On July 6, 2026, Broadcom (AVGO.US) submitted a filing to the US Securities and Exchange Commission that sent ripples through the entire semiconductor investment community. The document revealed that Broadcom and Apple (AAPL.US) had signed a new multi-year agreement, extending their long-term technology partnership through 2031. Under this deal, Broadcom will develop and supply Apple with a range of custom ASIC (Application-Specific Integrated Circuit) chip products for multiple generations of Apple devices.
The day after the announcement (Beijing time, July 7), Broadcom’s stock closed up 3.73% at $373.90, after surging as much as 6.3% to $383.16 during the session. This rally not only reflected the market’s optimism about the deal itself but also signaled a broader revaluation of the long-term value of the custom chip sector in the capital markets.
On the same day, the strategic positioning of traditional tech giants and volatility in crypto asset prices seemed to resonate. Is there a deeper industrial logic connecting these events? This article takes a closer look at the Broadcom-Apple partnership extension, breaks down the competitive landscape of the custom chip sector, and explores the "second curve" of AI infrastructure investment.
Broadcom and Apple Extend Partnership to 2031: Three Key Takeaways from the Agreement
Broadcom and Apple’s collaboration dates back nearly a decade. In January 2020, the two companies announced a wireless components supply agreement worth roughly $15 billion. In May 2023, they signed another multi-billion-dollar, three-year deal for Broadcom to develop and manufacture 5G RF components. This latest extension to 2031 is not just about longevity—it marks a significant deepening of their partnership.
First signal: Upgrading from RF components to custom ASIC capabilities. Broadcom has long supplied Apple with custom RF chips for iPhone, Wi-Fi and Bluetooth connectivity chips, and other network semiconductors. The core shift in the new agreement is the expansion from traditional RF components to custom ASIC chips. ASICs are custom-designed for specific application scenarios and are seeing surging demand in AI inference and high-performance computing. This move elevates Broadcom’s role in Apple’s supply chain from a "connectivity component supplier" to a "compute chip co-design partner."
Second signal: Official endorsement of Apple’s "Baltra" AI chip. Market reports had already suggested that Apple was working with Broadcom to develop an AI server processor codenamed "Baltra," expected to use TSMC’s N3P process and enter mass production in 2026. Bank of America Securities previously noted that Broadcom’s fifth new AI ASIC customer "could be Apple." The signing of this new agreement confirms those suspicions. With Apple on board, Broadcom’s AI ASIC client roster now includes Google, Meta, ByteDance, OpenAI, and Apple—the five tech giants.
Third signal: Locking in 20% of annual revenue with certainty. Analysts estimate that Apple accounts for about 20% of Broadcom’s annual revenue, making it Broadcom’s most important customer. Previously, there were concerns that Apple’s ongoing push into in-house chip development (including the launch of the C1 modem) could gradually reduce its reliance on Broadcom. Broadcom shares tumbled about 20% in June on these worries. The new agreement transforms this potential risk into a stream of long-term, stable orders, significantly boosting Broadcom’s revenue visibility for years to come.
Why Can’t Apple Do Without Broadcom? The Moat of Custom Chips
Despite Apple’s ongoing efforts to strengthen its in-house chip strategy—from M-series Mac processors to A-series iPhone chips and the self-developed C1 cellular modem—why does it still rely heavily on Broadcom for wireless communications and custom ASICs?
Technical barriers: Decades of expertise in RF and connectivity chips are hard to replicate quickly. Broadcom has accumulated decades of technical know-how and patents in RF front-end, Wi-Fi, and Bluetooth connectivity chips. These chips require complex analog circuit design, signal processing, and power optimization—skills distinct from Apple’s expertise in digital logic processors (like the A and M series). Even though Apple has launched its own N1 chip (integrating Wi-Fi and Bluetooth) for the latest iPhone, iPad, and Mac, it remains dependent on Broadcom for analog RF front-end chips in the near term.
