OpenAI Unveils Jalapeño Chip Built for Faster, More Efficient AI Inference

·1 views

OpenAI has developed a custom inference accelerator called Jalapeño, built in partnership with Broadcom and Celestica, designed specifically to serve large language models within its own production infrastructure. The chip moved from design to tape-out in just nine months and was announced in June 2026, with quantified performance results shared in an August 2026 engineering update. OpenAI reported efficiency gains of 1.5 to 1.9 times more AI work per watt and latency reductions of up to 3.6 times compared to existing setups, though these figures are self-reported and not independently verified. Unlike general-purpose AI chips, Jalapeño was designed from scratch around OpenAI's own LLM workload patterns, treating networking as an integrated architectural element to reduce data movement. While the chip is not available for direct purchase, its deployment could improve the speed, capacity, and long-term cost economics of AI services like ChatGPT and OpenAI's APIs.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

EU Cyber Resilience Act reporting obligations become enforceable on September 11

The European Union's Cyber Resilience Act (CRA) begins enforcing its incident and vulnerability reporting requirements on September 11, 2025, ahead of the full compliance deadline of December 11, 2027. Manufacturers of 'products with digital elements' must now report actively exploited vulnerabilities within 24 hours, provide fuller notifications within 72 hours, and submit final reports within 14 days to a month. Non-compliance with reporting rules can result in fines of up to €15 million or 2.5% of global annual turnover. Pure browser-based SaaS products generally fall outside CRA scope and under NIS2 instead, but companies shipping installable components such as desktop apps, SDKs, browser extensions, or on-premise agents are likely covered. Security experts warn that many B2B SaaS firms may unknowingly fall within scope due to supplementary installable tools alongside their core web products.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Shares Python Script to Quickly Audit Any New Server or Dev Machine

A developer named Arthur has published a beginner-friendly Python tutorial on DEV Community showing how to build a Developer Environment Checker tool. The script collects key system details — including OS type, CPU cores, RAM, disk usage, Python version, and local IP address — from a single run. It relies on standard Python modules such as platform, socket, shutil, and sys, along with the lightweight third-party package psutil. The tool is designed to save time when setting up or auditing a new server, VPS, or development environment by replacing multiple manual commands. Arthur presents the project as both a practical utility and a learning exercise covering concepts relevant to real-world server administration.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

OpenAI's data center chief exits amid wave of 13 senior departures in 2026

Chris Malone, who led data center operations at OpenAI, departed the company in August 2026 after roughly 16 months in the role, according to a TechCrunch report. His exit is part of a broader pattern of senior leadership turnover at OpenAI, with at least 13 executives leaving in 2026, including its chief revenue officer, chief operating officer, and head of product. Malone's team, responsible for physical infrastructure such as power, land, and cooling, was reorganized before his departure, with oversight shifting to VP Sachin Katti. The leadership churn comes as OpenAI is central to the large-scale Stargate infrastructure build-out involving Oracle, Nvidia, SoftBank, and Microsoft, and as its IPO has been pushed from 2026 to 2027. Analysts warn that instability in infrastructure leadership could affect AI capacity timelines, with downstream effects such as rate limits, regional access gaps, and price changes for developers relying on OpenAI services.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Nuxt Endpoints module brings Hono-style RPC type safety without rewriting routes

A new Nuxt module called Nuxt Endpoints introduces Hono RPC-style type safety to Nuxt server routes without requiring any router migration or directory restructuring. Developers replace defineEventHandler with defineEndpoint, co-locating request validation, client types, and OpenAPI documentation in a single declaration. Routes retain their existing file paths, HTTP methods, and Nitro routing, remaining callable as plain HTTP endpoints. The module is designed for incremental adoption — routes without response schemas still benefit from inferred typing, just like Nuxt's built-in typed $fetch. Only routes that explicitly use defineEndpoint join the shared contract, leaving all other Nitro routes untouched.

0 comments Read more at DEV Community