What looks mispriced: the market may still underweight how quickly Phison's enterprise and AI-storage solutions are displacing low-value retail and standalone controllers. Management said AI ecosystem modules were 38% of Q1 revenue, versus about 10% for controllers. S5
What can break the setup: Q1 generated NT$15.2B of net income but used NT$13.8B of operating cash as inventory doubled. Phison's May offering circular says it generally lacks firm long-term customer volume commitments and orders can be reduced or cancelled. Higher NAND prices can lift booked revenue while simultaneously raising replenishment costs. S4 S18 J1
Action discipline: for new capital, wait for the statutory margin and cash bridge. For existing exposure, treat the event as a high-volatility memory-cycle position; a low headline P/E is not a substitute for earnings durability. J1
Directional reaction framework
Thresholds are analyst-defined scenario markers, not company guidance. No event-isolating options surface was sourced, so this report does not invent an implied-move band. J1
Q2 sales are a fact; the Street's EPS bar is unusually wide and basis-uncertain.
Fiscal.ai figures were frozen at retrieval. Individual analyst timestamps and a precise basic/diluted, TIFRS/non-TIFRS label were not exposed by the returned estimate schema.
| Metric | Frozen bar | Range / count | What is now known | Decision use | Source |
|---|---|---|---|---|---|
| Q2 2026 revenue | Mean NT$61.778B Median NT$60.633B | NT$60.221–66.116B 5 estimates | Official monthly sum NT$67.888B: +9.9% vs mean, +2.7% vs high | Remove revenue as the principal event surprise; test conversion to gross profit. | S1S6 |
| Q2 2026 EPS | Mean NT$101.53 Median NT$103.08 | NT$68.97–119.33 6 estimates | Not reported | Wide dispersion encodes gross-margin, opex and non-operating uncertainty. | S6 |
| Q3 2026 | Revenue NT$65.697B EPS NT$93.44 | 5 revenue / 6 EPS estimates | Not guided by company | Tests whether Q2 is a peak or a durable AI/module run-rate. | S6 |
| Q4 2026 | Revenue NT$67.951B EPS NT$70.27 | 5 revenue / 6 EPS estimates | Not guided by company | Revenue resilience with lower EPS implies substantial margin normalization. | S6 |
| FY2026 | Revenue NT$240.071B EPS NT$301.29 | EPS NT$173.69–401.94 7 revenue / 9 EPS estimates | H1 revenue NT$108.855B | Annual EPS panel is not additive with quarterly estimates; use only on its own basis. | S1S6 |
| FY2027 | Revenue NT$249.020B EPS NT$165.15 | EPS NT$95.37–265.39 | Not guided by company | Mean EPS falls 45.2% on 3.7% revenue growth: consensus explicitly models a profit peak. | S6J1 |
Known before the earnings release
- April, May and June revenue were NT$20.207B, NT$22.828B and NT$24.853B. S2
- Q2 revenue rose 65.7% QoQ and 279.5% YoY using exact monthly sums. S1J1
- Management said June boot-drive revenue rose 5,600% YoY and mobile-controller shipments 47%; these are company claims, not audited segment KPIs. S1
- The external NAND tape is tight and price-led; memory producers reported strong enterprise SSD demand. S14S15S16
Still unknown—and therefore event-moving
- Q2 statutory gross margin, realized gross margin and inventory-obsolescence charge.
- Reported/basic/diluted EPS and the TIFRS-to-Non-TIFRS reconciliation.
- Equity-method/JV profit, FX, fair-value, tax and share-count effects.
- Inventory, receivables, payables, borrowings, CFO and free cash flow at June 30.
- AI ecosystem/module revenue share; controller and retail mix; enterprise volume versus ASP.
- Q3/H2 supply allocation, R&D spend and customer visibility.
What gross margin roughly reconciles to the EPS bar?
| EPS outcome | Implied net income | Implied statutory GM | Interpretation |
|---|---|---|---|
| Consensus low · NT$68.97 | NT$15.25B | 40.4% | A sharp reset from Q1; close to Q4 2025's 41.9% normalized statement margin. |
| Consensus mean · NT$101.53 | NT$22.45B | 52.9% | The central event hurdle under the stated analyst assumptions. |
| Consensus median · NT$103.08 | NT$22.79B | 53.5% | Only modestly above the mean hurdle. |
| Consensus high · NT$119.33 | NT$26.38B | 59.7% | Requires Q1-like economics despite higher opex and NAND replenishment costs. |
Q1's income statement was extraordinary; its cash conversion was the opposite.
Phison's inventory build may be commercially rational under supply allocation. The investor question is whether contracted demand and pricing protection are strong enough to make that inventory cash-generative rather than merely profitable on paper.
Q1 income statement: which line is durable?
