The story centers on Broadcom pivoting to AI-centric semiconductors and networking as the core of its business, driven by multi-year commitments from leading AI customers.

Broadcom reports strong profitability with free cash flow of $10.26 billion (46% of revenue) and adjusted EBITDA at 69% of revenue in Q2 FY26, as net income rises about 88% to $9.31 billion.

While customer concentration remains a risk, Q2 disclosures show orders from two additional core customers shipping in late 2026, broadening the base.

CEO Hock Tan highlights insatiable demand for XPUs and networking, supported by more than $30 billion in AI bookings in Q2 FY26.

commitments from Anthropic, OpenAI, and Meta illustrate long-term scale: Anthropic with over 1 GW in 2026 and 5 GW from 2027; OpenAI 1.3 GW in 2027 within a 10 GW by 2029 framework; Meta MTIA XPUs for 3 GW through 2028.

Broadcom positions itself as an alternative to Nvidia by giving customers direct access to TPU-roadmaps (Google), Meta MTIA, OpenAI, and Anthropic, strengthening its strategic moat.

Networking accounted for nearly 40% of AI revenue in Q2, with Tomahawk 6 as a flagship 100-terabit Ethernet switch and a 200-terabit successor in development.

The company benefits from accelerating demand for custom AI accelerators (XPUs) from hyperscale and LLM customers, with a strengthened backlog and a new major customer.

AI semiconductor revenue surged from $5.2B in Q3 FY25 to $10.8B in Q2 FY26, guiding for $16B in Q3 FY26 and a FY26 target of $56B, with expectations of surpassing $100B by FY27.

These multi-year contracts enhance revenue visibility tied to large-scale data center builds and AI compute architectures.

The thesis centers on Broadcom capturing hyperscaler AI capex that may bypass Nvidia by owning custom silicon and AI networking interconnect markets, leveraging XPUs and ASICs.

The AI business is highly profitable (69% EBITDA) and projected to grow from $56B this year toward $100B next year, with customers planning around Broadcom’s ship dates.