Forget AMD: This AI Hardware Stock Is the Smarter Bet Right Now

Broadcom's fiscal Q3 2027 AI semiconductor revenue reached $16.7 billion, up 221% year over year, versus AMD data-center revenue of $6.7 billion, up 107%. Broadcom raised its fiscal 2028 AI semiconductor revenue outlook to $115 billion from $100 billion and projects $230 billion in 2029, driven by growing hyperscaler demand for custom AI chips. The article argues Broadcom trades at roughly half AMD's forward valuation despite stronger AI growth prospects, favoring a rotation from AMD into Broadcom.

Analysis

The relevant question is not ASIC-versus-GPU revenue scale, but whether hyperscaler custom-silicon programs are incremental to accelerator budgets or displace merchant GPUs. AVGO benefits from both chip content and switching/networking attach, creating a higher-quality AI revenue mix than a simple accelerator comparison implies. If internal ASIC deployments expand, the most direct relative pressure is on AMD's already-narrower AI accelerator opportunity; NVDA is more insulated near term because its software ecosystem and broad workload flexibility remain the default for training and fast-changing inference workloads.

AVGO's forward targets should be discounted until customer concentration, program yields, and deployment cadence are independently visible. Custom chips carry meaningful execution risk: a delayed hyperscaler data-center build, lower token-demand growth, or a design-cycle miss can move revenue by billions because volumes are concentrated across a small customer set. Over the next 1-3 months, the catalyst is evidence of raised customer capex and booked production capacity; over 6-18 months, gross-margin progression and proof that ASIC volumes are not cannibalizing AVGO's networking mix matter more than headline AI sales.

Consensus may be too linear in extrapolating custom silicon adoption. ASIC economics are strongest for stable, massive workloads, but model architectures and inference requirements remain fluid; that favors programmable GPUs and can cap the addressable market for fixed-function designs. The better relative thesis is therefore AVGO upside versus AMD downside, not a blanket short of NVDA or a presumption that all AI spend shifts away from GPUs.

Key Decisions for Investors

• Initiate a 6-12 month long AVGO / short AMD dollar-neutral pair after confirming next-quarter hyperscaler capex guidance; target 15-20% relative outperformance, with a 7-8% relative stop if AMD AI revenue or MI-series customer traction materially exceeds expectations.
• For lower beta, own AVGO versus SOXX rather than outright: use a 3-6 month position sized for a 10% AVGO-specific drawdown. Falsify if AVGO guides AI semiconductor revenue below prior run-rate expectations or reports AI gross-margin dilution without offsetting network growth.
• Do not chase AVGO solely on long-dated management projections. Add only on earnings-related volatility or evidence that a second/third custom-chip customer reaches volume production; missing data to monitor are customer concentration, wafer commitments, and program-level margin contribution.
• Maintain NVDA neutral rather than short as the ASIC hedge. Revisit only if hyperscaler disclosures show sustained internal-silicon deployment reducing external GPU purchases, rather than merely supplementing capacity for mature inference workloads.