Broadcom slides 13% as Treasury yields pressure AI chip stocks

Broadcom's custom ASIC business keeps compounding even as a 19-year high in long-term Treasury yields knocked 12.87 percent off the stock in a single week, with Q2 AI semiconductor revenue up 143 percent year-over-year to $10.8 billion. The selloff hit the entire AI chip complex: the Philadelphia Semiconductor Index fell 4.96 percent, Micron dropped 7.02 percent, and Samsung Electronics and SK hynix slid 7.82 percent and 9.75 percent respectively on the Seoul exchange.

"Visibility now extends to 2028," CEO Hock Tan said on the earnings call, pointing to signed contracts including a 3-gigawatt Meta MTIA commitment through 2028 and a 10-gigawatt OpenAI deployment by 2029. Q2 AI bookings exceeded $30 billion against $10.8 billion shipped, and management guided Q3 AI revenue to $16 billion — over 200 percent year-over-year — while reiterating a fiscal 2027 target above $100 billion.

The financials back the growth story. Q2 operating income hit $10.788 billion, up 85.07 percent year-over-year. Free cash flow reached $10.262 billion, or 46 percent of revenue, with adjusted EBITDA margin at 69 percent. Cash on the balance sheet more than doubled year-over-year to $19.628 billion, and fiscal 2025 threw off $26.914 billion in free cash flow. Broadcom has raised its dividend for 15 consecutive years since fiscal 2011, with the current quarterly payout at $0.65.

Why Custom ASICs Win When Rates Rise

The rate environment is reshaping the competitive math in AI silicon. When borrowing costs climb, hyperscalers get pickier about capital expenditures — and pickier hyperscalers choose cheaper custom ASICs over Nvidia's general-purpose GPUs. Broadcom designs the custom chips that Alphabet, Meta, and OpenAI are locking into multi-year contracts to run their AI clusters. Nvidia remains the reflex pick for AI exposure, but Broadcom trades at roughly 20x forward earnings against Nvidia's premium multiple, while pairing that valuation with a 15-year dividend increase streak and free cash flow compounding at 60.07 percent year-over-year.

The rate-driven rotation has been brutal for the broader AI infrastructure cohort. CoreWeave fell 12.10 percent on Aug. 18 as investors reassessed its $35 billion to $39 billion capex target under rising financing costs. Nebius Group dropped as much as 14 percent intraday before paring losses. TeraWulf slid 6.5 percent. A Wall Street Journal analysis found nine top technology companies hold approximately $3 trillion in off-balance-sheet AI commitments, raising questions about the durability of hyperscaler spending that underpins the entire AI chip order book.

The Risk That Could Break the Thesis

Customer concentration is the main vulnerability. A handful of hyperscalers drive the AI ramp, and if Google, Meta, or OpenAI slow orders or shift to fully internal designs, the guidance breaks. Broadcom also carries $91.467 billion in total liabilities from the VMware deal, and higher rates make servicing that debt costlier. What blunts the risk: booking visibility through 2028, gigawatt-scale contractual commitments already logged, and shareholders' equity climbing to $87.691 billion while total liabilities declined 3.76 percent year-over-year.

The market is pricing rate fear into growth multiples across the semiconductor sector. AVGO sits at $364.03, well below its 52-week high of $494.18, with the analyst consensus target at $527.88. Full-year FY2026 AI revenue is guided to $56 billion, with the path to over $100 billion in FY2027 already backed by customer commitments. When the market marks down a business generating $26.914 billion in annual free cash flow with a 15-year dividend growth streak because the 10-year Treasury moved 5 basis points, the valuation gap between price and fundamentals becomes the trade.

This article is for informational purposes only and does not constitute investment advice.