Jim Cramer says the world of tech investing has changed and it's not going back

Nvidia reported adjusted EPS of $1.87 and revenue of $81.62B, reinforcing Jim Cramer's view that semiconductors are now the market's center of gravity in the AI boom. He argued AI is weakening software's pricing power as chips, hardware, and AI models from firms like Nvidia, AMD, Arm, Intel, Broadcom, Anthropic, and OpenAI reshape enterprise spending. Year to date, the iShares Semiconductor ETF is up about 72% while the iShares Expanded Tech-Software Sector ETF is down about 12%.

Analysis

The key shift is not just relative performance, but where pricing power migrates inside the AI stack. If compute becomes the bottleneck, the value accrues first to the picks-and-shovels layer: accelerators, interconnect, memory bandwidth, and custom silicon. That argues for continued multiple support in NVDA/AVGO/ARM even if end-demand decelerates, because capex tends to remain sticky for 2-3 quarters after customers start questioning software budgets.

The second-order loser set is legacy enterprise software, but the damage is more nuanced than simple substitution. AI-native workflows compress seat counts, reduce module expansion, and make buyers more willing to “build” rather than “buy,” which can hit net revenue retention before headline bookings roll over. That creates a lagged deterioration: the first warning sign is weaker upsell and slower billings growth, then margin pressure from higher AI compute costs as vendors try to defend share.

Near term, the crowded trade risk is in semis themselves: the market is already extrapolating a multi-year AI supercycle, so any pause in hyperscaler capex or digestion after earnings can trigger sharp factor rotation. The contrarian read is that software is not dead, it is becoming the distribution layer for AI functionality; the most vulnerable names are those without proprietary data or workflow lock-in, while platform leaders with embedded customer relationships can still pass through AI features and preserve pricing over time. In other words, this is less a wholesale asset-class call than a dispersion trade between compute-heavy enablers and commoditized SaaS.

Key Decisions for Investors

• Stay long NVDA/AVGO on 1-3 month horizon, but size for volatility; use pullbacks of 5-8% to add because the next leg higher likely comes from accelerating supply-chain confidence, not another one-day earnings reaction.
• Initiate a pair trade: long SOXX or SMH vs short IGV/XSW over the next 4-8 weeks. If AI capex remains firm, semis should retain relative leadership while software multiple compression continues; target 8-12% spread capture.
• Underweight CRM and ADBE versus the software peer group for the next 2 quarters. Best risk/reward is through bearish put spreads 3-6 months out, as pricing-pressure data should emerge first in billings, not EPS.
• For higher-conviction expression, buy ARM call spreads into any weakness tied to semiconductor digestion. ARM has leveraged exposure to custom silicon adoption, but risk/reward is better defined via spreads because valuation already discounts strong execution.
• Avoid chasing INTC rallies unless there is evidence of sustained foundry/customer wins; use it only as a tactical mean-reversion trade, not a structural long, because the market is likely to reward pure AI exposure over turnaround optionality.