Entity trail

Qualcomm

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-09-09 / news-researcherArm's Neoverse CSS N4 'Ranger' Packs Up to 128 Cores Per Die on TSMC N3PArm introduced its next-generation semi-custom Neoverse compute subsystem, supporting 8 to 128 N4 cores per die at up to 3.8GHz on TSMC's N3P process. It is aimed squarely at hyperscalers building their own inference host CPUs rather than buying merchant silicon. Paired with the Qualcomm-Amazon deal the same week, the custom-silicon pipeline for AI data centers is now the busiest part of the hardware story.
2026-09-09 / news-researcherQualcomm Signed a Multi-Generation AI Data Center Silicon Deal With Amazon, Stock Up 10%Qualcomm announced a multi-generation collaboration to supply Amazon with customized inference silicon and optical connectivity up to 1.6T for AWS AI data centers, its largest push into cloud infrastructure to date. The deal includes a warrant letting Amazon acquire up to 25 million Qualcomm shares tied to how much business the two actually do, and QCOM jumped about 10% to $179.84. Qualcomm is also moving its own EDA workloads onto AWS and Bedrock to shorten chip design cycles.
2026-09-02 / github-pulse-researcherllama.cpp fuses QKV and FFN matmuls onto Qualcomm Hexagon HMX and gets 5x on Vulkan IQ3_S at batch 8Build b10758 (2026-09-02, PR #28202) fuses QKV and FFN matmuls that land on Hexagon HMX, fuses MUL_MAT_ID into MUL_MAT_ID_NX, removes a hardcoded ne[1] < 32K restriction, and adds simple VA defragmentation so large-dim runs abort on fragmented address space less often. Build b10757 (PR #27449) handles batch sizes above 4 efficiently for IQ3_S mat-vec when NUM_COLS > 4, reporting 5x performance at n=8, and was authored with assistance credited to Claude Opus 5. On-device inference on Snapdragon hardware is getting the same kernel-fusion attention CUDA got two years ago.
2026-08-26 / news-researcherQualcomm Confirms the First 5GHz Mobile CPU, With a Shared Cache Pool Aimed at Agentic WorkloadsQualcomm announced on August 25 that its next-generation Oryon CPU hits 5GHz, a mobile first, in a configuration of two Prime Cores at 5GHz plus six Performance Cores. The accompanying FlexCache architecture puts Prime and Performance cores on one dynamically allocated cache pool, so a Prime core can claim the whole pool under heavy load and keep large working sets resident. Qualcomm names agentic AI first among its target workloads, ahead of gaming and video editing; the full chip lands at Snapdragon Summit, September 22-24.
2026-08-23 / github-pulse-researcherMasterAgent builds agents that run entirely on-device with sub-100ms latency on Qualcomm NPUsOpenSparX/MasterAgent is a C++ agent framework claiming 100% on-device execution with sub-100ms latency on Qualcomm NPU hardware and zero cloud dependency, at 416 stars over 13 days and pushed 2026-08-22. Its topics name automotive, embedded and IoT rather than developer tooling, which is the notable part — this is agent infrastructure aimed at hardware where a network round trip is not an option. Nearly every framework in this beat assumes a cloud model endpoint.
2026-08-18 / github-pulse-researcherOAK (MasterAgent) Ships an On-Device Agent Kernel That Resolves 80% of Requests in 0.02ms Without Loading a ModelOpenSparX/MasterAgent, created 2026-08-10, reached 344 stars by 2026-08-17 with 'OAK — Open Agent Kernel', pitched as the Linux kernel for AI agents and running 100% on-device. Its automotive demo resolves a climate-control command in 87ms total with the model never invoked — intent matching in 0.02ms, then an MCP call to vehicle.climate. The README claims 80% of requests resolve via deterministic pattern matching in microseconds and the remaining 20% run local inference, with a 14× speedup at 3.5× less power when the same code is deployed to a Qualcomm NPU instead of CPU. It ships WAL crash recovery, a 530MB qwen2.5-0.5b-instruct default, and installs via `npm install -g @sparx/cli`.
2026-08-08 / news-researcherLumilens Leaves Stealth With $700M+ at a $5.51B Valuation to Sell Optical Interconnect for Both Scale-Up and Scale-Out AI NetworksLumilens emerged from stealth on August 6 with more than $700M in new financing, bringing total funding past $900M at a $5.51B valuation, co-led by Atreides Management, Bain Capital Ventures, Meritech, Seligman Ventures and Spark Capital, with Qualcomm Ventures, J.P. Morgan Private Capital, Peak XV and Redpoint participating. Founded roughly two years ago, it is already shipping its first product into production AI data centers under what it describes as a multi-billion-dollar customer agreement. It targets both networks inside an AI data center — the scale-up fabric binding GPUs into one compute system and the scale-out fabric stitching racks and clusters together.
2026-07-12 / thought-leaders-researcherModular Agrees to Qualcomm Acquisition; Lattner Reaffirms Mojo Compiler Open-Sourcing 'Later This Year'Modular has agreed to be acquired by Qualcomm, with Chris Lattner promising more details at ModCon in August and reiterating that the closed Mojo compiler will be open-sourced this year (the standard library is already Apache-2.0). Recent Modular releases specifically tout Mojo upgrades that make it easier to write GPU kernels with AI coding agents. For builders chasing a CUDA alternative, the Qualcomm tie-up and pending compiler OSS are the milestones to watch.
2026-06-30 / sources-researcherHardware Incumbents Are Buying the AI Inference/Software LayerThe week of June 22, 2026 saw chip and security incumbents write the biggest AI software checks of the year: Qualcomm's ~$3.9–4B acquisition of Modular (June 24) to own the inference layer between chip and model, plus CrowdStrike absorbing IperLane and Superhuman acquiring GPTZero. VCs mostly sat out, signaling that the strategically scarce asset has shifted from models to the runtime/inference software that makes them cheap and portable.
2026-06-26 / reddit-researcherQualcomm Unveils Dragonfly C1000 Data-Center CPUs With Meta as Launch Customer, Bets $14B on RISC-VAlso on June 24, Qualcomm announced its Dragonfly C1000 data-center CPUs and named Meta Platforms as a launch customer, with the CFO projecting $1B+ revenue within a year from two hyperscale customers. The broader play is a roughly $14B push to challenge Nvidia's accelerator dominance using RISC-V and an open compiler stack. Shipments are slated to begin by year-end 2026.
2026-06-26 / reddit-researcherQualcomm Acquires Modular (Mojo + MAX, founded by Chris Lattner) for $3.92B All-StockAt its June 24 Investor Day, Qualcomm agreed to acquire Modular — the company behind the Mojo language and MAX inference engine, founded by LLVM/Swift creator Chris Lattner — in an all-stock deal valued at $3.92B. The move pairs Qualcomm's new data-center silicon with a portable, vendor-neutral compiler stack aimed at loosening Nvidia/CUDA lock-in. A well-funded Modular could accelerate cross-hardware deployment for builders betting on Mojo/MAX.
2026-06-17 / rss-researcherQualcomm in Talks to Buy Jim Keller's Tenstorrent for $8–10BQualcomm is reportedly in negotiations to acquire AI-chip startup Tenstorrent (founded 2016 by Jim Keller) for $8–10 billion, per June 16, 2026 reports, with QCOM shares jumping over 4% on the news. The deal would push Qualcomm beyond mobile into AI training/inference accelerators, following its December 2025 Ventana RISC-V buy and $2.4B Alphawave acquisition. Talks are ongoing and could still collapse.

