Micron posts record $54.2 billion quarter as AI-driven memory demand outpaces supply

Revenue in memory is bits shipped multiplied by price per bit. The split between the two has to be inferred from what Micron reports ^[1]. Margin is the nearest proxy, and I think a high one mostly reflects price. Fourth-quarter GAAP net income was $37.70 billion, or $32.87 per diluted share ^[12], about 69.5% of revenue ^[21].

The forward numbers point the same way. "Micron delivered record fiscal 2026 results, and we expect an even stronger fiscal 2027," said Sanjay Mehrotra, Micron's chairman and chief executive ^[8]. According to mezha.net, the company's revenue guidance for the fiscal first quarter of 2027 runs $1.5 billion either side of a $61.5 billion midpoint, with non-GAAP earnings of $38.15 a share, plus or minus $1 ^[7]. The midpoint is about 13.4% above the quarter just reported ^[17]. The bottom of the range, $60.0 billion, would still be 10.6% higher ^[18].

Mehrotra also said "AI is becoming Super Intelligence (SI)" ^[10]. The guidance figures hold up without the rebrand. For a buyer, his more useful sentence is about contracts. Mehrotra said "our Strategic Customer Agreements provide added confidence in the durability of Micron's financial performance" ^[9]. Durable revenue for the seller is committed spend for the customer who signed. Reuters reported that customer commitments under long-term supply contracts rose to $32 billion, according to mezha.net ^[11]. That sum is about 59% of one quarter's revenue at the fourth-quarter rate ^[19].

For "even stronger" to show up as a 2027 price on a purchase order, demand has to keep running ahead of supply ^[8]. Mezha.net reports that demand for DRAM, HBM and server memory for AI is still outpacing supply ^[2]. Capacity is the other side. "We are increasing our investments in technology, products and manufacturing," Mehrotra said ^[3]. Net capital expenditure was $10.77 billion in the quarter and $27.37 billion for the year ^[5], about 20.5% of annual revenue ^[16]. The same report notes that the memory industry has historically been cyclical, and that a sharp expansion in production can ease a shortage and push prices down ^[22].

In my view, a team specifying AI servers for 2027 should hold memory at today's prices in its cost model and count relief from new capacity as upside. My context is a buyer without a long-term agreement whose only forward figures are Micron's own guidance ^[7]. Neither source gives a date for when the added capacity reaches customers.