Q1 balance sheet: the inventory thesis is financed
Inventory: NT$72.199B, up 102.8% QoQ; it represented 51% of total assets. Accounts receivable rose 66.2% QoQ to NT$22.750B. S4J1
Funding: short-term borrowings rose to NT$16.845B from NT$2.351B; Phison also issued NT$6.054B of domestic bonds in Q1. Financing cash flow was +NT$19.233B. S4
Accounting: the quarter included a NT$616M inventory-obsolescence loss. Management described the impact as about 1.5 points of gross margin. S4S5
Post-quarter funding: Phison issued two US$400M zero-coupon overseas convertible tranches on May 26, due 2031, for foreign-currency raw-material procurement. Conversion begins August 27 at initial prices of NT$3,418.75 and NT$3,555.50, subject to adjustment. The proceeds are not included in the March 31 balance sheet. S18
The reported curve is vertical; the consensus curve says profits peak immediately.
Q2 revenue is official monthly data but the Q2 gross-margin point is intentionally blank. The EPS chart uses the vendor's normalized actual/estimate series so readers can see the expectation curve without pretending it is identical to company basic EPS.
Nine-quarter revenue and gross-margin path
Fallback: revenue rose from NT$15.895B in Q2 2024 to NT$40.967B in Q1 2026 and a known NT$67.888B in Q2 2026; statutory gross margin rose from 35.3% to 61.8% through Q1 2026, with Q2 unreported.
Vendor EPS actuals and frozen expectations
Fallback: vendor actual EPS rose from NT$3.60 in Q2 2025 to NT$68.00 in Q1 2026; frozen mean estimates are NT$101.53, NT$93.44, NT$70.27 and NT$51.96 for Q2 2026 through Q1 2027.
Accessible chart data
| Quarter | Revenue | Statutory GM | Vendor EPS actual / mean | Status |
|---|---|---|---|---|
| Q2 2024 (t−8) | NT$15.895B | 35.3% | NT$11.97 actual | Reported |
| Q3 2024 | NT$13.943B | 29.3% | NT$3.37 actual | Reported |
| Q4 2024 | NT$12.572B | 30.3% | NT$10.65 actual | Reported |
| Q1 2025 | NT$13.839B | 31.2% | NT$5.53 actual | Reported |
| Q2 2025 (t−4) | NT$17.890B | 29.2% | NT$3.60 actual | Reported |
| Q3 2025 | NT$18.137B | 32.6% | NT$10.01 actual | Reported |
| Q4 2025 | NT$22.799B | 41.9% | NT$21.05 actual | Reported |
| Q1 2026 (t−1) | NT$40.967B | 61.8% | NT$68.00 vendor actual | Reported |
| Q2 2026 (t) | NT$67.888B | Not reported | NT$101.53 mean | Revenue known / EPS estimate |
| Q3 2026 | NT$65.697B mean | Not available | NT$93.44 mean | Estimate |
| Q4 2026 | NT$67.951B mean | Not available | NT$70.27 mean | Estimate |
| Q1 2027 | NT$61.752B mean | Not available | NT$51.96 mean | Estimate |
Peers validate the NAND shortage—and raise Phison's quality bar.
The best read-through is not “memory is strong.” It is that enterprise/solution demand and NAND pricing are strong while consumer controllers, working capital and non-operating gains require separate treatment.
| Company / period | Company-reported evidence | Inference for Phison | Source |
|---|---|---|---|
| Silicon Motion Q2 2026 | Sales +32% QoQ; SSD controllers +5–10%, eMMC/UFS +15–20%, Ferri/boot solutions +110–115%; GM 50.2%. Inventory +30.6% QoQ, CFO −US$63.8M. GAAP EPS US$3.99 exceeded non-GAAP US$2.43 largely because of a US$74.7M investment gain. | The cleanest analogue says integrated solutions are outrunning controllers. Separate Phison's module mix, working capital and JV/FX effects from the headline. | S13 |
| SK hynix Q2 2026 | Revenue +51% QoQ; both DRAM and NAND prices rose significantly; demand exceeded supply and roughly ten customer LTAs were completed. | Validates tight supply and AI/eSSD demand. Higher NAND prices also raise replacement cost and can suppress consumer demand. | S14 |
| Samsung Q2 2026 | Memory revenue +62% QoQ; record DRAM/NAND bit sales and server mix; industry prices continued rising. PC/mobile demand moderated. | Enterprise strength can coexist with client weakness. Phison should split enterprise/module growth from controller/client exposure. | S15 |
| Micron FQ3 2026 | NAND revenue +99% sequentially: bits only mid-single-digit higher while price rose mid-80s. Data-center SSD revenue more than doubled; next-quarter outlook assumes price-increase moderation. | The current revenue surge is heavily price/mix-driven. Phison must show inventory economics and product value survive a slower pricing delta. | S16 |
Peer periods and business models are not perfectly aligned. Silicon Motion is a controller/solution analogue; SK hynix, Samsung and Micron are upstream memory producers. Read-throughs are analyst inference, not direct Phison results. J1
Eight questions that can distinguish value creation from a financed memory trade.
Management's Q1 commentary supplied a compelling strategic narrative. The Q2 call must turn it into auditable mix, margin and cash evidence.