Economies of scale: The cost-sharing logic behind custom chips. Developing and manufacturing ASIC chips is extremely expensive, requiring large shipment volumes to amortize costs. While Apple’s annual iPhone shipments in the hundreds of millions provide scale, Broadcom also serves other hyperscale customers like Google, Meta, and Microsoft, spreading R&D costs across a broader client base. This scale effect gives Broadcom a cost and technology iteration advantage that single-source suppliers can’t match.
Co-design capabilities: Closing the loop from specification to physical implementation. Custom ASICs are not simply "built to spec"—they require deep collaboration between chip suppliers and clients from the architectural design stage onward. Broadcom and Marvell together command about 95% of the hyperscale cloud custom AI accelerator co-design market. This co-design capability is built on years of trust and understanding, forming a moat that is hard for new entrants to cross.
Supply chain resilience: Apple’s diversified procurement strategy. Apple has been pushing for a more diversified supply chain, but in key chip areas, signing long-term agreements with core suppliers like Broadcom is itself part of supply chain resilience. In early 2026, memory prices spiked nearly 98%, prompting Apple to raise MacBook and iPad prices in June to reflect cost pressures. Against this backdrop, locking in long-term pricing and capacity for critical chips is a rational strategic move for Apple.
Broadcom vs. Marvell: The Duel in the Custom Chip Arena
In the hyperscale cloud custom AI accelerator co-design market, only two companies have true global competitiveness: Broadcom and Marvell. Together, they control about 95% of the market, but their strategies and market positioning differ significantly.
Market share and scale: Broadcom’s clear lead. Broadcom holds over 70% of the custom AI chip market. In Q2 of fiscal year 2026, Broadcom’s AI semiconductor revenue reached $10.8 billion, up 143% year-over-year. The company expects full-year AI semiconductor revenue to hit $56 billion in 2026 and surpass $100 billion in 2027. Broadcom’s AI ASIC clients now include Google (TPU series), Meta, ByteDance, OpenAI, and the newly confirmed Apple.
Marvell’s catch-up: Smaller, but faster. Marvell’s fiscal 2026 revenue was $8.2 billion, up 42% year-over-year, with data center revenue at $6.1 billion, up 46%. JPMorgan projects Marvell’s data center revenue will grow from about $6.1 billion in 2025 to $9.3 billion in 2026 and reach $14.6 billion in 2027. Marvell’s clients include Amazon (Trainium series) and Microsoft (Maia series).
Competitive dynamics: Differentiation, not zero-sum. Broadcom and Marvell are not locked in a simple zero-sum game. Broadcom’s advantage lies in broader client coverage and greater scale, while Marvell seeks breakthroughs through deep partnerships with specific customers. JPMorgan expects custom AI chip shipments could surpass GPUs by 2027, meaning the overall market is still expanding rapidly, leaving ample room for both players to grow.
Potential risks for Broadcom. Broadcom is not without challenges. In early June 2026, MediaTek secured Google’s TPU orders by successfully implementing a 336G SerDes solution, showing that even in Broadcom’s strongholds, the competitive landscape is shifting. Additionally, Broadcom’s stock remains about 24% below its 52-week high of $494.35, reflecting ongoing market debates about its valuation and growth sustainability.
Broadcom and the "Second Curve" of AI Infrastructure Investment
The Broadcom-Apple agreement extension is not just good news for a chip supplier—it reflects a deeper shift in the logic of AI infrastructure investment.
First curve: The GPU arms race. Over the past two years, AI infrastructure narratives have been dominated by compute investments centered on NVIDIA GPUs. However, the general-purpose nature of GPUs means suboptimal power and cost efficiency in certain scenarios. As AI workloads shift from training to inference, demand for customized, scenario-specific compute is rising.
Second curve: The custom compute revolution of ASICs. ASICs are designed for specific purposes and deliver higher performance and efficiency than general-purpose chips for targeted workloads. Counterpoint Research predicts that custom ASIC shipments will triple between 2024 and 2027. A June 23 report from Morgan Stanley shows global demand for advanced CoWoS packaging will reach 2.694 million units in 2027, up 93% from 1.394 million in 2026, with cloud providers’ in-house ASICs becoming a new growth driver for the CoWoS market.
