Worried about AI taking your job? New Anthropic research shows it’s not that simple

Anthropic's latest Economic Index finds AI adoption accelerating—44% of jobs can now use AI for at least a quarter of tasks (up from 36%)—and notes increasing capability with new models (Opus 4.5) and agent features like Claude Cowork that extend delegation and action-taking. Taiwan Semiconductor Manufacturing Company reported a blockbuster quarter with a 35% jump in profit, record revenue, and high‑performance computing chips now accounting for 55% of sales while advanced nodes (7nm and smaller) comprised over three‑quarters of wafer revenue, underscoring strong AI chip demand. Separately, leadership departures at Mira Murati’s Thinking Machines (co‑founders rejoining OpenAI) highlight talent competition in private AI ventures, and Google launched a US beta of Gemini’s opt‑in Personal Intelligence for paid subscribers with privacy controls.

Analysis

Market structure: The clear near-term winner is TSMC (TSM) — CNBC/TSMC data show HPC now ~55% of sales and >75% of wafer revenue from ≤7nm, implying sustained ASP power and scarce advanced-node capacity. Google (GOOGL/GOOG) is a second-tier winner as Gemini Personal Intelligence leverages ecosystem lock-in to monetize subscriptions; Apple (AAPL) is less exposed to this riff for now. Smaller AI startups and lower-skilled service providers face disruption or deskilling as Anthropic’s data shows rapid, uneven adoption across roles (44% of jobs now can use AI for ≥25% of tasks).

Risk assessment: Tail risks include a Taiwan/China escalation (massive supply shock to semis), swift privacy/regulatory action in US/EU targeting data-connected assistants (hit GOOGL), or a sudden model commoditization that collapses ASPs; probability medium but impact high. Immediate (days) effects: leadership churn hurts small private AI ventures and can lift incumbents like OpenAI; short-term (weeks–months): product monetization cadence and TSMC guidance; long-term (quarters–years): structural labor shifts and capex cycles. Hidden dependency: advanced-node concentration at TSMC creates single-point-of-failure risk and pricing leverage for foundries.

Trade implications: Tactical: overweight TSM (foundry exposure) and selective long GOOGL to play ecosystem monetization; size positions small (1–3% each) with options to limit downside. Use 6–12 month call-spreads on TSM (10%–30% OTM width) and 3–6 month call-spreads on GOOGL; hedge NVDA risk via small put protection or a modest short (0.5–1%) given elevated multiple. Key catalysts: Nvidia GTC (Mar 16–19), TSMC guidance, and any US/EU privacy bills in the next 90–180 days.

Contrarian angles: Consensus underprices privacy/regulatory backlash risk to Personal Intelligence—if regulation forces data constraints, GOOGL monetization could lag by >12–18 months, creating a drawdown of 15–30%. Conversely, the market may underappreciate TSMC’s pricing stickiness; historical parallel: 2016–18 memory/logic tightness produced multi-quarter margin expansion. Unintended consequence: talent churn (Thinking Machines -> OpenAI) historically concentrates capability at incumbents, increasing winner-take-most dynamics in AI infra.

Key Decisions for Investors

• Establish a 2–3% long position in TSM (TSM) within 1–4 weeks to play sustained AI chip demand; pair with a 6–12 month call spread (buy 10% OTM, sell 30% OTM) to cap cost; target +25–40% upside, stop-loss at -15%.
• Add a 1–2% long position in GOOGL (GOOGL/GOOG) and buy a 3–6 month call spread (buy 5–15% OTM, sell 25–35% OTM) to capture Gemini Personal Intelligence monetization; take profits at +20–30% or if major privacy legislation passes in US/EU within 90–180 days.
• Buy protective 3-month puts on NVDA (NVDA) ~5% OTM (size 0.5–1% portfolio) or short 0.5–1% outright to hedge valuation tail risk ahead of Nvidia GTC (Mar 16–19); close if implied vol falls >20% or NVDA rerates lower than peers.
• Implement a relative-value pair: long TSM (2%) / short NVDA (1%) over a 3–9 month horizon to capture foundry pricing power vs. endogenous model/accelerator valuation risk; rebalance if TSMC guidance misses by >3% or NVDA reports HPU-led demand >consensus.