AI rapidly improving, could vastly boost U.S. GDP, Anthropic's Clark tells policymakers

Anthropic co-founder Jack Clark said one internal-model forecast suggests AI could lift U.S. GDP to nearly $44 trillion by 2030, from just over $30 trillion, under conditions including continued technology improvement. He warned that the window for safeguards is shrinking to a few years, highlighting risks including biological and cyber misuse, while describing potential gains in scientific discovery and possible job creation alongside losses in some high-exposure fields. Anthropic is reportedly preparing for an IPO expected in November that could value it at up to $2 trillion.

Analysis

The market risk is treating a conditional productivity scenario as near-term earnings evidence. Even if AI improves output over time, the first-order investment cycle can precede broad customer savings: compute, power, cooling and networking costs arrive before adoption is reflected in productivity or margins. That favors infrastructure suppliers only while orders convert to revenue; it does not validate every AI software valuation. Conversely, customer-service automation may pressure exposed labor-intensive providers, but contract repricing and implementation costs could delay the earnings impact.

The more durable second-order risk is a split market: countries and firms with reliable power, capital and data access may capture gains, while constrained regions fall further behind. Tighter controls on biological and cyber uses could raise compliance costs and slow deployment, while increasing demand for security and audit tools. The claimed IPO valuation is not independently verifiable from these remarks; an offering would provide a useful disclosure catalyst, not proof of the macro thesis.

Near term, expect narrative-driven volatility rather than a reliable GDP signal. Over 1–3 months, watch IPO filings and AI infrastructure orders; over 6–18 months, test whether customers report measurable productivity, margin gains and new demand. The thesis weakens if data-center order growth or utilization rolls over, power constraints defer capacity, or customers fail to show AI-linked productivity gains.

Key Decisions for Investors

• Avoid buying broad AI exposure solely on the GDP projection. Treat it as a scenario, not a forecast; require evidence from customer adoption, realized savings and revenue conversion.
• Consider a relative-value expression: long diversified data-center power, grid and cooling suppliers against a basket of cash-burning AI application businesses, sized conservatively. Reassess after earnings disclosures on orders, utilization, capex and monetization; exit if infrastructure demand weakens or application revenue converts faster than expected.
• Keep customer-service and business-process outsourcing names on a watchlist rather than shorting indiscriminately. Seek evidence of contract losses, lower renewal pricing or falling headcount before positioning; wage savings could accrue to customers rather than the vendors.
• Before acting on the prospective Anthropic IPO, wait for offering documents and verify revenue mix, customer concentration, compute commitments, cash burn and governance. A materially different disclosure set or delayed offering would invalidate any valuation read-through.