Tech Retakes Lead Heading into Q4: Stocks to Watch

The FOMC raised rates 25bps to 3.75%-4.00%, while its median end-2026 rate projection rose to about 4.1% and its core PCE forecast increased to 3.4%; the 10-year Treasury yield briefly exceeded 5%. Despite the hawkish backdrop, the Nasdaq gained roughly 2% over the final two sessions of the week as capital rotated into semiconductors on strong earnings revisions. AMD closed at $559.82, up 161% year-to-date, supported by 50% Q2 revenue growth and doubled Data Center revenue, while SanDisk rose 10.99% Friday and more than 650% year-to-date amid NAND shortages and its pending S&P 100 inclusion.

Analysis

The relevant signal is not broad technology duration resilience; it is a narrow earnings-revision and supply-constrained hardware trade. That distinction matters because AI capex beneficiaries with visible backlog can sustain multiples while long yields rise, whereas software and ad-funded platforms without a comparable estimate-up cycle remain exposed to discount-rate compression. NVDA remains the benchmark risk: any evidence that accelerator supply is normalizing or hyperscaler capital budgets are being deferred would rapidly transmit to AMD, memory, networking and optical names.

AMD's upside over the next 1-3 months depends on converting customer commitments into recognized data-center revenue and, more importantly, proving gross-margin expansion despite competitive pricing against NVDA. A beat on revenue alone is insufficient after the move; the catalyst must be raised forward GPU revenue, improved mix, or a credible supply ramp. The cleaner second-order beneficiaries are HBM and memory suppliers such as MU and SK Hynix exposure proxies, plus AI-server component vendors, although NAND pricing strength is cyclical and historically self-correcting as capacity returns.

SNDK is a poor risk-adjusted fundamental chase following a large momentum move: index-related demand is finite, while passive-flow estimates are often overstated relative to actual float, active front-running, and rebalance implementation. The contrarian setup is that post-rebalance selling can begin within days unless spot NAND contract pricing continues to rise. Verify the reported price, index eligibility, free float, and the magnitude/duration of NAND contract-price increases before assigning any earnings value to the move.

The structural risk over 6-18 months is that high rates raise the hurdle for AI projects that lack immediate monetization; hyperscalers can fund capex, but their return thresholds eventually constrain incremental orders. The thesis is falsified near term by downward 2026 semiconductor EPS revisions, a material cut to MSFT/GOOG/META/ORCL capex guidance, or a break in SOX relative strength versus the Nasdaq after earnings.

Key Decisions for Investors

• Initiate a 1-3 month pair: long AMD / short SMH or NVDA-sized beta hedge, entered only on a pullback rather than a breakout. The objective is AMD-specific share and product-ramp upside while limiting sector-duration risk; exit if AMD's next results do not raise forward data-center revenue expectations or if the pair underperforms by 10%.
• Avoid adding SNDK into the index rebalance. Establish a post-event short watch only if price remains elevated while NAND contract-price data flatten; use defined-risk puts or a small short with a 12-15% stop because a genuine supply squeeze can extend sharply.
• Favor MU over SNDK for 6-12 month memory exposure if HBM demand and DRAM pricing remain firm: MU offers broader AI-memory participation and less reliance on a single NAND scarcity narrative. Reduce if quarterly gross-margin guidance fails to improve or memory spot prices decline for two consecutive monthly readings.
• Use MSFT, GOOG, META, and ORCL earnings as the capex-confirmation basket over the next reporting cycle. Any aggregate reduction in 2026 AI infrastructure spending should trigger de-risking across AMD, NVDA, MU and server supply-chain longs rather than being treated as an isolated hyperscaler issue.