Broadcom remains one of the semiconductor industry’s leading artificial-intelligence beneficiaries, even as MediaTek begins challenging its position in custom chips designed for Google.
Morgan Stanley reiterated its positive view of Broadcom, describing the company as a core AI winner despite concerns that MediaTek could capture part of its Google Tensor Processing Unit business. The bank expects the competitive threat to be real but manageable, with Broadcom potentially retaining roughly 80% of the relevant share.
The debate matters because custom AI accelerators have become a central part of Broadcom’s growth story. Large cloud providers increasingly want chips optimized for their own workloads rather than relying exclusively on general-purpose graphics processors. Broadcom supplies the design expertise, connectivity technology and manufacturing relationships needed to bring these application-specific integrated circuits, or ASICs, into production.
MediaTek’s expansion creates a new competitive variable. However, Broadcom’s existing customer relationships, supply capabilities and AI networking portfolio may make its position more difficult to disrupt than recent investor concerns suggest.
Broadcom’s AI Revenue Is Accelerating Rapidly
Broadcom’s latest financial results show why Morgan Stanley continues to classify the company as a major AI beneficiary.
The company generated $10.8 billion in AI semiconductor revenue during its fiscal second quarter of 2026, representing growth of 143% from the previous year. Broadcom said the increase came from rising demand for custom AI accelerators and AI networking products.
Total quarterly revenue reached $22.19 billion, up 48% year over year. Semiconductor solutions revenue increased 79% to approximately $15.01 billion, while infrastructure software revenue rose 9% to $7.18 billion.
Management expects the acceleration to continue. Broadcom forecast fiscal third-quarter revenue of approximately $29.4 billion and said AI semiconductor revenue should reach about $16 billion, more than tripling from the comparable period.
Those figures indicate that AI is no longer a small or experimental part of Broadcom’s business. It has become the company’s most important semiconductor growth driver and is contributing an increasingly large portion of consolidated revenue.
The central investment question is whether Broadcom can maintain that pace as competition expands and cloud customers become more involved in designing their own processors.
Why Custom AI Chips Matter
Custom AI chips are processors developed for the specific workloads of an individual cloud provider or technology company.
Unlike general-purpose GPUs, ASICs can be optimized for a narrower set of functions, potentially improving performance, energy efficiency and total operating cost. These advantages become increasingly important when hyperscale data centers deploy hundreds of thousands of processors.
Broadcom works with large customers to convert their chip architectures into manufacturable products. Its role may include intellectual property, high-speed interconnects, packaging expertise and coordination with semiconductor foundries.
This model allows cloud companies to retain control over important parts of the processor design while relying on Broadcom’s technical and supply-chain capabilities.
Google’s TPU program is one of the most prominent examples of custom AI silicon. Tensor Processing Units are designed to accelerate machine-learning workloads across Google’s internal infrastructure and cloud services.
MediaTek’s reported involvement in future Google chips suggests that Google wants a broader supplier base. Adding another design partner could improve negotiating leverage, increase capacity and reduce dependence on a single provider.
However, supplier diversification does not necessarily mean Broadcom will lose most of the program. Large AI deployments can support multiple partners, particularly when spending and computing requirements are expanding rapidly.
MediaTek Is Becoming a Credible Competitor
MediaTek is best known for processors used in smartphones and consumer devices, but the Taiwanese company is expanding aggressively into data-center custom silicon.
The company has doubled its 2026 data-center revenue forecast to $2 billion. It estimates that the custom AI ASIC market could reach $70 billion to $80 billion in 2027 and is targeting a 10% to 15% share of that opportunity.
MediaTek also says it can support both Taiwan Semiconductor Manufacturing Company’s CoWoS packaging and Intel’s EMIB technology. Advanced packaging connects processors, memory and other components within powerful AI computing systems. Supporting more than one packaging platform could offer customers additional flexibility.
Reuters reported that Intel’s EMIB technology was being considered for custom AI chips MediaTek was designing for Google, although MediaTek had not publicly confirmed Google as a customer.
These capabilities make MediaTek a legitimate competitor rather than a purely theoretical threat. It has experience designing complex chips, access to advanced manufacturing and ambitions to expand beyond its traditional mobile business.
Still, entering the custom AI market is different from displacing an established supplier. Broadcom has years of experience with hyperscale programs, a large portfolio of connectivity technology and established production processes.
Morgan Stanley’s view appears to be that MediaTek can win meaningful business without undermining Broadcom’s broader AI investment case.
Broadcom’s Networking Business Strengthens Its Position
Broadcom’s advantage extends beyond custom accelerators.
