Micron's Memory Supercycle Challenges Nvidia's AI Valuation Lead

Artificial intelligence profits are migrating from graphics processors to memory, and that shift reframes the entire chip debate. The argument here is blunt: Micron now earns roughly as much as Nvidia while trading at about one quarter of the valuation. If high-bandwidth memory is the binding constraint on AI servers, then the memory supercycle matters more than the next GPU launch. That single comparison sets up everything that follows.

Micron reported a record fiscal fourth quarter ended September 3, 2026, with revenue of $54.23B, up from $41.46B last quarter and $11.32B a year earlier. According to Micron, GAAP net income reached $37.70B, operating cash flow hit $43.97B, and full-year fiscal 2026 revenue totaled $133.19B. Chief executive Sanjay Mehrotra framed the result around AI becoming Super Intelligence. Importantly, this was a 14-week quarter, not the usual 13 weeks.

A record quarter on an extra week

The shock was profitability, with gross margin near 87% against about 46% a year earlier, according to the speaker. Manufacturing costs rose roughly 15%, so this looks like a pricing power story rather than an efficiency gain. CNBC, citing an LSEG poll, said the Street expected about $51.07B in revenue and $31.61 in earnings per share, a bar Micron cleared. The same coverage notes the stock rose nearly 500% over the past year.

Forward guidance was equally aggressive, pointing to roughly $61.5B in revenue plus or minus $1.5B and about $38.15 in earnings per share. The Street was near $57B and $35.40, so the guide implies another beat. Behind it sits committed capacity , with more than three-quarters of 2027 output already spoken for and early talks covering 2028. Customers have prepaid $12.7B in deposits, while 26 take-or-pay deals represent about $150B in future revenue.

Guidance backed by prepaid demand

Skeptics get their turn, because the next-quarter margin guide sits near 86.25%, slightly below the peak, alongside about $1B in extra bonus and capacity costs. Operating expenses are expected to rise $2.5B next year as new plants ramp. Management still uses margin-floor language for the year, suggesting confidence rather than caution. A Bloomberg headline about narrowing margins gets dismissed by the speaker as missing this broader contracted-demand context.

Nvidia still posted enormous numbers for its second quarter of fiscal 2027 ended July 26, 2026, with revenue of $96.221B, up 106% from last year. According to Nvidia, data center sales reached $89.0B, gross margin held at 75.0%, and third-quarter guidance points to $108B with no China data center sales assumed. Vera Rubin is now in full production. Yet SiliconAnalysts notes management called memory costs extreme, with supply covering only about 70% of demand.

Nvidia still grows, but memory binds

That contrast fuels the valuation gap at the heart of the pitch. Micron earned nearly $38B against roughly $40B for Nvidia, yet Micron was valued near $1.2T versus about $4.8T for Nvidia. Goldman Sachs is cited as attributing 18% of expected S&P earnings growth to Nvidia and 14% to Micron. If profits are converging while prices have not, the speaker argues, memory shareholders are buying the same AI earnings stream cheaper.

The Street still looks cautious, projecting only about 11% profit growth for Micron through fiscal 2028 versus roughly 68% for Nvidia, a forecast the speaker calls lazy. Part of the rebuttal is domestic supply, since Tomshardware reports Idaho Fab 1 comes online in the second half of 2027 under a 20-year $200B United States plan. HBM packaging in Virginia adds another onshore leg. With the only United States-based high-bandwidth-memory supply, scarcity supports the supply-constrained growth thesis.

Valuation math and the bear case

Even harsh stress tests look survivable, because halving memory prices would still leave margins near 74% from an 87% starting point. Counterpoint Research data cited by EastAsiaBrief puts second-quarter high-bandwidth-memory revenue share at 50% for SK hynix, 33% for Samsung and 18% for Micron. Samsung Securities is cited for allocating about 22% of Nvidia high-bandwidth-memory supply to Micron and 60% to SK hynix. The real threat is Chinese entrant CXMT, reportedly adding 85,000 wafer starts monthly versus 30,000 for Micron.

The closing argument returns to contracted demand rather than spot-market luck. Custom next-generation memory co-developed with Nvidia, multi-year take-or-pay coverage and deposits suggest visibility that memory has rarely enjoyed. The speaker discloses owning both Micron and Nvidia shares, with a larger Micron position, and runs a self-described contrarian growth strategy. Readers can follow more work through a single mention of rebellionaire.com. If memory stays the bottleneck, the rerating may still be early.