First screen, captured 9 days ago. Open the video.

Qualcomm Cloud AI 100 Ultra – 150W AI Accelerator Card with 128GB at 548GB/s

What HN thinks

A two-year-old card priced at one cent, sold out in the 63-cent variant, and somehow still the most interesting thing in the room.

Specs feel dated and unimpressive
Many note the card launched in 2023-2024 and its 548GB/s bandwidth is only M4 Max territory, with cloud rental pricing worse than renting an H200.
High compute suits prefill and batching
Others argue the 800 TOPS with large capacity gives high arithmetic intensity, good for prefill, embeddings, image generation and batched decode, at an attractive 150W.
Any inference hardware is welcome
With old GPUs like V100s and MI50s fetching high prices, some see value for businesses priced out of local inference, if the price is right.

Consensus: Most agree memory bandwidth limits decode performance and that the card is not cutting edge; whether it's worthwhile depends on an unknown price.

Recommended in the comments

AMD Instinct MI350P (product): Suggested as offering far more compute and bandwidth for those going down this path.
NVIDIA RTX PRO 6000 Blackwell (product): Recommended for much more compute and bandwidth, and can be underclocked efficiently.
NVIDIA H200 (product): Rentable at a similar hourly price while being better in every way.

Summarised by Claude from the comment thread 8 days ago.

Points: 164
Comments: 60
Best rank: #6
Posted: 2026-11-04 by peter_d_sherman
Category: Hardware

qualcomm cloud-ai-100-ultra ai-accelerator inference data-center llm

All captures:

Full page

Captured 9 days ago. Click for the whole screenshot.

Looks similar

Screenshot of altera.com: Altera Announces Agilex 2X, a Complete FPGA and SoC Portfolio

Altera Announces Agilex 2X, a Complete FPGA and SoC Portfolio (altera.com)

50 points by gmays 8 days ago | 16 comments

Altera press release announcing the Agilex 2X FPGA and SoC portfolio, led by the Agilex 27 family, with QuartusIQ AI design tooling support.

fpga altera agilex soc

Screenshot of blogs.nvidia.com: H100 GPUs Set Standard for Gen AI in Debut MLPerf Benchmark

H100 GPUs Set Standard for Gen AI in Debut MLPerf Benchmark (blogs.nvidia.com)

137 points by Anon84 3 years ago | 62 comments

Screenshot of intel.com: Intel D270 GPU AI Accelerator–Up to 480 GB LPDDR5X

Intel D270 GPU AI Accelerator–Up to 480 GB LPDDR5X (intel.com)

293 points by hurricanepootis 1 month ago | 168 comments

Intel product page for its data center GPU D-series AI inference accelerators, including the D270 with up to 480 GB LPDDR5X memory.

intel gpu ai-accelerator inference

Screenshot of newsroom.ibm.com: IBM Quantum Nighthawk Demonstrates New Levels of Speed and Quality

IBM Quantum Nighthawk Demonstrates New Levels of Speed and Quality (newsroom.ibm.com)

50 points by gok 2 days ago | 24 comments

IBM press release announcing the Nighthawk r3 quantum processor (520 qubits, 10,000 operations), Nighthawk r2 results, and plans for a 1,000+ qubit system in 2027.

quantum-computing ibm-quantum nighthawk qubits

Screenshot of intel.com: Intel: New Core Ultra Processors Deliver Breakthrough Performance

Intel: New Core Ultra Processors Deliver Breakthrough Performance (intel.com)

92 points by htk 2 years ago | 88 comments

Screenshot of github.com: Flawless UI-UX Skill 1.0.0

Flawless UI-UX Skill 1.0.0 (github.com)

15 points by 99lcccss 7 days ago | 19 comments

GitHub release notes for flawless-ui-ux v1.0.0, a UI/UX design skill for AI coding agents adding an interactive installer and support for twelve agents.

ai-coding-agents agent-skills ui-ux-design claude-code