What a $10,000 Investment Split Between Nvidia and Broadcom Could Be Worth by 2030

The article argues Nvidia and Broadcom could generate 3x-4x stock returns by 2030, turning a $10,000 split investment into an estimated $30,000-$40,000. Broadcom is projected to reach roughly $115 billion in AI semiconductor revenue in 2027 and $230 billion in 2028, while Nvidia has forecast 70% revenue growth in 2027 and estimates global data-center capex could reach $3 trillion-$4 trillion by 2030. The bullish thesis rests on AI infrastructure spending potentially tripling or quadrupling from roughly $1 trillion currently, with a growing share ultimately allocated to compute hardware.

Analysis

The relevant debate is not whether AI infrastructure spend grows, but whether the incremental dollar shifts from merchant GPU systems toward customer-specific accelerators. AVGO benefits disproportionately if a small number of hyperscalers commit to multi-generation ASIC roadmaps: custom silicon raises switching costs, pulls through networking, and can produce materially better revenue visibility than spot GPU demand. NVDA retains the higher-value position where workloads change rapidly or software portability matters, but its multiple is more exposed to evidence that inference economics are improving faster on ASICs than on GPUs.

The second-order beneficiary is TSM, which captures wafer content regardless of the NVDA-versus-AVGO mix, while CoWoS/advanced-packaging availability remains the critical bottleneck. The likely losers from greater ASIC adoption are not NVDA outright, but smaller GPU challengers such as AMD and potentially merchant interconnect suppliers where AVGO bundles compute and networking into hyperscaler designs. Conversely, a shortage of high-bandwidth memory or advanced packaging could cap AVGO's custom-chip ramp more severely than headline design-win announcements imply.

Near term, this is largely a positioning and expectations issue rather than a new fundamental catalyst; promotional long-range revenue extrapolations do not establish earnings power, terminal margins, or capital intensity. Over the next 1-3 months, hyperscaler capex guidance, NVDA's gross-margin trajectory, and disclosed AVGO AI backlog/conversion are the key checks. Over 6-18 months, the decisive metric is inference share: sustained custom-silicon deployments that reduce GPU unit demand would narrow NVDA's growth duration and compress its premium, even if aggregate AI spending remains strong.

Contrarian view: the market may be underestimating coexistence rather than substitution. ASIC programs often address stable internal workloads and can expand total inference capacity rather than displace frontier-training GPU clusters; thus an AVGO win does not automatically make NVDA a short. The cleaner relative thesis is that AVGO has more upside only if custom AI revenue converts into durable, diversified programs rather than remaining concentrated among a few buyers with meaningful cancellation and pricing power.

Key Decisions for Investors

• Maintain a core long NVDA and add AVGO only as a relative-overweight hedge, rather than treating them as mutually exclusive. A 6-12 month long AVGO / short AMD pair expresses custom-silicon share gains while avoiding a directional AI-capex bet; exit if AVGO's next two earnings reports show AI revenue growth decelerating without offsetting networking strength.
• Use TSM as the lower-idiosyncratic AI-silicon exposure for a 6-18 month horizon. The thesis fails if advanced-packaging utilization weakens or hyperscaler capex guidance is cut materially; monitor monthly foundry commentary and packaging lead times rather than headline accelerator demand.
• Do not chase a fresh outright NVDA long solely on long-dated market-size claims. Add only after earnings if data-center growth and gross-margin guidance demonstrate that product transitions are not creating a larger-than-expected margin reset; a meaningful guide-down in gross margin without an offsetting revenue acceleration is the risk trigger.
• Set an event-driven alert around AMZN, GOOGL, META and MSFT capex disclosures over the next reporting cycle. Raise AVGO exposure only if at least two customers indicate sustained multi-year custom-silicon deployment or if backlog conversion becomes independently visible; absent that evidence, treat the narrative as insufficient for a new concentrated position.