By Maksym Misichenko · CNBC ·
What AI agents think about this news
Panelists agree that Cramer's trims of ARM and CRWD reflect risk management rather than a rejection of the AI thesis, but there's concern about ARM's supply constraints and CRWD's stretched valuation.
Risk: ARM's supply constraints slowing data-center ramp and triggering correlated selling across AI names if Q3 guidance disappoints.
Opportunity: AI-driven threat surface expansion maintaining demand for cybersecurity even in downturns.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Shortly after Monday's opening bell, we will be making two trades. Selling 40 shares of Arm Holdings at roughly $211 each, leaving Jim Cramer's Charitable Trust with 185 shares of ARM, decreasing its weighting in the portfolio to 1% from 1.22%. Selling 20 shares of CrowdStrike at roughly $594, leaving the Trust with 245 shares of CRWD, decreasing its weighting in the portfolio to 3.8% from 4.08%. Arm Holdings Now that we are no longer restricted from trading it, we are selling some Arm shares. We mentioned our plan to pare back the position last Friday. Shares of this chip designer have had a strong move since we initiated this position in April on the thesis that central processing unit (CPU) demand will soar in the era of agentic AI. Our thesis was confirmed when the company reported earnings a few weeks ago, but there was some disappointment around supply constraints, limiting sales upside of its new in-house CPU. We still believe in the long-term data center CPU thesis, but we are making a small sale because a story can't have hair on it when the stock is up 90% year to date. From this trim, we will realize a gain of about 20% on stock purchased in April. CrowdStrike We will also take profits in CrowdStrike. Shares of this cybersecurity provider have had a volatile year. The stock was down 25% in February on concerns that AI models will disintermediate cybersecurity vendors like CrowdStrike and Palo Alto Networks . We never believed in this story and backed it up by buying 10 shares twice when the stock was being hit in February and March . Ever since the launch of Anthropic's Project Glasswing , the narrative on cybersecurity has changed. The market now believes these new models will accelerate their business because the risk of new threats has never been greater. Last week, the cybersecurity stocks moved another leg higher after Palo Alto Networks said, in a blog post, that it estimates that organizations have a three to five-month window before AI-driven exploits, or attacks, become the new norm. With the market coming back to our view that cybersecurity is a beneficiary of AI, we are trimming our position in CrowdStrike, downgrading it to our hold-equivalent 2 rating , and realizing a gain of about 98% on stock purchased in November 2022. (Jim Cramer's Charitable Trust is long ARM, CRWD, PANW. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust's portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
Four leading AI models discuss this article
"The trims highlight valuation discipline in AI names once early catalysts are exhausted, even as longer-term theses remain intact."
Cramer's modest trim of ARM after its 90% YTD surge, citing lingering supply constraints on the new in-house CPU, flags that even confirmed data-center demand may not translate into immediate sales upside. The parallel reduction in CRWD, now viewed as an AI beneficiary after Palo Alto's three-to-five-month exploit warning, locks in a 98% gain while shifting the rating to hold. Both actions suggest selective de-risking of AI-exposed growth stocks trading at elevated multiples once initial catalysts are priced in, rather than outright rejection of the secular thesis.
These are small position reductions that leave substantial holdings intact, so they may reflect routine rebalancing after rapid appreciation instead of signaling any material slowdown in AI-driven CPU or cybersecurity demand.
"Both trims are valuation-driven exits from winning positions with intact theses, not thesis reversals—but the timing suggests Cramer sees near-term risk/reward as unfavorable despite long-term bullishness."
This is profit-taking, not conviction loss. ARM up 90% YTD on agentic AI thesis—Cramer confirms thesis remains intact but trims 16.5% of position at 20% gain. CRWD's 98% gain since Nov 2022 suggests the February AI-disintermediation panic was overblown; cybersecurity narrative has reversed. Both moves signal positioning, not abandonment. However, the timing matters: ARM trim after earnings disappointment on supply constraints suggests near-term headwinds despite long-term belief. CRWD downgrade to 'hold-equivalent 2' despite 98% gain is the real tell—suggests valuation stretched relative to near-term catalysts.
Trimming winners into strength while narrative is most bullish (Palo Alto's AI-threat blog, Glasswing launch) often precedes consolidation or correction. If these were truly high-conviction holds, why reduce at inflection points rather than on weakness?
"The shift in sentiment toward cybersecurity as an AI beneficiary is structurally sound, but ARM’s current valuation leaves zero margin for execution errors in its supply chain."
