OpenAI’s first custom AI chip is no longer just a report: OpenAI and Broadcom unveiled it on June 24, 2026, under the name Jalapeño. It is an accelerator designed primarily to run large language models, not a consumer processor or an announced replacement for all of OpenAI’s Nvidia hardware. Initial data-center deployment is planned by the end of 2026.
What OpenAI announced
OpenAI calls Jalapeño its first “Intelligence Processor.” It is a specialized accelerator—hardware built to perform AI workloads—rather than a general-purpose CPU. OpenAI says the design starts from the needs of its models and products, including LLM kernels, memory movement, networking and serving patterns. The announcement describes a first generation of a broader compute platform, not a one-off retail chip. OpenAI’s June 24, 2026 announcement provides the company’s account.
The project was first reported in 2024, when Reuters reported that OpenAI was working with Broadcom and TSMC on a custom chip. Reuters reported in February 2025 that OpenAI was nearing completion of the design. The official unveiling means the old “reportedly working on” framing is now outdated: the chip exists as an announced project, though broad production and deployment are still ahead. Reuters’s October 2024 report and its February 2025 report document those earlier stages.
What Jalapeño is designed to do
Inference is the disclosed focus
Inference is what happens when a trained model processes a prompt and generates an answer, code, or other output. OpenAI’s announcement emphasizes serving interactive LLM products, where latency, throughput, energy use and hardware utilization affect both responsiveness and operating costs. OpenAI has not established that this first-generation processor will replace the GPUs used to train its frontier models or handle every kind of AI workload.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →#1 Best Overall
![Arduino® UNO™ Q 4GB [ABX00173]- Hybrid Board, Qualcomm Dragonwing QRB2210 microprocessor (MPU) & STM32U585 Microcontroller(MCU), AI Vision, Voice, IoT, Robotics, Linux Debian OS, Wi-Fi 5, USB-C](https://m.media-amazon.com/images/I/51cl4xygwGL._SL160_.jpg)
- Dual-Brain Hybrid Power: Combines the Qualcomm Dragonwing QRB2210 MPU (Quad-core Arm Cortex-A53 @ 2.0 GHz CPU, Adreno GPU, AI acceleration) and the real-time, low-power STM32U585 MCU for advanced applications like object recognition, voice commands, and motion detection.
- AI & Linux Capabilities: Unlocks AI-powered vision and sound solutions; runs Linux Debian OS for coding in Python and supports the Arduino ecosystem with libraries and Sketches; quick start with Arduino App Lab.
- Advanced Features: Equipped with 4 GB LPDDR4 RAM, 32 GB eMMC built-in storage, ideal for single-board computer (SBC) mode, running multiple simultaneous high-level processes, more complex AI or ML models, extensive logs. Dual-band Wi-Fi 5 (2.4/5 GHz), Bluetooth 5.1, and high-speed headers for vision, audio, and display peripherals.
- Seamless Expansion & Connectivity: Features the classic UNO form factor for shields compatibility, an 8x13 LED matrix, and a Qwiic connector for easy expansion with Modulino nodes; power and connect via the USB-C connector.
- Intended Use & Development: The perfect platform for prototyping robotics or IoT projects, empowering innovators with a unified development experience to mix Arduino Sketches, Python scripts, and containerized AI models in a single interface.
Why custom hardware may matter
A general-purpose accelerator must serve many customers and workloads. A custom design can instead be tuned to a company’s own models, kernels, memory access, serving software and latency targets. If the full system performs well, OpenAI could gain more control over capacity and improve efficiency for selected workloads. That does not guarantee lower prices for ChatGPT or the API: OpenAI has published no per-token savings, cost reduction or pricing change tied to Jalapeño.
