Broadcom’s expanding role in the artificial intelligence supply chain took an intriguing turn this week after company executive Charlie Kawwas denied that OpenAI was the mystery $10 billion customer revealed during the firm’s third-quarter earnings report. The clarification came as both companies announced a significant partnership to co-develop and deploy OpenAI-designed AI accelerators, underscoring the accelerating race to build next-generation computing infrastructure.

A Clarification Amid a Wave of AI Speculation

Appearing on CNBC’s Squawk on the Street alongside OpenAI President Greg Brockman, Kawwas, who serves as president of Broadcom’s semiconductor solutions group, sought to dispel growing market rumors. “I would love to take a $10 billion purchase order from my good friend Greg,” Kawwas said with a smile, “but he has not given me that PO yet.”

The denial comes after weeks of speculation that OpenAI — now valued at roughly $500 billion — was the unnamed customer responsible for Broadcom’s reported $10 billion custom chip order, a figure that significantly boosted Broadcom’s AI revenue outlook for 2025. Analysts had viewed OpenAI as the most likely candidate, given its massive compute demands and expanding infrastructure deals with chipmakers including Nvidia, AMD, and CoreWeave.

Despite the clarification, Broadcom’s partnership with OpenAI remains strategically important. The two firms will work together over the next four years to build racks of custom AI chips, with initial deployment set to begin in late 2026 and full-scale rollout expected by 2029.

Broadcom’s Expanding AI Footprint

Broadcom has quietly become one of the most critical players in the global AI supply chain. Its custom silicon and networking technologies power data centers operated by major technology firms including Google, Meta, and ByteDance — all of which are now investing aggressively in AI infrastructure.

During its September earnings call, Broadcom CEO Hock Tan said the new $10 billion order would come from a “large hyperscale customer” — a term typically used to describe global cloud providers with immense data-processing requirements. Analysts now believe the mystery client is likely one of Broadcom’s existing partners, though the company has declined to confirm any names.

This influx of demand has sharply altered Broadcom’s revenue mix. AI-related business now accounts for a growing share of its semiconductor segment, helping offset slower growth in traditional networking and broadband units. The company expects AI-related revenue to grow by double digits in 2025, driven by expanding orders for custom accelerators and networking gear.

OpenAI’s Infrastructure Ambitions

For OpenAI, the partnership marks another milestone in its bid to control more of its own technological stack. “By building our own chip, we can embed what we’ve learned from creating frontier models and products directly into the hardware,” Brockman said in a press release. The move reflects OpenAI’s strategic pivot toward vertical integration — reducing dependency on suppliers like Nvidia and optimizing performance for its large-scale AI models.

The collaboration also underscores the broader transformation of the semiconductor landscape, where custom silicon has become a key competitive differentiator. As AI workloads grow more complex, major players are increasingly designing their own chips tailored to their unique software ecosystems.

What Lies Ahead for the AI Hardware Race

Even as Broadcom’s mystery client remains unidentified, the announcement highlights a deeper trend — the consolidation of AI infrastructure among a handful of global technology giants. With companies racing to secure limited chip supply and accelerate innovation, the balance of power in the semiconductor industry is rapidly shifting toward those capable of building both software and hardware at scale.

Looking ahead, investors will be watching closely to see whether Broadcom can maintain its growth trajectory in AI infrastructure and whether OpenAI’s hardware ambitions yield meaningful competitive advantages. What’s clear is that the next phase of the AI revolution will be defined not just by algorithms — but by the chips that power them.