Intel's new GPU packs up to 480GB of memory — but skips 3D graphics
Intel launched the D270 GPU accelerator, part of its D-series, designed for AI inference with up to 480GB of LPDDR5X memory. The card, based on Xe3P architecture, offers 1.5TB/s memory bandwidth and 1.3 petaflops peak FP8 performance. It targets large language models and AI services, with customer sampling underway. Intel has not disclosed pricing or availability.
How this was made
The 30-second read
Why it matters
The D270's high memory capacity addresses a niche for trillion‑parameter models, potentially opening new data‑center revenue streams.
Market read
First‑time disclosure of a major AI‑focused GPU could influence Intel's stock and the broader AI‑hardware sector.
What to watch
Intel's choice of LPDDR5X over HBM could affect performance perception among AI developers.
Background
Intel is expanding its Xe architecture portfolio beyond graphics into AI inference, targeting large language model workloads.
Ticker impact
Intel unveiled its first D-series accelerator, the D270, with up to 480GB of LPDDR5X memory for AI inference.
likely modest upside as the market prices in potential AI‑accelerator demand
Product specs are new but no pricing or launch date is given, limiting immediate trading impact.
Market effects
Signals increased competition in the AI accelerator market, potentially pressuring rivals like Nvidia and AMD.
May benefit US data‑center providers adopting Intel hardware.
Adds to the broader AI‑hardware supply narrative influencing global tech equities.
Counterpoint
Without pricing or availability details, the announcement may not translate into near‑term revenue, limiting upside.
Key entities
- CompanyIntel
US‑listed semiconductor manufacturer launching the D270 AI accelerator.