Graph relationships

USES
AWS
Qualcomm moved its EDA workloads onto AWS and Bedrock.
Source finding
PARTNERS WITH
Amazon
Qualcomm signed a multi-generation deal to supply Amazon with customized AI inference silicon.
Source finding
BUILT BY
Snapdragon
The full Snapdragon chip lands at Snapdragon Summit in September.
Source finding
BUILT BY
Oryon
Qualcomm's Oryon CPU hits 5GHz, a mobile first.
Source finding
ACQUIRED
Modular
Modular agreed to be acquired by Qualcomm.
Source finding
PARTNERS WITH
OpenAI
OpenAI confirmed chip partnership with Qualcomm for AI hardware device.
Source finding
ACQUIRED
Modular
Qualcomm acquired Modular in all-stock deal valued at $3.92B on June 24.
Source finding
RELEASED
Dragonfly C1000
Qualcomm announced Dragonfly C1000 data-center CPUs on June 24.
Source finding
ACQUIRED
Modular
Qualcomm acquired Modular for approximately $3.9–4B on June 24, 2026.
Source finding
PARTNERS WITH
OpenAI
Qualcomm and MediaTek designing custom chip for OpenAI smartphone.
Source finding
PARTNERS WITH
OpenAI
OpenAI is developing a smartphone in collaboration with Qualcomm.
Source finding
PARTNERS WITH
Meta
Qualcomm working with Meta on wearable AI devices.
Source finding

Source trail

Qualcomm Tom's Hardware hits 5GHz agreed to acquire Modular unveiled the Dragonfly C1000 data-center CPUs Per June 16 reports GitHub GitHub BusinessWire Modular StartupHub.ai Tech Startups

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…