Nvidia, Micron, and Sandisk Are Up Over 1,000% Since the AI Boom Started. These 2 AI Stocks Could Be the Next Big Winners.

AI infrastructure demand is set to lift connectivity/networking winners like Astera Labs and Arista Networks as data transmission needs surge with larger, more complex AI workloads. Astera reported Q2 revenue up 104% to $392M and non-GAAP net income up 82% to $0.80/share, with Q3 revenue growth guided to accelerate to 139% as Scorpio switches scale; Arista posted Q2 revenue up 38% to $3.0B and non-GAAP net income up 40% to $1.02/share, guiding Q3 growth to 43% for both revenue and net income. The article frames both names as potential “next big winners,” despite valuations that appear elevated (Astera ~144x earnings, Arista ~61x), supported by consistent earnings beats versus consensus.

Analysis

This is less a broad AI beta trade than a capex mix shift. As cluster sizes rise, networking content per dollar of AI spend increases faster than compute units, so ANET is the cleaner beneficiary than NVDA at the margin, while CSCO risks losing share in the highest-growth part of the data-center stack. ALAB has more torque because it sits closer to the bottleneck, but that also makes it more sensitive to any pause in hyperscaler rollout schedules.

Near term, the catalyst is not the narrative but whether order flow converts into billings and whether management can defend margins as customers multi-source. Over 1-3 months, the stocks can keep rerating if AI capex stays front-loaded; over 6-18 months, the main falsifier is a shift toward custom interconnect architectures or a digestion phase in server deployments, which would cap ANET's multiple and hit ALAB hardest because it trades as a high-duration growth asset.

The consensus may be missing that networking is a second-wave trade, not a clean secular line item: it is lumpy, customer-concentrated, and often delayed until compute and memory bottlenecks are partially solved. That makes the upside real but timing-sensitive; if memory or storage remains the binding constraint, the market may be early on the revenue inflection for network vendors. In that scenario, quality names with pricing power should hold up, while the weakest-moat growth names can de-rate quickly on any guide miss.

Key Decisions for Investors

• Long ANET vs. short CSCO for 1-3 months: preferred way to express AI networking share gain. Use a relative-value trade because ANET should capture incremental AI switch spend while CSCO is exposed to slower legacy enterprise mix. Falsifier: CSCO AI/data-center commentary improves materially or ANET order growth decelerates.
• Start a small ALAB position only on pullbacks or via a 6-12 month call-spread structure rather than outright stock. The upside is operating leverage if AI racks keep densifying, but the multiple can compress fast if the ramp slips. Keep size modest because this is a high-beta execution story, not a low-risk compounder.
• Avoid chasing NVDA here; the marginal increment from networking spend is likely to accrue to adjacencies before it meaningfully changes GPU share economics. If owning NVDA, treat it as the core platform bet, not the best expression of the next leg of AI infrastructure capex.
• Watch MU and SNDK as secondary beneficiaries, but require confirmation from pricing and gross-margin trends before adding. If memory tightness eases faster than expected, it can actually delay network procurement, which would be negative for ALAB/ANET despite being positive for the broader AI stack.