inference

Etched unveiled working A0 silicon for Sohu, its transformer-only inference ASIC, disclosing $800 million raised at a $5 billion valuation and more than $1 billion in signed customer contracts.

AWS added Grok 4.3 to Bedrock on June 15, making xAI a model provider on the platform and debuting a separate Mantle inference path.

OpenAI's first custom silicon, the Broadcom-built Jalapeño inference accelerator, went from design to tape-out in nine months and targets deployment by the end of 2026.

The startup says its 3nm Napier inference chip turns matrix multipliers into adders, with commercial sales of a 72-chip system slated for the second half of 2027.

d-Matrix says its Corsair inference accelerator is in full production, pairing on-chip SRAM with GPUs to cut AI response times by more than 10x.

A new MIT CSAIL inference method has a model inspect its prompt in a Python REPL and recursively call itself over snippets, processing inputs beyond its context window.

London-based Fractile closes a $220M Series B led by Accel, Founders Fund, and Factorial Funds to tape out SRAM-based inference chips promising radically faster and cheaper AI output generation.

Cloudflare detailed the custom infrastructure behind its Workers AI expansion to host models like Kimi K2.5, including a Rust-based inference engine and a lossless weight-compression system.

MLCommons' April 2026 inference benchmark round adds text-to-video and vision-language tests while NVIDIA posts a 2.7x jump on DeepSeek-R1 through software alone.

Alibaba's Damo Academy has announced the XuanTie C950, a 5nm RISC-V CPU core capable of natively running hundred-billion-parameter language models, marking RISC-V's first serious entry into high-end AI computing.

Google Research unveils TurboQuant, a training-free compression algorithm that reduces LLM key-value cache memory by 6x and boosts inference throughput up to 8x, sending SK Hynix and Samsung shares down 5-6 percent.

The five largest US cloud providers have committed up to 690 billion dollars in 2026 capex, nearly doubling 2025 levels, as inference workloads surpass training to consume 60 to 70 percent of AI compute demand.