Broadcom’s strategic positioning: From "selling shovels" to "building shovels." What sets Broadcom apart is its deep involvement in every stage of the ASIC process—from architectural definition to physical implementation. This "co-design" model makes Broadcom a core builder of foundational compute for AI infrastructure. The Jalapeño inference chip, co-developed by Broadcom and OpenAI, went from design to tape-out in just nine months and is expected to cut inference costs by about 50%—a testament to Broadcom’s technical moat.
Implications for the crypto industry. Crypto mining has followed a similar evolution from general-purpose GPUs to custom ASICs. The progression of Bitcoin mining hardware from CPU to GPU to ASIC closely mirrors the shift in AI compute from GPU to ASIC—when workloads become standardized and large-scale, the efficiency and cost advantages of custom chips become irreversible. The move by Bitcoin miners like TeraWulf into AI compute further demonstrates the fluidity of compute infrastructure across different application scenarios.
Conclusion
The extension of Broadcom and Apple’s partnership through 2031 is more than just a renewed supply agreement—it’s a confirmation of the strategic value of custom chips in the new era. It underscores three key points:
First, in the semiconductor industry, "in-house development" and "outsourcing" are not mutually exclusive. Even a giant like Apple still needs specialized partners like Broadcom in areas such as RF, connectivity, and custom ASICs. The moat of custom chips lies not just in technology, but in the systematic capabilities built through long-term co-design.
Second, the ASIC sector is emerging from the shadow of the GPU narrative to become a standalone focus for AI infrastructure investment. From Google’s TPU to Apple’s Baltra, from OpenAI’s Jalapeño to Amazon’s Trainium, hyperscale customers are voting for custom compute with real capital.
Third, for investors, understanding the Broadcom-Marvell duopoly offers more long-term value than chasing short-term stock price swings. Broadcom’s lock-in with Apple through 2031 is not just a certainty anchor for AVGO’s custom chip business—it’s a landmark event marking the transition of the custom chip sector from "concept" to "performance."
Back to the market on July 7, 2026—Bitcoin broke above $64,000, and Broadcom closed up 3.73%. Two seemingly unrelated asset prices moved on the same day, but the shared narrative is clear: compute power is becoming the scarcest asset of the digital era. Whether for decentralized crypto networks or centralized AI cloud services, the hunger for compute is reshaping the entire technology sector’s valuation framework. And custom chips are the blueprint at the very foundation of this transformation.
FAQ
Q: How does Broadcom’s new agreement with Apple differ from the 2023 deal?
The 2023 agreement focused primarily on the development and manufacture of 5G RF components, with a scale of several billion dollars. The new agreement expands the collaboration from RF components to custom ASIC chips, covering multiple generations of Apple products, and extends the partnership through 2031. This marks a substantial upgrade from "component supply" to "co-design of compute chips."
Q: What percentage of Broadcom’s annual revenue comes from Apple?
According to multiple institutions, Apple contributes about 20% of Broadcom’s annual revenue, making it Broadcom’s largest single customer. This proportion means the stability of Apple’s orders has a direct impact on Broadcom’s valuation.
Q: What is Broadcom’s market share in custom AI chips?
Broadcom holds over 70% market share in custom AI chips. In the hyperscale cloud custom AI accelerator co-design market, Broadcom and Marvell together account for about 95% of the market.
Q: How was Broadcom’s AI business revenue in 2026?
In the second quarter of fiscal 2026, Broadcom’s AI semiconductor revenue reached $10.8 billion, up 143% year-over-year. The company expects full-year AI semiconductor revenue to reach $56 billion in 2026.
Q: What’s the main difference between custom ASIC chips and GPUs?
GPUs are general-purpose compute chips capable of handling a variety of tasks. ASICs are custom-designed for specific uses and deliver higher performance and efficiency for targeted workloads. As demand for AI inference grows, ASICs are becoming a key source of compute power for AI infrastructure.
