Texas Instruments Q2: analog snap-back, or scarcity spike?
The continued effect of the AI data center build out on analog chip suppliers
Texas Instruments (TXN) reported Q2 CY2026 on June 22nd. Texas instruments has recently seen a pretty massive surge in demand driven by the AI data center build out. Given analog semis are usually a highly cyclical business I thought I’d dig into what their most recent commentary tells us about continued demand. I built this report and generated these graphics with Easel. Easel is our newest agent surface currently in beta.
The report is as follows:
Setup
In October 2025, TXN’s F3Q25 print guided gross margin down 250 bps and the stock closed down ~9%. Sell-side takeaways were unanimous: shallowest analog recovery on record, weighed down by capex depreciation, auto/industrial deferrals, and tariffs. Only Data Center was working.
Three quarters later, TXN’s own quarterly tape describes the opposite. Revenue reaccelerated from an F4Q25 trough of $4.42B to $5.46B this print (+23% YoY, +13% seq), gross margin recovered from 55.9% at the trough to 61.4% — 550 bps of expansion in two quarters — and on-book DIO drained from 222 days to 196 while lead times ticked up for the first time this cycle.
Grounded in TXN 8-K earnings releases and earnings-call transcripts, F3Q25 through F2Q26; F3Q26 shown at $5.90B guide midpoint with the $5.65–6.15B range as whiskers.
The debate ahead of TXN’s July 22 print was no longer “is the recovery real?” — it was “how much is panic-buying?”
Peer cross-check
Six analog peers have reported into the CY2026 cycle. Across five signals — on-book DIO trend, distributor channel weeks, fab utilization, lead-time language, and any pull-in acknowledgment — the tape is directionally uniform.
Cell color: green = bullish signal, amber = mixed / strategic build, grey = not disclosed. ON's DIO shown as 201-day headline / 126-day base ex-strategic.
Two reads: (i) On-book DIO is elevated everywhere (>160 days) but draining sequentially where it's working stock — TXN 196, −13 QoQ, MCHP 185, −16. ADI (168) and NXPI (165) are running small strategic prebuilds, not working inventory. ON's 201 headline strips to 126 ex-strategic. (ii) The channel is the cleaner customer-facing tell and is decisively lean — MCHP 26 days ("lower end of historical"), ADI 6.5 weeks and declining inside a 6-7 week target, STM "further decreased and now normalized," IFX flagging "supply chain inventory reaching low levels." ON at 10.8 weeks is the only flat print. No peer is reporting excess channel inventory.
Utilization and lead times
Channel drawdown alone could just mean distributors tightening working capital. Cross-check against how hard peers are running factories:
NXPI guides front-end utilization from low-80s H1 to mid-80s H2 — a step, not a drift.
ON took utilization from sub-70 to 77% mid-quarter to meet demand (each point ≈ 25-30 bps of GM).
Infineon describes 300mm fabs as “very, very high,” AI data-center on allocation, €25B backlog +25% YoY, orders booked “well into next fiscal year.”
MCHP is ramping every large fab; underutilization charges of $46.6M expected to shrink each quarter through FY27.
ADI at 73% GM (+180 bps seq), at/near peak utilization — any further upside requires outsourcing.
Lead-time language got explicit. MCHP’s Steve Sanghi, May 7: “in another quarter or so, there could be nothing available in 4-6 weeks.” ON’s Ganesh Moorthy, May 4: “some technologies already on allocation… lead times are starting to extend.”
STM’s pull-in denial
STM’s Jean-Marc Chery on the April 23 call: “Strong booking of Q1 has shown absolutely no pull-in order — it is a well-balanced loading of the 2026 quarter-to-quarter.” STM has the deepest European Tier-1 (Bosch, Continental, Denso, Aptiv) footprint in the group, so a clean denial there matters. The bull reconciliation: the channel is so under-inventoried that even normal restocking looks like scarcity — with true end-demand not yet hot.
The F2Q26 print
Revenue $5.46B, +23% YoY, +13% seq, ~4% above the $5.25B Street.
Gross margin 61.4%, +200 bps vs consensus, +550 bps off the F4Q25 trough of 55.9%.
EPS $2.14 (incl. $0.05 discrete tax benefit).
Q3 CY26 guide $5.65-6.15B (midpoint +8% seq — well above typical Q3 seasonality).
DIO landed in the bull-tell zone. CFO Rafael Lysecki: “Inventory was $4.6 billion, down $90 million from the prior quarter, and days were 196, down 13 days sequentially.” Draining 13 days while revenue grew 13% seq means production is at most in step with shipments.
