Dario Amodei Predicted at Davos in January 2025 That AI Could Surpass Humans "at Almost Everything" Within 2 to 3 Years. With Anthropic's Revenue Up 7-Fold This Year, Is That Timeline Still on Track?

Anthropic's annualized revenue run rate reportedly reached $65 billion, up 622% from $9 billion last year, and the company estimates 2028 sales could approach $195 billion ahead of a potential IPO. Despite the growth, the article advises investors to avoid the offering due to reported AI-model misalignment, cybersecurity risks, potential labor-market disruption, and the prospect of stricter government regulation. CEO Dario Amodei, along with OpenAI CEO Sam Altman and Elon Musk, has warned that AI development may need stronger independent safety oversight or a slowdown.

Analysis

The key investable signal is not the reported private-company run rate but the widening gap between AI demand narratives and independently verifiable economics. Before extrapolating any read-through to NVDA, investors should require confirmation of whether the figure represents contracted bookings, recognized revenue, or an annualized short-period usage spike; each implies materially different compute demand and durability. A regulatory or safety-review regime would likely shift spend from frontier-training velocity toward inference controls, audit trails, model monitoring and identity/security layers—supportive for PANW, CRWD, ZS and OKTA relative to pure accelerator beta.

Near term (days to 1 month), IPO speculation can reinforce private-AI valuations and public proxy demand, but NVDA’s incremental upside depends on hyperscaler capex translating into sustained GPU orders rather than model vendors recycling venture-funded compute commitments. Over 1-3 months, a high-profile model-security incident would compress frontier-model valuation multiples while benefiting cybersecurity vendors with enterprise incident-response exposure. The 6-18 month structural effect is more ambiguous: compliance costs create a moat for scaled labs and cloud platforms (MSFT, GOOGL, AMZN), but can reduce training cadence and lower the market’s assumed semiconductor demand slope.

Contrarian view: calls for independent safety review are not necessarily bearish for incumbent AI platforms. Prescriptive compliance can raise barriers to entry, concentrate enterprise workloads in a few auditable providers, and make customers more willing to deploy AI once liability frameworks are clearer. The bearish thesis is falsified if major labs maintain release cadence and cloud capex guidance while reported enterprise AI workloads continue converting into paid inference revenue; it is validated by capex cuts, deferred model launches, or a regulatory event that imposes pre-deployment approval requirements.

Key Decisions for Investors

• Do not treat an Anthropic IPO as actionable until the registration statement distinguishes recognized revenue, contracted backlog, gross margin and customer concentration; use the filing as a catalyst watch item rather than underwriting the reported run rate.
• Initiate a 3-6 month pair: long PANW or CRWD / short a modest NVDA basket exposure only if NVDA trades materially above its pre-earnings capex-implied range. The thesis is that governance/security spend rises even if frontier-training growth decelerates; cover the short leg if hyperscalers raise aggregate AI capex guidance or NVDA raises supply outlook.
• Prefer MSFT and AMZN over unlisted frontier-model exposure on a 6-18 month horizon: distribution, cloud capacity and enterprise compliance tooling monetize a regulated deployment environment. Risk is aggressive model commoditization or cloud AI margins failing to improve; reassess after the next two quarterly cloud-margin disclosures.
• Set an event-driven alert for material U.S. or EU pre-deployment safety-review rules. On confirmation, add cybersecurity/compliance exposure and reduce high-multiple AI infrastructure beta; absent binding rules, avoid positioning on voluntary industry rhetoric alone.