Broadcom Reveals Hyperscalers' Ambitious Plans for AI Chip Deployment, Challenging Nvidia's Dominance

Apple's new BFF, Broadcom, reveals three hyperscalers want to deploy 1,000,000 GPUs or XPUs by 2027; something that will make Nvidia wince

The move comes as hyperscalers look to reduce their dependency on AI chips from Nvidia Nvidia has ridden the generative AI boom to record-breaking revenues and profits over the past two years, and while it remains well ahead of its competitors, the company is facing growing pressure - not only from rival AMD but also from hyperscalers which have traditionally relied on Nvidia GPUs but are now looking to reduce their dependence on its hardware. As The Next Platform notes, "Nvidia's biggest problem is that its biggest customers have massive enough IT expenditures that they can afford to compete with Nvidia and AMD and design their own XPUs for serial and parallel computing. And when they do so, it is chip design and manufacturing houses Broadcom and Marvell, who have vast expertise running chippery through the foundries of Taiwan Semiconductor Manufacturing Co, who will be benefiting." In its most recent earnings conference call, Hock Tan, President and CEO of Broadcom, told investors, "Specific hyperscalers have begun their respective journeys to develop their own custom AI accelerators or XPUs, as well as network these XPUs with open and scalable Ethernet connectivity. As you know, we currently have three hyper-scale customers who have developed their own multi-generational AI XPU roadmap to be deployed at varying rates over the next three years. In 2027, we believe each of them plans to deploy one million XPU clusters across a single fabric." Without naming specific companies, Tan added, "To compound this, we have been selected by two additional hyperscalers and are in advanced development for their own next-generation AI XPUs." It is widely believed that Broadcom is working with Google and Meta, and as we previously reported, with ByteDance and OpenAI on custom AI chips. Apple is also thought to be developing its first artificial intelligence server chip, codenamed "Baltra," with Broadcom providing the advanced networking technologies essential for AI processing. During the Q&A portion of the earnings call, when Tan was asked about market share, he responded, "All we are going to do is gain our fair share. We're just very well positioned today, having the best technology, very relevant in this space. We have, by far, one of the best combination technologies out there to do XPUs and to connect those XPUs. The silicon technology that enables it, we have it here in Broadcom by the boatloads, which is why we are very well positioned with these three customers of ours."

Broadcom CEO on AI investment frenzy: million-chip clusters on the horizon

Broadcom has reached a historic milestone, surpassing a US$1 trillion market valuation amid the rapid growth of the artificial intelligence (AI) market. CEO Hock Tan expressed confidence in a continued AI investment surge through the late 2020s. He noted that within three years, Broadcom's clients aim to build extensive computing clusters equipped with millions of AI chips, driving significant market growth. Earnings call recap and highlights On December 12, Broadcom posted better-than-expected fourth-quarter earnings, despite slightly underperforming in revenue. The results underscore the company's strong position in the generative AI infrastructure market, which continues to fuel robust demand. Broadcom reported a 220% year-over-year growth in AI-related revenue for its 2024 fiscal year, totaling US$12.2 billion. This surge, driven by its AI XPU and Ethernet product portfolio, underscores AI's pivotal role in shaping Broadcom's business expansion. Broadcom is collaborating with three major clients to develop AI chips, with plans to deploy 1 million chips in networked clusters by 2027, according to CEO Hock Tan. He estimated the total market opportunity for its XPUs and AI networking components at US$60 billion to US$90 billion by 2027, as reported by CNBC. While Broadcom has not officially named its chip customers, analysts indicate it is working with Google, Meta, and ByteDance to accelerate AI system training and deployment. The company has reportedly developed custom processors for these purposes, according to the Financial Times (FT). Broadcom CEO projects AI investment boom through 2029 In an interview with the FT, Tan predicted that the AI spending boom among major tech firms will persist through 2029. He revealed that Broadcom's Silicon Valley clients are intensively planning 3- to 5-year AI infrastructure investments, advancing at full speed until financial or shareholder constraints arise. xAI's "Colossus" facility in Memphis, launched in September, houses 100,000 Nvidia GPUs, setting a benchmark in AI computing power. By 2027, however, Broadcom's clients are expected to construct clusters featuring up to 1 million AI chips, according to Tan. Broadcom's AI focus overshadows speculation on Intel assistance Hock Tan, renowned for his acquisition expertise, dismissed rumors about aiding Intel during its operational struggles, clarifying that he has not been approached. He reiterated that Broadcom's focus remains firmly on the AI semiconductor market. "An agreement can only be reached if it is practicable," said Tan. "Feasibility means someone is going to come ask me. There's one thing I've learned since Qualcomm: don't make hostile offers." In 2018, then-US President Donald Trump blocked Broadcom's US$142 billion hostile takeover bid for Qualcomm, marking an unprecedented intervention, as reported by the FT. Tan's comments highlight Broadcom's conviction that AI opportunities will remain robust in the near term. Broadcom leads AI chip market with key tech partnerships Broadcom is expanding its presence in the AI hardware sector through collaborations with tech giants like OpenAI and Apple to develop custom AI server chips. As companies seek alternatives to Nvidia, the US$3 trillion leader in processors for training large language models (LLMs), Broadcom has become a key partner for AI hardware solutions. Apple is developing its first AI server processor, codenamed "Baltra," in partnership with Broadcom for networking technology. The chip, set to be manufactured using TSMC's N3P process, is expected to enter mass production by 2026, according to reports from Reuters, TechCrunch, and CEO Times. Similarly, OpenAI is working with Broadcom on AI inference processors and has secured production capacity from TSMC through this partnership, with production slated for 2026, per Bloomberg and the South China Morning Post. OpenAI has also assembled a 20-member chip development team led by a former Google engineer who contributed to Tensor Processing Units (TPUs). Beyond OpenAI and Apple, Broadcom is reportedly partnering with Google, Meta, and ByteDance to create custom processors tailored for accelerating AI training and deployment. Analysts highlight these collaborations as evidence of Broadcom's pivotal role in the AI ecosystem. Broadcom and Marvell dominate the custom chip market, which Marvell projects will grow to US$45 billion by 2028. Both companies are expected to maintain their leadership, with Marvell predicting no significant third competitor emerging in the near term. As tech giants invest heavily in AI to achieve transformative goals, such as building human-level general AI pursued by OpenAI and Anthropic, Tan views the challenges as opportunities, emphasizing the immense potential of the AI hardware market. Promising outlook for Broadcom as big tech boosts AI spending Goldman Sachs raised Broadcom's 12-month price target from US$190 to US$240 on December 15, citing growth in its custom chip client base, according to CNBC. Morgan Stanley projects that Amazon, Google, Meta, and Microsoft will collectively spend around US$300 billion on capital expenditures in 2024, with higher levels expected by 2026. Business Insider reports that much of this investment will focus on fixed assets like data centers and real estate. During the latest earnings call, Tan expressed optimism about Broadcom's AI growth, emphasizing the demand for custom AI accelerators. "Massive specific hyperscalers have begun their respective journeys to develop their own custom AI accelerators. We see an opportunity over the next three years in AI," he said, as reported by MSN.