BofA's Vivek Arya Sees Nvidia at $350 as Agentic AI Drives an “Unprecedented” Chip Cycle

Bank of America’s Vivek Arya raised NVIDIA’s price target to $350 versus a stock around $215, citing unprecedented generative AI demand and agentic AI adoption as drivers of sustained semiconductor demand. NVIDIA’s Q1 FY27 revenue rose 85.23% year over year to $81.61B, gross margin held at 75.0%, and the board authorized an additional $80B buyback while lifting the quarterly dividend from $0.01 to $0.25. The article argues the supply squeeze is benefiting the broader AI semiconductor complex, including AMD, Broadcom, Lam Research, Marvell, and Intel.

Analysis

The key second-order read is that AI demand is no longer being driven only by model training, but by sustained inference workloads that scale with user activity and task complexity. That matters because agentic systems are structurally more GPU-intensive per unit of end-user value than chatbots, which means capex can keep accelerating even if headline model launches slow. The beneficiaries are not just the obvious accelerator vendors; the entire bottleneck stack—HBM, advanced packaging, substrates, optics, wafers, and foundry capacity—should retain pricing power longer than the market currently discounts.

The most interesting implication is that hyperscaler capex restraint may now be self-defeating. If compute intensity is tightly linked to growth, then companies that underinvest risk slower product iteration, weaker retention, and lower monetization per customer, while aggressive spenders compound share gains. That creates a winner-take-more dynamic in which the strongest balance sheets can defend or expand moats, but weaker cloud/enterprise AI players may be forced into catch-up spending at inferior returns.

Near term, the stock reaction is likely to remain concentrated in the “picks-and-shovels” layer rather than in application software, because the market is still pricing a longer duration of supply scarcity. The main risk is not demand destruction but normalization: if Blackwell ramp, China exposure, or customer digestion slows for even one quarter, the market could compress multiple expansion quickly despite still-strong fundamentals. The contrarian point is that the setup may be less about a perpetual scarcity trade and more about a 12-18 month earnings revision cycle; if so, the best risk/reward is owning the highest operating leverage names and fading any assumptions that the whole AI complex deserves equal upside.

Key Decisions for Investors

• Add NVDA on any post-earnings consolidation; use a 3-6 month horizon. Risk/reward remains favorable as long as revenue growth stays >60% and gross margin holds near 75%; downside is a guide cut or China contribution miss that could trigger a 10-15% multiple reset.
• Express relative value with long NVDA / short a basket of lagging AI infrastructure beneficiaries that depend on the same capex cycle but have lower pricing power; target 2-3x beta-adjusted upside if the earnings revision cycle broadens, but cut if supply-chain lead times normalize faster than expected.
• Stay long AVGO and LRCX as secondary beneficiaries of the bottleneck trade; these names offer cleaner exposure to AI infrastructure demand with less headline-risk than NVDA, and should outperform over the next 1-2 quarters if AI semis and capex remain above consensus.
• Use AMD as a tactical catch-up long only on weakness, not strength. The trade works if investors keep rotating into alternative AI compute suppliers, but the risk/reward is less attractive than NVDA because the market will likely demand proof of share gains before re-rating.
• Avoid treating INTC as a pure AI beneficiary until execution is visible; the foundry narrative is a 2-4 year story, so any long should be hedged with a short in a more direct beneficiary of current demand to isolate roadmap optionality from near-term disappointment risk.