Will HBM represent more than 30% of Micron's FY2026 total revenue?

The Q2 data dramatically changes the denominator calculus. FQ2 revenue of $23.86B (+75% QoQ) puts the annualized run-rate at ~$95B, and FQ3 is guided at $33.5B — implying FY2026 total revenue could reach $80-90B. For HBM to represent 30% of that, HBM revenue would need to be $24-27B for the full year. The previous estimate of ~$8B annualized HBM run-rate was already insufficient against a $75B denominator; now it must clear a far higher bar. HBM4 volume shipments have begun (36GB 12-Hi for NVIDIA Vera Rubin), which is a positive step, but the critical new fact is that non-HBM margins now exceed HBM margins — meaning the DRAM/NAND pricing supercycle is inflating the denominator faster than HBM-specific revenue can grow as a share. The 30% threshold has become materially harder to reach.

Let me work the math carefully. FQ1 revenue was $13.64B, FQ2 was $23.86B, FQ3 guided ~$33.5B. If FQ4 reaches ~$38-42B (reasonable given trajectory), FY2026 total is roughly $109-113B. For 30% HBM share, that requires $33-34B in HBM revenue for the full year. Even with aggressive HBM growth — say HBM doubles from ~$2B/quarter in FQ1 to ~$4B in FQ2, ~$7B in FQ3, ~$10B in FQ4 — that cumulative ~$23B still falls short. However, HBM4's higher trade ratios (>3:1 to 4:1) and the volume shipment confirmation provide a plausible upside path. The five-year SCA signed with a large customer could include significant HBM commitments. I give a slightly higher probability than the pure math suggests because of HBM4 pricing power uncertainty — if HBM4 commands a substantially higher ASP than HBM3E, the revenue share could accelerate nonlinearly in FQ3-FQ4.

The updated data makes the arithmetic nearly impossible. With FQ2 at $23.86B and DRAM at $18.8B (79% of total), the DRAM pricing supercycle is the dominant story — not HBM-specific growth. Sanjay explicitly confirmed non-HBM margins are now HIGHER than HBM margins, which is the clearest evidence that the denominator is outpacing the numerator. The committee's original finding that '30% requires significant acceleration' was based on a ~$75B annualized denominator; the actual denominator is now tracking 25-40% higher. No specific HBM quarterly figure was disclosed for FQ2, suggesting Micron is no longer highlighting HBM as a share-of-revenue metric — likely because the denominator is making the percentage less impressive. HBM4 shipments are confirmed but the revenue impact in FQ2 is uncertain and FQ3-FQ4 ramp timing cannot close the gap. The question is effectively answered by the denominator explosion.

The Q2 earnings data is unambiguously negative for this market. FQ2 revenue at $23.86B and FQ3 guided at $33.5B project a FY2026 total well above the $75B estimate in the baseline context. The denominator has grown dramatically while HBM revenue, though growing, cannot keep pace as a percentage share. The most telling new data point: non-HBM margins exceeding HBM margins. This means the extreme DRAM/NAND scarcity pricing is making conventional memory MORE profitable per bit than HBM — the opposite of the thesis that HBM would become a progressively larger share. Even with HBM4 volume shipments begun, the broad-based pricing supercycle is lifting all boats, with the non-HBM boats rising faster. The 30% threshold was a stretch at the old denominator; it is approaching implausible at the new one.

I want to give more weight to the bullish case than pure arithmetic suggests. Several factors could accelerate HBM share in the back half: (1) HBM4 volume shipments are now confirmed, and HBM4 has significantly higher ASPs than HBM3E — the revenue per unit is substantially higher. (2) The five-year SCA likely includes HBM commitments at premium pricing. (3) Data center bits TAM exceeding 50% of industry for the first time in CY2026 means the structural mix shift favors HBM-adjacent products. (4) HBM4 16-Hi (48GB) sampling opens a higher-ASP product tier. However, the critical constraint remains: no HBM-specific FQ2 revenue was disclosed, and the mid-60s% DRAM price increase confirms the denominator is running away. Even with these bullish factors, 30% for the full year is very demanding. I assign 25% — acknowledging the path exists but is narrow.

The updated context explicitly states the key tension: 'the DRAM/NAND pricing supercycle is inflating the total revenue denominator faster than HBM-specific revenue can grow as a share.' This is confirmed by the non-HBM margins exceeding HBM margins data point. The math is damning: FQ1 HBM was approximately $2B+ against $13.64B total (~15% quarterly share). Even if HBM grew 50% QoQ to ~$3B in FQ2, that is only 12.6% of $23.86B — the share actually DECLINED because total revenue grew 75% QoQ. This pattern will likely continue in FQ3 with $33.5B guided revenue. Cumulative FY2026 HBM share would need an extraordinary FQ4 (perhaps $10B+ HBM in a single quarter) to reach 30% for the year. That is implausible given the ~$2B quarterly baseline just two quarters ago.

The denominator problem has gotten significantly worse. FQ2 $23.86B and FQ3 guided $33.5B project FY2026 total revenue far above the original $75B estimate. HBM would need $24-33B cumulative for 30% share — roughly 3-4x the previous annualized run-rate. HBM4 shipments beginning is positive but insufficient to close the gap. Non-HBM margins exceeding HBM confirms the broad pricing surge is the dominant revenue driver.

The Q2 data makes 30% HBM share near-impossible on a cumulative FY2026 basis. Total revenue of $23.86B in FQ2 alone means FY2026 will likely exceed $90B. For 30% HBM share, cumulative HBM revenue needs to exceed $27B. With FQ1 HBM at ~$2B and likely FQ2 at ~$3B (generous estimate), that means FQ3+FQ4 combined would need ~$22B in HBM revenue — roughly $11B per quarter. Given HBM quarterly revenue was just ~$2B two quarters ago, this is not achievable.

HBM4 volume shipments and 16-Hi sampling provide upside optionality, and the five-year SCA could include material HBM commitments. Data center TAM exceeding 50% of industry for the first time supports structural HBM demand. But the arithmetic remains challenging — the denominator grew much faster than expected. Assigning 23% to account for the possibility that HBM4 ASPs are dramatically higher than HBM3E and could drive a nonlinear revenue acceleration in the back half that the simple math misses.