experimental vintage — single-model run (fable, prompt postcall-v3); the slate re-run supersedes this page when it lands · transcript 288555e234756340-whisper-llm-v1 · response 39fee597385f67d2…
| ✓8-K results, 2026-08-12 | SEC filing · byte-verified · print + new guide |
| ✓Earnings call, 2026-08-12 | self-transcribed webcast replay · lower trust class, every citation checked against it |
| ◷10-Q | not yet filed · this page revises again when it lands |
| metric | our number | company guide | ||||||
| Revenue $M | 18,300 | 18,100 ± 100 | 18,050 18,650 was 16,450 pre-earnings — raised +1,850 on the print, guide, and call | |||||
| Adjusted EPS | 1.35 | 1.33 ± 0.01 | 1 1 was 1.11 pre-earnings — raised +0.24 | |||||
| Free cash flow $M | 4,300 | not guided | 3,700 4,950 | |||||
| ||||||||
| amplify | The company beat the high end of its own guide across revenue, margin, and EPS, and its new guide should be read the same way (floor-setting). · turn 4 | The graded record shows a 452M beat vs its own guide midpoint (+2.7%) and an EPS beat vs the in-force guide; a single observation, but combined with 'prudent' language on the call it argues the FY27 guide is a floor. |
| amplify | AI infrastructure revenue will grow to $7.5B in FY27, from ~$4B recognized on $9.3B of FY26 orders. · turn 3 | Backlog math is corroborated on the call itself ('we did about $4 billion in revenue and $9 billion in orders for FY26') and management called $7.5B 'a good prudent guide' — coverage from existing backlog alone is high, so modest upside is more likely than shortfall. |
| discount | We are 'only at the beginning' of a multi-year networking supercycle. · turn 3 | Classic testimony at a moment of strength; orders are lumpy hyperscaler capex, ~5 points of Q4 growth was price, and RPO (+7%) and ARR (+3%) grew far slower than orders (+35%), so I do not extrapolate the FY27 growth rate into out-years. |
| accept | Price increases contributed about five points of Q4 top-line growth and ~4-5 points are planned into FY27, lapping in the second half. · turn 20 | Management volunteered a quantified, unflattering decomposition (organic growth is lower than headline), which is credible precisely because it cuts against their narrative. |
| accept | Q4 security growth of 14% was flattered by sizable, longer-duration on-prem Splunk deals; underlying trajectory is high-single-digit for FY27. · turn 14 | Management itself flagged the anomaly against its own headline, so I underwrite security at high-single-digit FY27, not the Q4 print. |
| discount | No significant lead-time issues; supply is secured directly with TSMC and adequate even above the guide. · turn 17 | Unverifiable from the record, memory-cost inflation is acknowledged elsewhere on the call as an industry issue, and gross margin already fell 210bps y/y — I hold gross margin at the low end of the guided band and fading through the year. |
| accept | FY27 operating margin of about 35% would be a company high watermark, because hyperscale growth requires minimal incremental opex. · turn 11 | Q4 independently demonstrated the mechanism (opex down 3.7 points as % of revenue more than offsetting a 2.1-point gross margin decline), so the leverage claim is corroborated by reported numbers. |
| discount | Mythos/last-day-of-support refresh is a large coming driver, though 'we haven't seen a massive amount of impact from mythos yet'. · turn 23 | Pipeline anecdotes (a CEO text message) are not bookings; I underwrite zero incremental mythos revenue and treat it only as optionality above my out-year path. |
| discount | FY27 AI orders should be 'meaningfully higher' than FY26. · turn 29 | Simultaneously, management withdrew the annual order target in favor of a revenue target — a disclosure change that reduces verifiability exactly when orders are the bull case, so I take the direction but not the magnitude. |
| accept | Q2-Q4 implied growth of ~13% reflects tougher comps and normal linearity, not demand deceleration. · turn 11 | The comp math is verifiable from the reported FY26 quarterly ramp (14,883 to 17,252) and Q1-at-25%-of-year matches the stated three-year average. |
| year | revenue $M · EPS | growth, was → now | why it changed |
| FY2026 | 63,325 was 63,003 EPS 4.33 was 4.29 | +11.2% → +12.0% | Fully reported: four quarters sum to 63,325 (14,883+15,349+15,841+17,252); adj EPS 4.33 and capital return of 12.7B at 99% of FCF pin FCF near 12.8B. Aligned — these are reported facts; my sealed 63,003/4.29 was close but underestimated the Q4 hyperscale surge. |
| FY2027 | 73,300 was 67,925 EPS 5.15 was 4.66 | +7.8% → +15.8% | Re-rated up from my sealed 67,925/4.66: company issued 72.2-73.4B / 5.05-5.11 with $7.5B AI revenue largely covered by FY26 backlog (9.3B orders vs ~4B recognized). I underwrite the top of the revenue range and a nickel above the EPS range top because the one resolved guide observation shows a +2.7% midpoint beat and management repeatedly used 'prudent'; I do not go further because the pricing tailwind (~4-5 pts) is partly mechanical, not demand. |
| FY2028 | 79,800 was 72,500 EPS 5.70 was 5.08 | +6.7% → +8.9% | Raised from sealed 72,500/5.08 on the higher FY27 base and continued AI backlog conversion, but growth deliberately decelerates to ~9% as price increases lap and hyperscaler comps get very tough. Well below the management-implied supercycle continuation: I discount 'orders meaningfully higher in FY27' into only moderate FY28 revenue growth because RPO/ARR growth (7%/3%) shows the recurring base is not compounding with the hardware surge, and hyperscaler capex is cyclical. |
| FY2029 | 85,400 was 76,900 EPS 6.25 was 5.53 | +6.1% → +7.0% | Raised from sealed 76,900/5.53 on the compounding of the FY27 step-up, with growth fading toward high-single digits; Silicon One rollout across the portfolio by FY29 supports margin via silicon-margin capture even as hardware mix rises. Materially below a straight-line of management's supercycle narrative: silence on hyperscaler digestion risk, export exposure, and enterprise budget stretch is reflected as a ~7% growth rate rather than the ~10% ex-AI core management touts today. |
| FQ4 reported vs our sealed number (mechanical) | +322 |
| new guide plus call, combined (no pre-call vintage to split them) | +0 |
| total revision | +322 |