Apple's last tower topples… and the others will follow

Apple has discontinued the Mac Pro (the M2 Ultra tower launched June 2023, ~ $7,000), a signal that expandable, upgradeable tower desktops are being phased out in favor of highly integrated SoC designs. The Mac Pro lacked RAM expansion and support for add-on GPUs, illustrating the broader industry trend toward integrated CPU/GPU/memory/storage that reduces addressable aftermarket sales for DIMMs, discrete GPUs and primary storage. Expect a gradual sectoral repositioning: downside pressure on discrete GPU/component vendors over the medium term, and relative upside for firms positioned around integrated chiplet/SoC supply (e.g., AMD).

Analysis

The Apple-driven push to fully integrated SoCs accelerates a structural reallocation of TAM from discrete component makers to integrated-SoC winners and fabs. Over the next 24–36 months expect wallet-share to shift away from aftermarket DIMM/PCIe ecosystems (volume down high-single-digits CAGR) and toward higher-margin, factory-configured SKUs and services recurring revenue; that amplifies gross-margin leverage for firms that control both silicon design and supply (vertical integrators).

AMD is best positioned to monetize this cycle because its chiplet strategy decouples node risk from system-level integration, letting it scale GPU and fabric IP across PC and server form factors without the same single-monolithic-yield penalty; conservatively, win rates in OEM bid processes of +10–20% would translate into mid-teens EPS upside over 12–24 months. Nvidia’s valuation is most exposed to a re-rating if discrete volumes compress—yet its software and datacenter entrenchment create a large tail risk: a sustained enterprise AI capex surge could keep discrete HBM-based cards in tight demand for 6–24 months.

Second-order winners include TSMC/OSAT ecosystems and open-driver-friendly GPU stacks (which favor AMD on enterprise Linux deployments); losers include aftermarket PCIe/GPU vendors, DIMM suppliers over the medium term, and any OEMs that can’t pivot to factory-configured upsell economics. Key catalysts: OEM design-win announcements and server procurement cycles (next 2–8 quarters), GPU inventory weeks on hand reports (monthly/quarterly), and TSMC capacity allocations (semi-annual guidance).

Key Decisions for Investors

• Pair trade (6–18 months): Long AMD equity or buy AMD 12–18 month LEAPS (target +35–45% upside) funded by selling covered NVDA 12–18 month calls. R/R: asymmetric—capture secular integration wins while using option premium to finance exposure. Risk: NVDA AI capex surge; hedge with a protective call on NVDA if NVDA re-rates >20%.
• Relative-value short (6–12 months): Short NVDA modestly (5–10% book) with a strict stop-loss (20%); hedge by buying a small long position in AMD or an AMD call. R/R: short thesis captures potential compression in discrete GPU TAM; tail risk is sustained datacenter demand that keeps NVDA revenue growing >25% YoY.
• Tactical long AAPL call spread (3–12 months): Buy a near-ATM 6–12 month call and sell a higher strike to fund cost—position size small (2–4% net long). R/R: play premium pricing/services lift as Apple-led integration normalizes; downside limited by spread width. Risk: consumer cyclical softness or supply constraints that delay upgrade cycle.
• Selectively avoid/underweight Intel (6–24 months): Reduce exposure to INTC hardware growth bets and reallocate proceeds to AMD or AAPL exposure. Use a protective put for any short-term allocation you must hold; rationale: secular headwinds in discrete/upgradable PC demand and execution risk vs chiplet competitors.