
About Qualcomm (NASDAQ:QCOM)
Qualcomm Incorporated is a global semiconductor and telecommunications equipment company headquartered in San Diego, California. Founded in 1985, the company is known for its development of wireless technologies and for playing a central role in the evolution of digital cellular standards, including CDMA and subsequent generations of mobile standards. Qualcomm’s business combines the design and sale of semiconductor products with a patent licensing program for wireless technologies and related intellectual property.
The company’s product portfolio includes system-on-chip (SoC) platforms marketed under the Snapdragon brand, cellular modem and RF front-end components, connectivity solutions for Wi‑Fi and Bluetooth, and processors and platforms aimed at automotive, IoT, networking and edge-computing applications.
