Software Bear Market: 2 Monster Artificial Intelligence (AI) Stocks With up to 70% Upside to Buy Now, According to Wall Street

Anthropic released the Claude Cowork plugin suite targeting enterprise software, coinciding with a sector sell-off (application software -21% YTD, software infrastructure -14% YTD). Palantir (PLTR) is highlighted for >50% YoY revenue growth, a 1,900% rally since ChatGPT, shares down ~16% YTD, $4.4B remaining U.S. commercial deal value and 325 Q4 deals closed, with Citigroup assigning a $260 price target (~70% upside). Amazon (AMZN) reported strong Q4/2025 results and guided ~$200B capex for 2026; AWS run-rate rose from $97B (end-2023) to $142B (Q4 2025) with margins climbing to ~35%, and analyst price targets ranging roughly $300–$360 (high end ~70% upside).

Analysis

The arrival of broadly capable models is accelerating a bifurcation: capital and revenue will concentrate toward providers of compute, inference plumbing, and firms that own persistent, proprietary data schemas (ontologies). Expect semiconductor winners (NVDA) to capture more training/inference share over the next 12–36 months as customers prefer off-the-shelf high-efficiency accelerators for near-term rollouts, while hyperscalers push bespoke inference silicon that will only materially dent GPU volume after ~18–36 months. Second-order supply-chain effects include faster reuse cycles for datacenter power/cooling, a step-up in wafer capacity allocation decisions at foundries, and outsized services demand (integration, security, governance) that benefits consultancies and premium data vendors.

Key risks are not headline model quality but trust, governance, and deployment economics. Near-term price action (days–weeks) will be dominated by sentiment and analyst updates; meaningful enterprise migration and contract renewals play out over 6–24 months, and a structural shift to bespoke silicon is a 12–36 month conditional catalyst that can truncate baseline TAM expectations for GPU incumbents. Regulatory limits on data portability or safety-driven constraints could strand parts of the market and create flight-to-quality dynamics benefiting incumbents with enterprise controls.

Market dislocations are creating tactical entries: software multiple compression is likely overstated for specialized data-analytics franchises and for market-structure players that monetize transaction and listing flows (NDAQ). Expect M&A and advisory fee tailwinds which support cyclical financials and market infrastructure revenues; conversely, broad horizontal SaaS names without sticky data assets remain vulnerable to consolidation or margin compression if they fail to prove differentiated ROI within 12 months.

Key Decisions for Investors

• Long NVDA (6–12 month call spread): Buy 3–6 month bulls if a <10% pullback forms; target asymmetric +30–60% upside over 6–12 months vs a Tmax 15% downside stop. Rationale: capture incremental GPU demand and higher ASPs while limiting theta risk.
• Pair trade — Long NVDA / Short INTC (6–18 months): Size 1.5:1 NVDA exposure vs INTC to express share shift toward accelerators; use INTC puts or a small outright short. Reward: NVDA upside capture if GPU demand persists; risk: Intel secures hyperscaler custom silicon wins — cap losses at 20% of position.