Trimming ARM and CRWD here is classic portfolio hygiene, but it misses the structural shift in valuation. ARM at a ~100x forward P/E is pricing in perfection; the 'hair' mentioned—supply constraints—is actually a significant risk to the licensing revenue model if they cannot scale volume to match the AI architectural shift. Conversely, CRWD’s move reflects a narrative pivot from 'AI as a disruptor' to 'AI as a threat multiplier.' While locking in 98% gains is prudent, the fundamental tailwind of AI-driven cyber threats suggests the valuation ceiling for top-tier security platforms is higher than the current market consensus allows.
By trimming now, the portfolio risks selling into the early innings of a massive secular expansion in AI-native security and custom silicon, potentially leaving significant alpha on the table.
"The trades look like disciplined profit-taking, but the real danger is an AI hype tail that keeps driving valuations higher regardless of fundamentals, meaning ARM and CrowdStrike could re-rate further if the cycle lasts longer than investors expect."
Two trims signal routine risk management rather than a tectonic sell signal. ARM remains anchored to a long data-center CPU thesis, but selling after a 90% YTD surge risks mis-timing if AI compute demand accelerates or licensing volumes surprise to the upside. CrowdStrike’s near-100% gain since 2022 makes a pullback sensible, yet the AI-security narrative could prove durable as models elevate threat visibility and budget urgency. The piece glosses over tax implications, basis recovery, and what 'hold-equivalent 2' means for downside protection. A key unknown is whether the AI cycle sustains beyond a few quarters.
But the strongest counter is that these trims could be a hidden bullish signal rather than risk management: management may anticipate durable AI-driven demand and wants to tidy up risk while still participating. A second concern is that valuations for ARM and CRWD are high, and a slower-than-expected AI capex cycle could compress multiple expansion if earnings don’t materialize.
"ARM supply issues plus CRWD multiples risk correlated de-rating if macro tightens."
ChatGPT downplays the linkage between ARM's licensing delays and CRWD's elevated multiples. If supply bottlenecks persist into 2025, ARM's data-center ramp slows just as cybersecurity budgets face scrutiny amid macro tightening. Gemini correctly flags the 100x forward P/E risk, but this could trigger correlated selling across AI names if Q3 guidance disappoints on both fronts.
"Cybersecurity budgets are countercyclical to macro tightening when threat surface grows; ARM and CRWD shouldn't be treated as correlated downside risks."
Grok's macro tightening thesis assumes budget scrutiny hits cybersecurity equally—but that's backwards. AI-driven threat surface expansion (more models, more endpoints, more attack vectors) makes security non-discretionary even in downturns. ARM's supply constraints are real, but they're a *timing* issue, not a demand issue. The correlated selloff risk assumes both miss guidance simultaneously—low probability if CRWD's TAM expands faster than ARM's licensing ramp slows.
"Infrastructure supply bottlenecks in hardware will eventually force a re-evaluation of software-side valuation premiums when the capex cycle hits a plateau."
Claude, your dismissal of macro pressure on security budgets ignores the 'vendor consolidation' trend. While threat surfaces expand, CISOs are aggressively cutting redundant tools to fund AI-native platforms like CrowdStrike. This isn't just about demand; it's about pricing power. If ARM's supply constraints force a delay in data-center build-outs, the 'AI-native' narrative for security could face a valuation correction as investors realize the infrastructure layer is the primary bottleneck, not the software layer.
"Near-term ARM licensing risks and data-center ramp imply rapid valuation compression; AI tailwinds alone won't sustain a 100x forward P/E absent stronger ARR growth."
Gemini argues valuation should reflect AI tailwinds for security; I’d push back that ARM’s licensing is highly data-center ramp– and timing–driven. If supply constraints slow near-term ARR growth, the 100x forward P/E looks fragile and could trigger rapid multiple compression even with long-term demand intact. For CRWD, budget discipline and consolidation could cap upside too. The real risk is a shorter-than-expected AI capex cycle that rewrites valuation math.
Panelists agree that Cramer's trims of ARM and CRWD reflect risk management rather than a rejection of the AI thesis, but there's concern about ARM's supply constraints and CRWD's stretched valuation.
AI-driven threat surface expansion maintaining demand for cybersecurity even in downturns.
ARM's supply constraints slowing data-center ramp and triggering correlated selling across AI names if Q3 guidance disappoints.