Leaks Predict $5000 RTX 5090 GPUs in 2026 Thanks to AI Industry Demand

Korean outlet Newsis reports unverified insider leaks that AMD and NVIDIA plan gradual, permanent GPU price increases starting January–February 2026, driven by AI demand and rising DRAM costs; the RTX 5090 (launched at $1,999) is claimed could reach as high as $5,000 by year-end. The report cites memory as comprising up to 80% of average GPU BOM and forecasts up to a 40% increase in memory prices by Q2 2026, while OEMs such as ASUS have signaled hardware price hikes; implications include potential margin upside for GPU vendors, demand compression in the gaming market, and supply-driven risks for OEMs and console launch timing — monitor confirmed vendor pricing, DRAM spot and contract trends, and OEM inventory decisions.

Analysis

Market structure: AI-driven GPU demand is creating asymmetric pricing power — vendors of DRAM (Micron MU, SK Hynix, Samsung/SSNLF) look like net beneficiaries if DRAM rises ~40% by Q2 2026, since memory is cited as ~80% of GPU BOM. OEMs (HPQ, DELL) and consumer-focused AMD (AMD) are direct losers from cost pass-through and softer volume; NVIDIA (NVDA) can sustain ASP hikes short-term but risks demand elasticity if flagship ASPs move from $1,999 → $5,000 (a >150% increase).

Risk assessment: Tail risks include regulatory export controls on AI accelerators, a simultaneous GPU+DRAM supply squeeze, or demand destruction if flagship pricing exceeds consumer willingness-to-pay (>~$3k threshold). Immediate (days) volatility likely on leaks; short-term (weeks–months) fundamentals driven by DRAM spot indices and OEM inventory turns; long-term (quarters) depends on enterprise AI accelerator adoption and custom silicon displacing GPUs. Hidden dependency: console and OEM contract windows — a missed console launch could amplify PC demand shock.

Trade implications: Tactical trades: long DRAM exposure (MU) into Q1–Q3 2026 as a play on +40% DRAM; hedge NVDA downside with 3–6 month put spreads sized to 1–2% portfolio risk. Consider a relative trade: long MU (2–3% net) / short AMD (1–2%) to capture margin divergence if memory costs compress AMD’s gaming margins more than NVDA’s datacenter mix.

Contrarian angles: Consensus assumes persistent scarcity; but leaks are unverified and history (2021 crypto GPU spike) shows rapid mean-reversion once supply rebalances. If NVDA sell-off >7% on these headlines, buy-dip size-limited because enterprise GPU demand is sticky; conversely, sustained DRAM price momentum >20% by Mar 2026 validates overweight DRAM suppliers.

Key Decisions for Investors

• Establish a 2–3% long position in Micron (MU) within 2–6 weeks, target +25–40% upside by Q3 2026 if DRAM spot rises 30–40%; set a 12% stop-loss and reassess after MU Q1 2026 results and DRAM spot index moves.
• Enter a pair trade: long MU (2% allocation) / short AMD (AMD, 1% allocation) — expect MU margin tailwind vs AMD gaming margin compression; target 20% pair spread tightening within 3–9 months, stop if AMD outperforms NVDA by >10% on product surprise.
• Buy a protective 3–6 month put spread on NVDA (size = 1–1.5% portfolio risk) to hedge downside from demand shock or regulatory action; choose strikes ~7–12% OTM to limit premium while capturing headline-driven moves.
• Trim 30–50% of consumer-PC exposure in HPQ and DELL over the next 30 days and redeploy proceeds into DRAM suppliers or NVDA datacenter exposure if NVDA sells off >7% — monitor OEM inventory turns and ASUS pricing announcements as triggers.