Broadcom’s AI chip blitz: $100b revenue run, mega-deals, and a bid to outshine Nvidia

By early 2026, Broadcom's AI chip revenue was surging exponentially, with projections exceeding $100 billion for the following year—reaching a scale comparable to Nvidia's revenue three years prior. A significant driver of this growth is the Tomahawk 6 networking chip, which alone accounted for over one-third of the quarter’s AI revenue, benefiting from strong demand and a favorable margin mix that underscores Broadcom’s robust product portfolio.

Broadcom’s strategic diversification of its AI customer base, including heavyweights like Google, Meta, OpenAI, and Anthropic, has been pivotal in cementing its rising market prominence. This broad engagement not only mitigates dependency risks but also supports massive deployment scales—such as Anthropic’s anticipated 3+ gigawatt deployment generating close to $60 billion in revenue—demonstrating Broadcom’s ability to secure long-term, high-value partnerships across the AI ecosystem.

Despite emerging competition from MediaTek, particularly around Google’s v7 inferencing chip, Broadcom’s two decades of technology leadership and execution expertise position it strongly to capture leading market share within the expanding AI customer base. This resilience, combined with a 100% increase in AI-related revenue and a market capitalization surpassing $1.5 trillion, firmly establishes Broadcom alongside Nvidia as a dominant force in the AI chip sector.

Broadcom CEO Hock Tan’s projection of AI chip revenue exceeding $100 billion by 2027 highlights not only the company’s rapid growth trajectory but also its massive, yet often underappreciated, role in AI infrastructure. This underscores Broadcom’s expanding prominence as a foundational player in the AI chip ecosystem, complementing the more visible AI compute leaders and signaling a strategic leadership position in the broader AI market landscape.

By early 2026, Broadcom had strategically fortified its supply chain to navigate the persistent industry-wide shortages of critical components like HBM memory, securing supply agreements through 2028 with key Japanese suppliers outside Taiwan. This proactive approach, initiated in 2024 under CEO Hawk's direction, not only circumvented long wait times and inflated prices but also reinforced Broadcom's reputation for reliability among customers seeking stability amid global supply disruptions.

Beyond merely ensuring availability, Broadcom has enhanced its value proposition by offering tailored customization of memory products to boost performance, a capability that distinguishes it as a flexible and dependable supplier. This ability to 'tweak the product for you,' as noted in 2026 analyses, allows Broadcom to meet specific customer needs even amid constrained supply conditions, thereby deepening client trust and strengthening its foothold in the AI chip ecosystem.

By early 2026, Broadcom had cemented its strategic foothold in the AI hardware ecosystem through multi-year agreements with tech giants Google and Anthropic, committing to supply custom AI chips and TPU compute capacity through 2031. These partnerships not only underscore Broadcom’s expanding role in powering generative AI workloads but also mark a significant scaling of compute infrastructure, with Anthropic alone projected to demand over 3.5 gigawatts of TPU capacity starting in 2027.

The financial stakes of these deals are monumental, with analysts projecting Broadcom could generate up to $42 billion in AI revenue from Anthropic in 2027, highlighting the lucrative nature of long-term AI compute supply contracts. This surge in demand for Broadcom’s custom silicon reflects a broader industry shift seeking alternatives to Nvidia’s GPU dominance, offering Broadcom enhanced revenue visibility amid intensifying competition in the AI accelerator market.

By early 2026, Anthropic's revenue growth has been nothing short of meteoric, skyrocketing from approximately $9 billion ARR at the end of 2025 to over $30 billion ARR within just a few months. This surge is underpinned by a rapidly expanding customer base, with the number of business clients spending more than $1 million annually doubling to over 1,000, reflecting the broad market adoption of its AI model Claude. This explosive growth has been strategically supported by Anthropic's diversified multi-cloud, multi-chip compute partnerships, including deals with Google for TPUs, AWS for Trainiums, and anticipated collaborations with AMD, which collectively enhance Anthropic’s leverage against competitors like Nvidia and foster a resilient infrastructure ecosystem.