AI data centers require high-speed switches, optical components and networking chips to connect thousands of processors. The performance of the overall cluster depends not only on individual accelerators but also on how efficiently data moves between them.
Broadcom is a major supplier of Ethernet networking technology used in these systems. That gives the company exposure to AI infrastructure even when customers use accelerators designed by other suppliers.
This diversified position is strategically important. A loss of some custom-chip share at one customer may be partially offset by higher demand for networking products across a wider range of data centers.
Broadcom’s Q2 results showed that both custom accelerators and AI networking contributed to its $10.8 billion of AI semiconductor revenue.
The company is also working with additional large technology customers. Meta expanded its custom-chip partnership with Broadcom through 2029, including an initial deployment exceeding one gigawatt of computing capacity. Broadcom’s networking hardware is also expected to support Meta’s expanding AI infrastructure.
Broadcom and OpenAI separately announced a collaboration involving the deployment of 10 gigawatts of custom AI accelerators, further broadening the company’s exposure beyond Google.
These programs reduce the importance of any single customer over time, although customer concentration remains a relevant risk.
Profitability and Cash Flow Support the Bullish Case
Broadcom’s AI expansion is producing substantial earnings and cash flow rather than revenue growth alone.
The company reported fiscal Q2 adjusted EBITDA of $15.24 billion, equal to 69% of revenue. Free cash flow reached $10.26 billion, representing 46% of sales.
Strong cash generation gives Broadcom flexibility to fund research, return capital to shareholders and manage debt associated with its infrastructure software acquisitions.
The company also benefits from combining semiconductor growth with recurring software revenue. Its VMware-related infrastructure software business can provide a more stable earnings base when non-AI semiconductor markets weaken.
Investors should nevertheless monitor whether rapid AI revenue growth maintains Broadcom’s current margins. Custom programs can involve significant engineering investment, advanced packaging costs and commitments to secure manufacturing capacity.
The fiscal third-quarter outlook calls for adjusted EBITDA of approximately 68% of projected revenue, suggesting management expects profitability to remain strong despite the sharp increase in sales.
What Could Challenge AVGO Stock?
The most immediate risk is faster-than-expected custom-chip share loss.
Should MediaTek win a significantly larger portion of Google’s TPU program than Morgan Stanley anticipates, investors may reduce their estimates for Broadcom’s future accelerator revenue.
Customer concentration is another concern. Custom AI projects involve a relatively small number of hyperscale companies, meaning changes in the spending plans of one major customer can materially affect growth.
Broadcom also faces competition from Marvell, internal chip-design teams and emerging suppliers attempting to make custom silicon development more flexible.
Valuation presents an additional risk. Expectations for AI revenue have risen rapidly, so even strong results could disappoint if guidance does not exceed optimistic forecasts.
Supply constraints involving advanced manufacturing, memory, packaging and networking components could also delay deployments. Demand may remain strong while revenue recognition shifts into later quarters.
What Morgan Stanley’s Call Means for Investors
Morgan Stanley’s argument is not that Broadcom faces no competition. Instead, the bank appears to believe the market has overestimated how disruptive MediaTek’s entry will be.
Broadcom’s position is supported by accelerating AI revenue, long-standing custom-chip expertise, advanced networking products and relationships with several major AI infrastructure customers.
MediaTek may gain meaningful share in Google’s future programs, but the overall custom ASIC market is expanding rapidly. In a growing market, Broadcom may continue increasing revenue even if its percentage share declines modestly.
Investors should focus on fiscal third-quarter AI revenue, new custom accelerator customers, networking growth and management’s long-term outlook. These indicators will show whether Broadcom’s AI opportunity remains broad enough to offset competitive pressure.
FAQ
Why does Morgan Stanley consider Broadcom a core AI winner?
Broadcom benefits from custom AI accelerators, Ethernet networking products and large hyperscale customer relationships. Its fiscal Q2 AI semiconductor revenue grew 143% year over year to $10.8 billion.
Is MediaTek taking business from Broadcom?
MediaTek is reportedly participating in custom AI chip development associated with Google. Morgan Stanley views the threat as genuine but expects Broadcom to retain most of the relevant share.
What is the company’s AI revenue forecast?
Broadcom expects approximately $16 billion in AI semiconductor revenue for fiscal Q3 2026, representing growth of more than 200% from the prior-year period.
How does Broadcom benefit from AI networking?
It supplies Ethernet switches and connectivity chips that move data between accelerators in large AI clusters. This gives it exposure to AI spending beyond its custom processor business.
What are the main risks for AVGO stock?
Major risks include custom-chip competition, customer concentration, supply constraints, elevated valuation and AI revenue guidance that falls short of market expectations.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. Always conduct your own research before making any investment decisions.