Who is building and making it
“In-house” describes OpenAI’s role in designing a processor around its own needs; it does not mean OpenAI manufactures every component itself. OpenAI says it designed the accelerator with Broadcom, which contributes semiconductor implementation, networking and connectivity. Celestica contributes board, rack and system integration expertise. Earlier Reuters reporting identified TSMC as the intended fabrication partner for the first custom chip; the June 2026 announcement’s central named partners are OpenAI, Broadcom and Celestica.
Rank #2
- Dual-Brain Hybrid Power: Combines the Qualcomm Dragonwing QRB2210 MPU (Quad-core Arm Cortex-A53 @ 2.0 GHz CPU, Adreno GPU, AI acceleration) and the real-time, low-power STM32U585 MCU for advanced applications like object recognition, voice commands, and motion detection.
- AI & Linux Capabilities: Unlocks AI-powered vision and sound solutions; runs Linux Debian OS for coding in Python and supports the Arduino ecosystem with libraries and Sketches; quick start with Arduino App Lab.
- Advanced Features: Equipped with 2 GB LPDDR4 RAM, 16 GB eMMC built-in storage, ideal to develop in PC-connected mode, running the OS, Python scripts, and basic network services (SSH) without a demanding GUI or heavy multitasking; great for lightweight AI and memory-optimized TinyML applications, needing local storage for basic OS and core libraries. Dual-band Wi-Fi 5 (2.4/5 GHz), Bluetooth 5.1, and high-speed headers for vision, audio, and display peripherals.
- Seamless Expansion & Connectivity: Features the classic UNO form factor for shields compatibility, an 8x13 LED matrix, and a Qwiic connector for easy expansion with Modulino nodes; power and connect via the USB-C connector.
- Intended Use & Development: The perfect platform for prototyping robotics or IoT projects, empowering innovators with a unified development experience to mix Arduino Sketches, Python scripts, and containerized AI models in a single interface.
A precise description is OpenAI’s first custom-designed AI accelerator, developed and industrialized with Broadcom and other manufacturing and systems partners. This is not evidence that OpenAI has become a vertically integrated chip manufacturer.
What is known about performance—and what is not
OpenAI and Broadcom say engineering samples have run machine-learning workloads in the lab at production-target frequency and power, including GPT‑5.3‑Codex‑Spark. They also claim early testing shows substantially better performance per watt than the current state of the art. Those are company-reported results, not independently verified benchmarks. The companies say OpenAI models helped accelerate parts of design and optimization and describe a nine-month path from initial design to manufacturing tape-out; that timeline is likewise a company claim, not an independently established industry record.
Rank #3
- Single core ARM Cortex-A7 32-bit core, integrated with NEON and FPU
- Built in Micro's self-developed 4th generation NPU, with high computational accuracy and support for mixed quantization of int4, int8, and int16. Among them, int8 has a computing power of 0.5 TOPS and int4 has a computing power of up to 1.0 TOPS
- Built in self-developed 3rd generation ISP3.2, supports 4 million pixels, and supports various image enhancement and correction algorithms such as HDR, WDR, and multi-level denoisin
- It has powerful encoding performance, supports intelligent encoding, adapts to save bit rates according to the scene, and saves more than 50% of the bit rate compared to conventional CBR mode, making the captured images high-definition, smaller in size, and doubling the storage space
- The design with built-in RISC-V MCU supports low-power fast startup, 250ms fast capture, and simultaneous loading of AI model library, enabling facial recognition to be completed within 1 second
OpenAI has not yet released the detailed technical report it says will follow. The public announcement does not provide verified figures for process node, transistor count, memory capacity or bandwidth, FLOPS or TOPS, benchmark scores, manufacturing yield, production volume, or per-chip cost. Tape-out—the point at which a design is sent for fabrication—is not the same as proven mass production: yield, packaging, system integration and reliable operation at scale still matter.
Why OpenAI wants custom silicon
- More supply options: Custom capacity could reduce dependence on a single accelerator supplier at a time when AI infrastructure demand is high. It does not show that OpenAI is leaving Nvidia.
