By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panelists agreed that Micron's pivot towards long-term supply agreements and HBM integration is a significant strategic move, but they differ on the sustainability of AI demand and the risks associated with Micron's massive capex commitment. They also debated the protective power of the CHIPS Act subsidies.
Risk: Massive overcapacity risks by the 2027 fab completion date if AI inference demand plateaus or if custom silicon designers successfully optimize memory usage to reduce total chip count.
Opportunity: De-risking 'boom-and-bust' volatility and potentially re-rating Micron's valuation from a commodity manufacturer to a critical infrastructure provider.
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 →
Micron CEO Sanjay Mehrotra said on Thursday artificial intelligence has fundamentally changed the memory business, an industry prone to boom-and-bust cycles.
"Today there is no AI without memory. AI systems need more memory. They need higher performance memory. They need lower power memory," Mehrotra told Jim Cramer on "Mad Money." "So, the value of memory, that equation has totally changed."
His comments came in the shadow of a massive semiconductor fabrication site under construction near Micron's headquarters in Boise, Idaho, part of the company's planned $250 billion investment in U.S. manufacturing and research. The Boise site alone will eventually include two fabs, each roughly the size of 10 football fields; a single fab will have enough steel rebar to circle Earth twice, according to Mehrotra. The first Boise fab is expected to begin producing wafers in mid-2027. The scale of that investment reflects how dramatically Mehrotra thinks AI has altered the outlook for memory.
Memory has historically been a cyclical business, with periods of strong demand encouraging manufacturers to add capacity, only for excess supply to eventually drive down prices. However, Mehrotra — an engineer by trade who's worked in the chip industry for over 40 years and previously co-founded SanDisk — said AI is creating a more durable source of demand.
The opportunity extends beyond data centers, he said. Mehrotra said he expects autonomous vehicles, robots, and AI-enabled consumer devices to require increasingly large amounts of memory in the years ahead.
"Memory today is essential," Mehrotra said. "That's why I call it the strategic infrastructure of the AI era."
That demand is also changing the value customers place on memory, according to Mehrotra. Instead of customers simply soliciting bids and buying from whichever supplier offers the lowest price, he said memory must increasingly be designed alongside the processors and systems in which it will operate. Mehrotra said that makes memory essential to the performance of the broader system rather than simply another component.
"We are working closely with them earlier and earlier in their development cycle," he said. "Our customers recognize the value of memory, because memory is what is enabling them to design products that are driving growth engines for them."
He said Micron still can't produce enough to satisfy that demand.
"All our customers across our end markets will buy everything that we make," Mehrotra said, adding that data-center customers currently want roughly 50% more supply than Micron is able to commit.
The memory maker is gaining greater visibility into that demand through long-term customer agreements, another important shift for a business historically exposed to swings in the spot market. During the company's most recent earnings call in late June, Micron announced that it had signed five-year strategic agreements with 16 customers. Mehrotra said the company has since inked additional deals.
"They have committed to taking the supply," Mehrotra said. "So, this gives us assurance of demand."
*Sign up now** for the CNBC Investing Club to follow Jim Cramer's every move in the market.*
Questions for Cramer?
Call Cramer: 1-800-743-CNBC
Want to take a deep dive into Cramer's world? Hit him up!
Mad Money Twitter - Jim Cramer Twitter - Facebook - Instagram
Questions, comments, suggestions for the "Mad Money" website? [email protected]
Four leading AI models discuss this article
"Micron is successfully transitioning from a commodity-driven cyclical model to a contract-based 'strategic partner' model, which justifies a higher P/E multiple."
Micron's pivot toward long-term supply agreements and HBM (High Bandwidth Memory) integration is a structural evolution, not just cyclical hype. By locking in 16+ customers for five-year commitments, Micron is effectively de-risking the 'boom-and-bust' volatility that historically compressed margins. If they can maintain their current HBM market share lead against Samsung and SK Hynix, Micron’s valuation should re-rate from a commodity manufacturer to a critical infrastructure provider. However, the $250 billion capex commitment is immense; if AI inference demand plateaus or if custom silicon designers successfully optimize memory usage to reduce total chip count, Micron faces massive overcapacity risks by the 2027 fab completion date.
The 'strategic infrastructure' narrative ignores that memory remains a capital-intensive industry where technological obsolescence occurs rapidly, and long-term contracts offer little protection if a customer's own AI product fails to gain market traction.
