AI Bots Crashed $10K in 2 Weeks While Our "Beat-the-Bot" Divvie Sails Higher

The article argues AI trading models performed poorly in a two-week experiment, losing about 33% of capital overall and underscoring limits in automated portfolio management. It then turns constructive on Amgen, highlighting 70% gross margins, a 29% share count reduction over a decade, and $5.2B of rare-disease revenue last year, with Tepezza contributing $490M in Q1 2026. Management’s $27.8B Horizon Therapeutics acquisition is cited as a key growth driver, while Wall Street price targets have moved into the $307-$375 range versus a recent pullback from above $388.

Analysis

The real market signal here is not that AI is “bad at trading,” but that generic pattern-recognition systems are structurally vulnerable in markets with low signal-to-noise and reflexive execution. The failure mode is overtrading plus style drift: when a model is forced to express a view continuously, it will manufacture certainty, which amplifies turnover and transaction-cost drag faster than any edge can compound. That matters for crowded systematic strategies more broadly, because the same pathology shows up in quant sleeves that optimize on recent history and then overfit into the next regime change.

AMGN is the cleaner takeaway. This is a capital-return compounder with multiple self-reinforcing levers: dividend growth, buybacks, and a post-M&A earnings mix that is less dependent on single-product sentiment than the market appears to price. The second-order effect is that the stock can re-rate without needing multiple expansion from a broad biotech cycle; it only needs sustained execution and continued capital return, which makes it more resilient than AI-adjacent names that require perpetual estimate upgrades.

Consensus likely misses how much of the downside risk is now embedded in a narrow narrative around one drug or one quarter. That kind of headline risk can create repeated entry points for long-only capital, especially when the balance sheet and cash generation allow management to keep shrinking the float. The setup is a slow-moving “price magnet” where the path matters less than the endpoint, and the endpoint can be pulled higher by buybacks and dividend hikes even if revenue growth moderates.

The broader implication is that this is a relative-value world, not a binary growth-vs-value world. If markets keep rewarding perceived AI certainty, the under-owned, cash-generative healthcare compounders become even more attractive on a risk-adjusted basis because they monetize today’s cash instead of today’s narrative. That makes AMGN more interesting as a funding leg against higher-multiple, story-driven exposures than as a standalone long.

Key Decisions for Investors

• Go long AMGN on weakness in the next 2-6 weeks; target a 10-15% move back toward prior highs over 6-12 months, with downside limited by dividend support and buyback flow.
• Pair trade: long AMGN / short NVDA on a 6-12 month horizon as a valuation-and-cash-return hedge; the thesis is not bearish NVDA outright, but that AMGN has lower execution variance and better downside carry.
• Sell cash-secured puts on AMGN at strikes ~5-8% below spot for the next earnings cycle; attractive if implied volatility stays elevated from product-specific headline risk.
• If already long biotech, rotate from smaller-cap single-asset names into AMGN as a lower-beta capital-return anchor; use any post-earnings selloff as a staged entry rather than chasing strength.
• Avoid initiating fresh short exposure in AMGN unless the stock loses its dividend-growth narrative for two consecutive quarters; the combination of buybacks plus payout growth makes outright shorting unfavorable on a 3-6 month horizon.