AI Research
Samsung's In-Memory AI Compute Needs the CPU's Caches Switched Off
The eight-times bandwidth figure from Hot Chips is measured inside the package. Using it costs the host processor its cache, prefetch and out-of-order execution.
1d ago
#inference
AI Chips
d-Matrix Claims 20x HBM4 Bandwidth Density, Names No Date
A challenger architecture aimed squarely at the wall Micron spent the previous session describing.
6d ago
AI Chips
Cerebras' CS-4 Hits 750 PFLOPS With Three Wafers and a Better Cooling System
The headline number is a package count, and the per-wafer gain comes from pushing more power through existing 5nm silicon.
Aug 19, 2026
AI Chips
Etched Raised $700m at $21B, Led By Its Only Named Customer
A valuation that roughly doubled in 26 days, on one shipped rack — delivered to the firm that led the round.
Aug 19, 2026
AI Chips
OpenAI Is Serving a Frontier Model on Cerebras Silicon at 750 Tokens a Second
Ultrafast runs GPT-5.6 Sol up to 14x faster than standard inference. It is the first time OpenAI has publicly put a frontier model on someone else's accelerator.
Aug 14, 2026
AI Chips
Alibaba's 64-Card Supernode Went Commercial: Interconnect Is Chinese
The M890 is rated for inference on models up to 10 trillion parameters and already serves Kimi K3 and Qwen3.8-Max. Alibaba has not said what silicon is inside.
Aug 13, 2026
AI Models
OpenAI Cut Its Cheapest Model 80% and Says Its Best One Rewrote the Serving Stack
Luna is down 80%, Terra 20%, and the flagship is unchanged. The efficiency gain OpenAI actually measured was 20%.
Jul 31, 2026
AI Models
A Spanish Startup That Shrinks Models by 95% Wants €500 Million for It
Multiverse Computing opened a Series C targeting up to $570 million at a $1.7 billion pre-money valuation — five times its last round, and not yet closed.
Jul 27, 2026
AI Chips
Etched Doubles to a $10.3 Billion Valuation Three Weeks After Showing Its Silicon
Sequoia led a $300 million Series C into an inference-only chip company that every major investor turned down in 2023 — and SK hynix, the world's leading AI-memory maker, came along.
Jul 24, 2026
Google AI
Google Is Designing a Chip With Gemini's Architecture Baked Into the Silicon, Report Says
A server chip codenamed 'Frozen v2' would freeze parts of Google's model design directly into hardware for far more efficient inference, targeted for 2028, according to The Information.
Jul 20, 2026
AI Chips
AI Chip Startup Etched Is in Talks for a $20 Billion Valuation
The transformer-chip maker is negotiating two concurrent rounds — one near $20 billion, another near $10 billion — that would quadruple the $5 billion valuation it held just months ago, though neither has closed.
Jul 19, 2026
AI Chips
A $400M Loan Backed by Inference Chips: Upper90 Bets General Compute Against Nvidia's Grip
The firm that financed Crusoe's first GPUs is now lending against SambaNova silicon — in what may be the first major credit deal collateralized by non-Nvidia inference hardware.
Jul 17, 2026
AI Chips
SambaNova Raises $1B at an $11B Valuation as Inference Demand Surges
The chip startup's first close of a Series F, led by General Atlantic, values it at $11 billion — a sharp rebound built on enterprises like JPMorgan Chase buying Nvidia alternatives to run AI in-house.
Jul 9, 2026