- Workload-specific efficiency: Hardware designed alongside models and serving software may improve performance per watt or latency on OpenAI workloads. The size and real-world breadth of any improvement remain undisclosed.
- Operating economics: Efficient inference could lower the cost of serving products such as ChatGPT, Codex and the API, but no savings figure or customer price reduction has been announced.
- System-level control: Coordinating chips, memory, networking, racks and serving software gives OpenAI more influence over how its infrastructure is built and deployed.
The approach also carries risk. A design optimized for today’s model patterns may be less compelling if workloads change; software such as compilers, kernels and runtimes must perform well; and advanced packaging, memory, networking, power, cooling and data-center construction can constrain capacity even when a chip design is ready. Strong results on selected internal workloads would not automatically transfer to every model or outside customer’s software.
Rank #4
- 【POWERFUL ESP32‑S3 CONTROLLER】Built‑in Xtensa 32‑bit LX7 dual‑core processor, 512KB SRAM, 8MB PSRAM, 16MB Flash for stable AI voice computing and multitask processing.
- 【Preloaded Dual AI Platforms】Comespre-installed with complete Deepseek and OpenAI voice dialogue projects.Experience intelligent voice interaction instantly. (Note: OpenAI functionality requires your own API key.)
- 【STABLE WIRELESS & CLEAR AUDIO】Integrated 2.4GHz Wi‑Fi + Bluetooth 5 (LE); dedicated audio decoding module for natural, responsive voice interaction.
- 【USER‑FRIENDLY VISUAL & PLUG‑AND‑PLAY】2” TFT‑SPI color screen shows real‑time chat; modular design, no extra wiring, ready to use after setup.
- 【FULL LEARNING SUPPORT】45 programmable GPIOs, rich interfaces, online web tutorials, free technical support for beginners & developers.
Timeline: from report to planned deployment
| Date | Milestone |
|---|---|
| October 2024 | Reuters reports that OpenAI is working with Broadcom and TSMC on a first custom AI chip, with production targeted for 2026. Report |
| February 12, 2025 | Reuters reports that OpenAI is nearing completion of its first chip design and plans to send it to TSMC for fabrication. Report |
| October 13, 2025 | OpenAI and Broadcom announce a plan for 10 gigawatts of OpenAI-designed AI accelerators, with deployment targeted to start in the second half of 2026 and finish by the end of 2029. OpenAI announcement |
| June 24, 2026 | OpenAI and Broadcom publicly unveil Jalapeño. Announcement |
| By the end of 2026 | Initial deployment of the Jalapeño-based platform is planned, according to OpenAI and Broadcom; this is a target, not confirmation of general availability. |
| By the end of 2029 | The previously announced 10-gigawatt accelerator deployment is targeted for completion. Broadcom announcement |
The 10-gigawatt figure refers to planned accelerator and networking infrastructure scale, expressed as power capacity—not a stated number of chips. Converting it to a processor count would require details such as chip and rack power, cooling configuration and utilization that have not been disclosed.
Does this mean OpenAI will replace Nvidia?
No. OpenAI has announced a custom platform, not a complete Nvidia exit. Jalapeño’s disclosed inference focus makes it plausible that custom silicon could serve particular workloads alongside other accelerators, while training and other tasks continue to use different hardware. OpenAI has also been reported to use or evaluate alternatives such as AMD and specialized inference hardware. Nvidia’s position includes a mature software, networking and developer ecosystem as well as chips; any alternative must work economically and reliably across the whole system, not just post a promising silicon result.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Can customers buy or rent Jalapeño?
There is no announced retail product, standalone card, licensing offer or self-service cloud instance for Jalapeño. OpenAI’s disclosures concern its own data-center platform. A business or developer seeking OpenAI models can use the OpenAI API; those customers consume model services rather than provision Jalapeño hardware. Organizations that need externally available accelerator infrastructure would need to evaluate commercial cloud or hardware offerings separately—the announced OpenAI chip is not one of them.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