End-markets confirmed — with Data Center accelerating rather than digesting:
Grouped bars: F1Q26 (prior-quarter) YoY growth vs F2Q26 (this print) YoY growth by end-market. Communications F2Q26 not quantitatively disclosed — management stated only that it "grew both YoY and sequentially.
Industrial +30% YoY, +10% seq, breadth across every sector and region — and Haviv Ilan flagged industrial is still 5-6 points below the 2022 peak, i.e., cyclical ceiling not in view.
Data Center roughly doubled YoY, +20% seq — an acceleration from Q1’s +90%, not the moderation the preview penciled in. Ilan: “strong demand… I don’t expect that to change in the foreseeable future.”
Automotive mid-teens YoY, up upper-single-digits seq — the seasonal inflection the preview was watching for, led by China EVs/hybrids.
Automotive pull-in and lead-time uptick
Two nuances:
Lead times ticked up. Ilan acknowledged TXN’s lead times were sub-13 weeks in Q2 but are now “a couple of weeks higher, simply because the demand is growing.” Small in absolute terms, but directionally a partial validation of Sanghi’s MCHP call.
Auto pull-in got acknowledged. Industrial denial was clean — “customers are early and are not yet building inventory.” Automotive was different: “our automotive customers have taken their inventory to very low levels. And now as there is a little bit more demand, they find themselves in a situation that is not sustainable. I think that also drove part of the demand.” Read literally: some Q2 auto demand was OEM/Tier-1 refill, not end-market pull. First explicit acknowledgment in the peer set that lean customer inventory pulled CY26 demand forward. Isolated to autos, framed as normalization rather than panic — but it’s the sentence to quote back at ON and NXPI on their prints.
What to watch
ON Q2 (early Aug) — does Moorthy adopt the “not sustainable” auto framing? Where does the 10.8-week channel go?
NXPI Q2 (late Jul) — mid-80s H2 utilization confirmation; any auto pull-in acknowledgment.
MCHP F1Q FY27 (early Aug) — did Sanghi’s “nothing available in 4-6 weeks” call land? Channel below 26 days?
TXN F3Q (late Oct) — does the lead-time uptick extend? Does the industrial denial hold?
IFX FQ4 (mid-Nov) — €25B backlog growth; AI allocation breadth.
ADI F4Q (late Nov) — does the outsourcing decision materialize?
Sell-side GM revisions — the 200 bps TXN beat, if mirrored by ON/NXPI/MCHP in the next four weeks, forces the revision cycle that has lagged peer earnings-call language by ~one quarter.
Bottom line. July 22 was a green light: revenue upside, 200 bps of GM upside, DIO draining, Data Center accelerating. The one honest yellow flag — Ilan’s auto pull-in language — is narrow and testable inside four weeks against ON and NXPI. If those prints don’t echo “not sustainable,” the pull-in question quiets and the cycle prices closer to a durable multi-quarter recovery than a scarcity spike.
That’s it, thanks for the read. Reach out to us at Portrait Research if you’re interested in what we’re building.
Sources
Texas Instruments Q1 2026 Earnings Call, Apr 23, 2026 — 209-day DIO baseline, +30% industrial, +90% data center
Texas Instruments Q2 2026 earnings release (8-K exhibit), Jul 22, 2026 — $5.46B revenue, 61.4% GM, $2.14 EPS, Q3 guide $5.65-6.15B
Texas Instruments Q2 2026 Earnings Call, Jul 22, 2026 — 196-day DIO, industrial +30% and 5-6 pts below 2022 peak, data center doubled, auto “not sustainable”, lead times “a couple of weeks higher”
Analog Devices Q2 2026 Earnings Call, May 20, 2026 — 73% GM, 168-day DIO, 6.5-week channel, near-peak utilization
NXP Semiconductors Q1 2026 Earnings Call, Apr 28, 2026 — front-end utilization low-80s H1 → mid-80s H2, 165-day DIO
Microchip Technology F4Q 2026 Earnings Call, May 7, 2026 — 185-day DIO, 26-day channel, $46.6M underutilization, “nothing available in 4-6 weeks”
ON Semiconductor Q1 2026 Earnings Call, May 4, 2026 — 201-day headline DIO (base 126), 10.8-week channel, 77% utilization, “some technologies already on allocation”
STMicroelectronics Q1 2026 Earnings Call, Apr 23, 2026 — channel “further decreased and now normalized”, “absolutely no pull-in order”
Infineon Q2 2026 Earnings Call, May 6, 2026 — 300mm fabs “very, very high”, AI data-center on allocation, €25B backlog +25% YoY
Portrait sell-side takeaways — Texas Instruments Q3 2025 print, Oct 2025 — the “shallow cycle” anchor now inverted
Portrait Texas Instruments F2Q26 earnings preview, Jul 21, 2026 — pre-print bull-tell baseline