"AI has created a genuine multi-year demand tailwind for memory, but Mehrotra is extrapolating scarcity into permanence while underweighting the capex arms race now underway across all three major memory makers."
Mehrotra's thesis rests on three pillars: (1) AI durably shifts memory from commodity to strategic component, (2) long-term contracts de-risk cyclicality, (3) $250B capex bet signals conviction. The data-center demand signal is real—50% unmet demand is material. But the article conflates *current* AI intensity with *structural* demand. Memory intensity per AI workload will improve (efficiency gains, quantization, pruning). Autonomous vehicles and edge AI are real but multi-year, not imminent revenue drivers. Most critically: Micron's capex assumes no supply-side competition. TSMC, Samsung, SK Hynix all investing heavily in memory. Long-term contracts are valuable but not ironclad—customers negotiate out during downturns. The cyclicality hasn't vanished; it's been *delayed* by scarcity.
If memory efficiency improvements accelerate faster than AI workload growth, or if competitors ramp capacity before 2027, Micron's $250B capex becomes stranded assets in a normalized market—and the 'strategic component' thesis collapses back into commodity pricing.
"AI-driven memory demand may be durable, but the magnitude and duration of that tailwind are uncertain, and a multi-hundred-billion capex program plus long-term supply commitments risk margin compression if demand normalizes or supply outpaces demand."
The article portrays AI as a durable, structural demand driver for memory, with Micron planning a massive capex push (Boise fab, part of a $250B plan) and locking in demand via long-term contracts. That could support volumes and perhaps pricing power if AI adoption stays elevated. Yet the piece glosses over key risks: the memory market is historically cyclical, capex intensity can lead to oversupply and margin compression, competition from Samsung and SK Hynix, potential delays/cost overruns on giant fabs, and the ROI of such a large investment given uncertain AI demand durability. Missing context includes exact ROI expectations, debt/financing implications, and how demand splits across DRAM vs NAND and data-center vs consumer/SaaS.
Even with AI demand, memory pricing can fall as new supply comes online and workloads diversify; long-term contracts may bind Micron to volumes and prices that underperform if AI growth cools or if a technology shift reduces memory content per AI operation.
"N/A"
[Unavailable]
"Micron's massive capital expenditure is de-risked by geopolitical necessity and government subsidies, creating a regulatory floor that transcends standard cyclical commodity risks."
Claude is right about the 'delayed' cyclicality, but everyone is ignoring the geopolitical leverage of Micron's $250B capex. By aligning with the CHIPS Act, Micron isn't just betting on AI; they are buying a regulatory moat. Domestic production subsidies effectively socialize the risk of their massive capex. Even if AI demand softens, the US government's desire for supply chain sovereignty makes Micron 'too important to fail,' creating a floor that pure commodity players lack.
"Government backing de-risks capex deployment, not the underlying demand thesis that justifies $250B in the first place."
Gemini's 'too important to fail' argument conflates political intent with economic reality. CHIPS Act subsidies reduce *capex risk*, not demand risk. If AI memory demand genuinely softens by 2027, the US government won't force customers to overpay for excess capacity—they'll redirect subsidies elsewhere. The regulatory moat protects Micron's *existence*, not its margins. That's a meaningful distinction Claude and ChatGPT both missed.
"CHIPS Act subsidies reduce capex risk but do not guarantee demand or margins; a large buildout could become stranded if AI demand stalls."
Gemini overstates the protective power of the CHIPS Act moat (speculative). Subsidies may lower capex risk, but they don't lock in demand or margins, and policymakers can reallocate support. Even with $250B in subsidies, a demand downturn or rapid efficiency gains could leave Micron with stranded capacity and squeezed pricing. The bigger risk is not 'being too important to fail' but a long, expensive buildout that doesn't translate into durable profitability if AI demand stalls.
[Unavailable]
The panelists agreed that Micron's pivot towards long-term supply agreements and HBM integration is a significant strategic move, but they differ on the sustainability of AI demand and the risks associated with Micron's massive capex commitment. They also debated the protective power of the CHIPS Act subsidies.
De-risking 'boom-and-bust' volatility and potentially re-rating Micron's valuation from a commodity manufacturer to a critical infrastructure provider.
Massive overcapacity risks by the 2027 fab completion date if AI inference demand plateaus or if custom silicon designers successfully optimize memory usage to reduce total chip count.