Broadcom Forecasts >$100B AI Chip Revenue; Large Orders From Anthropic, OpenAI

Broadcom pushed more aggressively into commercial AI silicon with a multi‑year demand outlook that assumes AI‑accelerator sales will surpass $100 billion by 2027. Management tied that forecast to scheduled, large‑scale deliveries measured in gigawatts: a multi‑gigawatt commitment to Anthropic (commonly reported around ≈3 GW) and a separate, over‑one‑gigawatt shipment planned for OpenAI, both concentrated in the 2027 build window. The company concurrently guided second‑quarter revenue near $22 billion — above consensus — and authorized up to $10 billion of share repurchases to shore up capital structure and signal confidence in cash generation.

Market participants responded with an immediate stock move higher pre‑market, reflecting a mix of optimistic demand read‑throughs and a renewed debate over whether headline design wins can be converted quickly into recurring volume. Supporting Broadcom’s addressable market thesis are corroborating upstream reads: ASML’s recent results showed sizable new bookings (reported around €13 billion) and strong sales that signal foundries and IDMs are committing to multi‑year equipment buys, while TSMC reported demand verification from hyperscalers and flagged a stepped‑up capex profile (management cited a roughly 30% revenue uplift scenario for 2026 in contemporaneous reporting). Those supplier signals make the capacity expansion story more plausible but are not a guarantee of immediate wafer‑level output.

Analysts offer divergent takes. Jefferies published a notably bullish scenario that links heavier hyperscaler capex to a potential Broadcom share capture in an initial Google‑scale build (the bank modeled an initial window of about six million server‑accelerator units and suggested Broadcom could take roughly 85–90% of that tranche), producing a price objective implying roughly 60–62% upside from recent levels. Other sell‑side and industry voices caution that NVIDIA’s entrenched software stack, commercial relationships and broad GPU applicability will protect much of its market share except in tightly defined, high‑volume ASIC use cases.

Critical supply‑side constraints remain: substrate availability, packaging and test throughput, advanced‑node wafer allocation and lengthy qualification cycles are repeatedly cited by the supply chain as potential bottlenecks that can slow shipment conversion. Geopolitical and regulatory dynamics — selective clearances for high‑end NVIDIA parts into certain markets and evolving U.S.–Taiwan arrangements that affect tariffs and onshore investment — further complicate near‑term access to advanced nodes and customer footprints in specific regions.

Taken together, the data point to a hybrid outcome: Broadcom has secured anchor customers and upstream commitments that materially increase the probability of scaling, but realizing the full revenue pathway requires overcoming manufacturing, integration and policy frictions. For hyperscalers the practical procurement response is likely diversification — deploying ASICs where efficiency and unit economics dominate, and retaining GPUs where software breadth, tooling and vendor neutrality remain decisive. For investors, the announcement tightens the link between hyperscaler capex intent and semiconductor valuations, but execution timing will determine whether the upside is front‑loaded or stretched across several years.