To address its critical compute capacity crunch amid rapid scaling, Anthropic has secured a landmark multi-gigawatt TPU compute agreement with Broadcom and Google, commencing in 2027 and extending through 2031. This massive infrastructure commitment, involving approximately 3.5 gigawatts of next-generation TPU capacity and an estimated $175 billion in capital expenditure, not only guarantees Anthropic’s long-term access to cutting-edge AI compute resources but also exemplifies Broadcom’s strategic positioning as a dominant custom silicon provider alongside hyperscalers. However, this partnership has also introduced competitive tension within Google itself, as Anthropic’s TPU consumption has constrained Google Cloud’s own Gemini AI development, illustrating the high-stakes resource competition in the AI chip ecosystem.

Anthropic’s exploration of in-house AI chip design signals a forward-thinking response to persistent hardware shortages and a desire for greater control over its compute destiny. While still in nascent stages without a dedicated design team, this initiative aligns with a broader industry trend among AI leaders like Meta and OpenAI, who are investing heavily—upwards of half a billion dollars—to develop proprietary AI silicon. This strategic move complements Anthropic’s multi-cloud, multi-chip approach, leveraging its collaborative ecosystem of partners including Google, Amazon, and Broadcom, and underscores CFO Krishna’s expanded role in orchestrating not just financial stewardship but also compute capacity strategy to sustain Anthropic’s rapid growth trajectory.

Google’s 14% ownership stake in Anthropic and its preference for Anthropic to run on Google Cloud TPUs rather than competitors like OpenAI using Nvidia GPUs outside GCP highlights a nuanced strategic alliance within the AI ecosystem. While Google aims to dominate with its own Gemini AI, it pragmatically supports Anthropic’s ascendancy on its TPU infrastructure as a preferable alternative, effectively hedging its bets in the competitive AI landscape. This dynamic illustrates how Anthropic’s collaborative ecosystem not only fuels its growth but also plays a pivotal role in shaping cloud providers’ strategic positioning and chip deployment decisions in the rapidly evolving AI market.

Meta's $2.3 billion deal with Broadcom in 2025 marks a decisive strategic shift toward AI chip independence, aiming to reduce reliance on Nvidia's vertically integrated hardware dominance. By partnering with Broadcom, which embraces an open-source hardware model, Meta is positioning itself within a competitive landscape defined by two contrasting approaches: Broadcom's open ecosystem versus Nvidia's closed system. This move not only signals Meta's intent to control its AI infrastructure but also highlights Broadcom's rising role as a credible alternative in the AI chip market.

Mirroring Amazon's strategy of building custom infrastructure to handle vast data and user bases, Meta is leveraging its scale to develop proprietary AI chips that could support both internal applications and external services. As one analysis put it, Meta is effectively 'pulling an Amazon move' by transforming its massive data advantage into bespoke hardware solutions, potentially opening new avenues for monetization beyond its core platforms.

Amid soaring demand and supply constraints in the AI chip sector, Meta's substantial investment underscores the critical importance of securing dedicated compute resources. With Nvidia unable to meet all market needs, Meta's cash-rich position enables it to forge strategic deals like the Broadcom partnership, ensuring prioritized access to the compute power essential for scaling its AI ambitions.

By early May 2026, Broadcom's dominance in the AI semiconductor space was unmistakable, with its AI chip revenue soaring 106% year-over-year to $8.4 billion, underpinned by a massive $73 billion backlog from hyperscale clients like Google, Meta, and Anthropic. This robust demand and multi-year revenue visibility exceeding $100 billion by 2027, combined with industry-leading profitability margins of 51%, solidify Broadcom's position as a powerhouse in custom ASICs and networking solutions for AI data centers.

The market's confidence in Broadcom's AI chip leadership was vividly reflected in its stock performance, which surged from around $314 to over $350 following the announcement of multi-year partnerships with top hyperscalers. This rally underscores investor optimism about the company's growth trajectory amid accelerating AI adoption, even as insider trading data reveals a contrasting narrative with 211 sales versus a single purchase in the past six months, including significant share disposals by top executives, suggesting nuanced internal perspectives on valuation or liquidity needs.