Broadcom CEO Hock Tan Says Demand Already Exceeds Broadcom's $115 Billion AI Forecast. Here's What That Means for the Stock.

Broadcom CEO Hock Tan said AI revenue is expected to double to about $115 billion in fiscal 2027 and double again to roughly $230 billion in fiscal 2028, with demand already exceeding the 2027 outlook. The company also targets more than $30 in fiscal 2028 EPS, over six times fiscal 2025 earnings, following Q3 sales growth of 86% to $29.6 billion and a 96% increase in non-GAAP EPS to $3.32. Demand for Broadcom's custom AI ASICs from customers including Alphabet, Anthropic and OpenAI underpins the outlook; analysts' average $533 price target implies 46% upside.

Analysis

The investable issue is not headline AI growth but whether custom silicon shifts the profit pool away from merchant GPUs. AVGO is becoming the key enabler for hyperscalers seeking lower inference/training cost per token, which can reduce their dependence on NVDA’s full-stack pricing over a 12-24 month horizon. The second-order beneficiary is GOOG: internally differentiated TPU capacity can improve Gemini unit economics and lower external accelerator procurement, although the benefit is more likely to appear in cloud gross-margin durability than near-term revenue.

The guidance has unusually high expectations embedded: investors should underwrite supply availability, advanced-packaging capacity, yield, and customer deployment schedules—not management’s demand signal. A large portion of ASIC economics is customer-concentrated and program-specific; a delayed model launch, change in architecture, or customer decision to multi-source can shift revenue recognition materially despite unchanged end-demand. AVGO’s software cash flows provide downside support, but the AI multiple will compress quickly if the conversion of booked designs into volume shipments slips.

Near term, positive revisions can continue as investors capitalize a multi-year earnings path. Over the next 1-3 months, the relevant catalyst is evidence that supply expansion is translating into incremental shipment capacity rather than merely longer lead times; over 6-18 months, monitor whether custom silicon expands from a few flagship programs into a repeatable multi-customer platform. Consensus may be underestimating custom ASIC disruption to NVDA at the margin, but is likely overconfident that all announced/custom programs produce GPU-like margins and uninterrupted volume ramps.

Key Decisions for Investors

• Maintain an overweight AVGO versus SOXX for the next 6-12 months, but add only on post-results volatility rather than chase guidance-driven strength. Target a 2:1 upside/downside framework; exit or reduce if AI revenue guidance is not raised alongside evidence of supply expansion, or if semiconductor gross margin falls materially despite mix improvement.
• Initiate a measured AVGO long / NVDA short pair (roughly beta-neutral) over a 12-18 month horizon to express custom-silicon share gains while reducing broad AI-capex risk. Thesis fails if NVDA sustains accelerating data-center growth without material gross-margin pressure, or if AVGO discloses customer-program delays or weaker-than-expected ASIC margins.
• Keep GOOG as a watch-list beneficiary rather than a direct ASIC trade. Add only if Cloud margin expansion and TPU-driven capacity utilization become visible in quarterly disclosures; the missing data are internal TPU economics, external cloud demand mix, and incremental capex requirements.
• Do not use NFLX as an AI-chip read-through; it has no identifiable earnings sensitivity to this mechanism from the available information.