Grok’s autonomous stock portfolio has built a sizable lead over Claude’s newer fund on the Autopilot mirror-trading platform. The result exposes a widening performance gap between two AI agents picking real-money trades.
The two X accounts, @grkportfolio and @theaiportfolios, run separate experiments by AI Finance Labs. The wider lineup of AI-managed strategies on Autopilot holds roughly $150 million in mirrored capital.
Grok Builds a Multi-Quarter Lead in AI Stock Picking
Grok’s portfolio has returned 59% over its first nine months, according to publicly available Autopilot data, with $17 million currently invested. The S&P 500 climbed 36% in the same window.
Over the past three months, the agent added another 12.6%, against the SPY’s 9.75%. The performance has stayed concentrated in AI infrastructure and energy names.
Heavy positions in semiconductor and memory stocks captured the hyperscaler capital expenditure cycle. Defense and power exposure provided ballast during macro shocks earlier in 2026.
The hardware bet aligns with the broader pattern documented across coverage of AI agents now moving billions in real markets. The xAI-branded portfolio operates without human override, according to its public posts.
Follow us on X to get the latest news as it happens
Claude’s Defensive Tilt Misses the AI Rally
Claude’s portfolio launched in April 2026 with $50,000 in seed capital. Every decision runs through the agent with no human override.
Recent posts show the fund rotated into ServiceNow (NOW) and Zeta Global (ZETA) while trimming Microsoft (MSFT).
The agent justified each trade through probability-weighted scenarios, kill conditions, and forward catalysts.
The discipline has not produced the same headline returns. Operators have acknowledged in public posts that the fund trails the S&P 500. The gap covers several points across its first two months on the platform.
“It’s now been about two months since Claude’s AI agents started picking stocks So far, they’ve underperformed the market SPY: +8.3% Claude: 2.6%,” they said.
Claude leaned into second-order AI plays such as enterprise software, fintech, and power. That tilt missed parts of the mega-cap rally that lifted direct semiconductor names.
Outside the experiment, independent traders have used Anthropic’s models to power Polymarket bots that reportedly cleared millions in profit.
What the Gap Means for AI-Driven Investing
The headline numbers tell only part of the story. Grok has roughly a year of public history. Claude’s track record covers weeks.
Live AI-led trading carries real risks. Anthropic recently received legal warnings about how its name appears on retail products.
Market professionals have also questioned the value of retail AI trading bots.
“They lack real intelligence, so expecting them to trade and consistently beat humans in a reasonable timescale doesn’t make sense,” highlighted Raullen.eth, an AI builder and popular user on X.
Anyone mirroring either agent faces fees and concentration risk. Strong recent returns may not repeat once the cycle turns.
The two portfolios offer a rare public test of how different AI models translate market data into trades. Still, events such as earnings and sector rotations could reveal whether aggressive infrastructure bets or hedged software exposure hold up better.
The post Grok Outpaces Claude AI in Stock Trading With 60% Profit appeared first on BeInCrypto.
