OpenAI’s move into custom silicon is reshaping the AI hardware race, promising cheaper, more efficient model deployments while quietly turning up competitive pressure on longtime supplier Nvidia.

In late 2025, OpenAI and Broadcom began co-developing a custom accelerator tailored to large language model (LLM) inference, part of OpenAI’s bid to control the “full stack” behind its products.¹ Over nine months, the team designed an ASIC dubbed “Jalapeño,” built specifically to run models like ChatGPT and Codex rather than train them.¹ OpenAI says early lab tests show “performance per watt substantially better than current state-of-the-art,” and that engineering samples are already running workloads including GPT‑5.3‑Codex‑Spark.¹

On June 22, 2026, OpenAI formally unveiled Jalapeño as its first “Intelligence Processor,” the opening chip in a “multi-generation compute platform” meant to make advanced AI “faster, more reliable, and more accessible to more people.”¹ Broadcom contributed silicon implementation and high-speed networking, positioning the chip for large data-center deployments.¹

By June 24, reporting framed Jalapeño as part of a broader effort to reduce dependence on Nvidia and secure more capacity amid a global compute crunch.² Axios noted OpenAI had begun testing the “first in a family of homegrown chips,” with commercial use at Microsoft and other partners expected by year-end and larger rollouts in 2027.² OpenAI aims to have its custom chips powering 10 gigawatts of compute by 2029.²

External analysts highlighted competitive stakes. The Verge reported Broadcom CEO Hock Tan claims Jalapeño “matches the performance of Nvidia’s Blackwell chips and Google’s Tensor processing units,” while emphasizing its efficiency gains.³ Ars Technica cast the chip as the first step in a long-term, vertically integrated platform designed to “squeeze out more capacity amid a global compute crunch.”⁴

By June 26, TechCrunch described Jalapeño as “Big Tech’s spiciest move away from Nvidia,” part of an industry-wide shift as OpenAI joins Google, Apple, Amazon, Meta, and SpaceX in building custom chips to manage single-supplier risk and tune hardware to specific AI workloads.⁵

On social media, OpenAI leaders amplified the technical and cultural stakes. President Greg Brockman wrote that Jalapeño was “designed from scratch for LLM inference over nine months” and that performance per watt was “looking incredible,”Greg Brockman (gdb)⁶ while CEO Sam Altman offered a terse verdict on the team’s work: “team cooked, spicily.”Sam Altman (sama)⁷

Meanwhile, some in the open-source community argued that this centralization only strengthens the case for decentralized alternatives. Hugging Face CEO Clément Delangue reposted a claim that “Local and Opensource AI are going to win,” suggesting that as giants move into highly specialized data-center silicon, others will double down on edge and open models.clem 🤗 (ClementDelangue)⁸

Across these perspectives, Jalapeño marks both a technical milestone and a strategic inflection point: a bid by OpenAI to secure its own compute future while intensifying the race to define the next generation of AI infrastructure.