AI Firms Use Food Codenames for Projects
OpenAI and Broadcom on June 24, 2026 unveiled Jalapeno, OpenAI's first custom LLM-optimized inference chip, co-developed over roughly nine months. OpenAI's statement reads: "Chips are foundational to the AI economy. Building our own expands our full-stack platform from products to models to infrastructure, and will help us scale intelligence, serve more people, and expand access to AI." Broadcom confirmed the partnership in an investor relations release. The chip represents OpenAI's move toward custom silicon to reduce reliance on third-party accelerators. Business Insider also reports that AI companies increasingly use food-themed internal codenames: OpenAI model projects reportedly include garlic and shallotpeat (per The Information), Google's Gemini image model is publicly named Nano Banana Pro, and Meta reportedly used avocado for a future model.
What happened
OpenAI and Broadcom unveiled Jalapeno on June 24, 2026 - described as OpenAI's first advanced custom inference chip designed and optimized for large language models. OpenAI's official statement reads: "Chips are foundational to the AI economy. Building our own expands our full-stack platform from products to models to infrastructure, and will help us scale intelligence, serve more people, and expand access to AI." Broadcom confirmed the partnership and chip name in an investor relations release. CNBC and TechCrunch both report the chip went from design to tape-out in approximately nine months.
Why it matters
Custom silicon designed around LLM inference workloads - rather than adapted general-purpose GPUs - can deliver gains in throughput-per-watt and unit economics at scale. OpenAI's move into custom silicon follows a broader industry pattern established by Google (TPUs), Amazon (Trainium/Inferentia), and Microsoft (Maia). For practitioners and infrastructure teams, Jalapeno signals that OpenAI is pursuing vertical integration of its compute stack, which affects future pricing, capacity planning, and vendor dependency for API users.
Food codenames context
Business Insider reports the food-theme naming is a broader pattern: OpenAI model projects reportedly carry codenames garlic and shallotpeat per The Information; Google's Gemini image model is publicly released as Nano Banana Pro; and CNBC reported Meta used avocado for a future model. Food codenames are a common internal convention in silicon and software development to maintain shorthand across teams without revealing product specifics.
What to watch
Technical details on Jalapeno's architecture, node process, and deployment timeline have not been disclosed. Watch for follow-up engineering blog posts, chip vendor documentation, and OpenAI infrastructure announcements that clarify where and when Jalapeno enters production inference.
Key Points
- 1OpenAI and Broadcom unveiled Jalapeno, OpenAI's first custom LLM inference chip, on June 24 - a nine-month design-to-tape-out milestone that advances OpenAI's vertical compute integration.
- 2Custom silicon tailored to LLM inference (vs. adapted GPUs) can improve throughput-per-watt and reduce API unit costs at scale, following the path Google, Amazon, and Microsoft already took.
- 3Chip architecture, process node, and production deployment timeline have not been disclosed; practitioners should watch for follow-on engineering or infrastructure announcements.
Scoring Rationale
OpenAI's first custom inference chip co-developed with Broadcom is a confirmed, multi-sourced hardware milestone - OpenAI's own announcement, Broadcom IR release, CNBC, and TechCrunch all cover it. Custom silicon for LLM inference at OpenAI's scale has direct implications for AI infrastructure economics, vendor dependency, and long-term API pricing. Score reflects the underlying chip announcement; the Business Insider food-codenames framing dilutes narrative prominence but does not diminish the substantive news.
Sources
Primary source and supporting public references used for this report.
View 5 more sources
- OpenAI and Broadcom Unveil LLM-Optimized Inference Chipopenai.com
- OpenAI and Broadcom reveal Jalapeno, first AI chip in partnershipcnbc.com
- OpenAI unveils its first custom chip, built by Broadcomtechcrunch.com
- OpenAI and Broadcom Unveil LLM-Optimized Intelligence Processorinvestors.broadcom.com
- Broadcom and OpenAI unveil custom-built Jalapño inference processortomshardware.com
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