What management said in Q1
- AI ecosystem modules were 38% of revenue and could exceed 50% over time.
- Standalone controllers were about 10%; retail was below 10% and expected to decline.
- Inventory was built because components could become difficult to secure from August through December.
- Some customer fulfillment was only 30–35%; three hyperscale designs were expected to enter production in Q3/Q4.
- 2026 R&D expense could exceed NT$20B and possibly NT$23B.
All are management claims from the Q1 call, not independently audited KPIs. S5
What would falsify the thesis
- AI/module mix fails to rise despite 66% sequential revenue growth.
- Statutory GM falls below 50% without a clearly temporary accounting bridge.
- Equity-method/FX gains explain a material part of the EPS beat.
- Inventory and borrowings rise again while CFO stays deeply negative.
- Q3 revenue guide/commentary falls below the NT$65.7B vendor mean or enterprise ramps slip.
- Management cannot quantify firm demand, inventory aging or customer concentration.
Cheap on FY2026 peak earnings; less cheap on the earnings reset that follows.
Price uses the last completed Taiwan session before the report date. The live August 3 session opened during research and was excluded to prevent timestamp mixing.
Frozen valuation anchors
NT$1,640 July 31 close; NT$362.6B market capitalization using 221.086M point-in-time shares. S8S6J1
5.44x FY2026 mean EPS; 5.17x NTM EPS for Q2 2026–Q1 2027; 9.93x FY2027 mean EPS; 10.39x FY2028 mean EPS. S6J1
The share price fell 31.2% in July and 41.8% from its June 2 closing peak. The de-risking is real, but so is the consensus profit cliff. S8J1
What the multiple does—and does not—say
It says: the market does not capitalize 2026 earnings as permanent. The annual panel expects revenue to rise 3.7% in 2027 while EPS falls 45.2%. S6
It does not say: that Phison is automatically cheap. Fiscal EPS coverage is dispersed, annual and quarterly panels are not synchronized, and working-capital funding can change the per-share outcome.
Omissions are deliberate: no price target is used because timestamped estimate provenance was unavailable; no implied move or whisper is shown because no auditable event-isolating source was obtained.
Illustrative operating scenarios—not price targets
- Revenue
- NT$67.888B
- Gross margin
- 45%
- Opex
- NT$12.0B
- Non-op
- NT$0.5B
- Tax
- 20%
Known sales do not convert; inventory/replenishment costs and lower non-op support dominate.
- Revenue
- NT$67.888B
- Gross margin
- 53%
- Opex
- NT$12.0B
- Non-op
- NT$2.5B
- Tax
- 15%
Low-50s statutory margin and Q1-like non-operating support reconcile to the vendor mean.
- Revenue
- NT$67.888B
- Gross margin
- 60%
- Opex
- NT$12.5B
- Non-op
- NT$2.75B
- Tax
- 15%
AI/module mix preserves near-Q1 economics even as R&D and revenue scale.
All cases use 221.086M shares and span the vendor consensus range. They are constructions, not forecasts or valuation targets. J1
Primary facts, vendor consensus and analyst inference are kept separate.
Cut-off: August 2, 2026 at 23:50 BRT / August 3 at 02:50 UTC. Market price is the July 31 completed close. Company claims are labeled; calculations can be reproduced from the stated inputs.
Monthly, Q2 and H1 revenue; management commentary on demand and product indicators.
Exact 2026 monthly revenue series used to sum Q2 and H1.
TIFRS and Non-TIFRS results, margin and EPS presentation.
Balance sheet, comprehensive income and cash-flow statement; filing copy distributed by Quartr.
Management claims on mix, inventory, supply, design wins, R&D and JV contribution.
Quarterly and annual mean, median, high, low, count and NTM valuation. Identifier resolved as TPEX-8299.
Historical revenue, gross profit, operating income and net income used for the trajectory.
July 31 completed close and June–July path; underlying data attributed by the page to S&P Global Market Intelligence.
Official meeting list and ten-day quiet-period policy; no Q2 meeting shown at cut-off.
Third-party August 13 Q2 release/presentation indication.
Third-party August 7 estimated earnings date, illustrating calendar disagreement.
Initial board authorization for up to US$800M for foreign-currency raw-material procurement; superseded for final terms by S18.
Product-line growth, margin, earnings bridge, balance sheet, cash flow and Q3 outlook.
Memory pricing, supply/demand, eSSD and customer LTA commentary.
Memory revenue, NAND/DRAM bit sales, server mix, pricing and end-market outlook.
NAND revenue, bits, pricing, data-center SSD, cash flow and pricing-cadence outlook.
Filed July 17 at 14:36:44; confirms August 13 at 17:00 Taipei, online, hosted by SinoPac Securities CLST Branch.
Two US$400M zero-coupon tranches, May 26 issue, raw-material use of proceeds, conversion terms and customer-order risk disclosure.
Q2 sum/surprise, margin-to-EPS bridge, scenarios, working-capital bridge, market cap, valuation, drawdown and reaction framework. Inputs and formulas are disclosed in the relevant sections.